Steve says "just wait." [Archive]

Icaras

Jun 27, 2010, 10:49 PM

Don't get excited, he may have been talking about the recent Mac Mini Update.;)

I think the Mini was a big part of Steve's statement, but it surely isn't just that. The mini is great, but not at all Apple's best selling desktop mac, let alone the entire Mac line. According to the Apple Online Store, the Mac Mini is 4th place in the top 10 selling list (which is followed by the Mac Pro in 5th place, which I admit is actually very questionable to begin with :eek:)

But I think we can all safely assume that the top 3 has merit. The Macbook, Macbook Pro, and iMacs are definitely Apple's best selling Mac computers. In fact I just visited my local SF store yesterday and those three machines are without a doubt, the most prominently displayed Mac computers in the store.

So I don't think the Mac Mini is everything Apple has up it's sleeve for the Mac. If we are to believe statistics, then we should be able to expect some more surprises for the other Macs as well, hopefully with the Mac Pro very much included. The recent Mini redesigns gives me much hope that Apple may be doing the same with the Mac Pro. In my opinion, I think we'll see some similar drastic changes in aesthetics to the Mac Pro, at the very least. I honestly didn't think of it very much of before, but after seeing the new Mac Mini, I am optimistic. The 7 year old design, as great and functional as it is, doesn't sit right with the rest of the Mac family anymore. I think 2009-2010 marks a significant and important relaunch of the Mac family, starting with the redesigned Macbook and iMacs from last year.

As glum as everyone on these forums tend to get about the demise of Apple's professional market and Mac OS, this is probably THE best time to really emphasize their computer line, as its very clear that their iDevices are producing that so called halo effect for Apple. I can speak out of experience as I've already converted many traditional PC users with iPod/iPhone into crossing over to Macs. And such, it takes really little effort, as they are the ones that usually approach me with the questions and the interest.

Hey, but that's my hunch. I just don't think the Mac Mini is all Steve told us to wait for. There's just got to be more to it than that.

Gregintosh

Jun 27, 2010, 10:57 PM

I really hope he wasn't talking about just the Mac Mini. Though as people place their orders today for 10 month old hardware, it allows Apple to make extra margins (as the cost of components has dropped, yet the savings will not be passed onto consumers).

I hope the iMacs are updated ASAP because I really need to upgrade mine. I will be ready to pull the trigger as soon as it happens on either a brand new one or a refurb or clearance existing one (depending on how great the upgrade is).

I've also noticed that the refurb store on Apple.com doesn't have as many iMacs as it used to about a week ago. In fact, the cheapest iMac on there now is $1299. Perhaps that's a hint that a refresh is imminent?

Red Comet

Jun 27, 2010, 11:42 PM

I would like Apple to throw in a decent stock gpu solution for the Mac Pro.

Bakafish

Jun 28, 2010, 12:32 AM

Light Peak is holding things up I think. Obviously CPU pricing and availability are affecting the new Pro's, but they could launch a new high end model now on the current architecture easily if they wanted to. I think that adding CPU doesn't really help Pro users that much at this point since there are more significant bottle necks in the current platform, mostly IO related. The trend is supporting GPU based CPU offload anyway, so hexacore-shmexacore.

What I'm guessing is that the new Pro's will focus on very high speed PCI, Memory and Light Peak throughput. Figure lots of space, power and data channels for GPU's with Light Peak being the interface that allows for NextGen networking, storage and Apples so far under exploited distributed processing capabilities. I think that the amount of onboard drive storage capacity will take a hit to give space for GPU cards and power and cooling, as NAS and SAN's are really coming down in price and there will be a big push from Apple and Intel to replace Firewire, FiberChannel, 10 Gigabit Ethernet and even Infiniband in a lot of applications.

I don't know if it will really catch on and be able to replace everything that it potentially could, but I expect that that will be the intention. The cost of high end network infrastructures is a large percentage of the overall cost of the potential customer base of the Pro machines and typically limits the usefulness and scalability of the product. It is expensive to even support Gigabit speeds for any reasonable number of machines and I can picture Intel really looking hard at getting into the networking game by bypassing the current status quo of Ethernet. Remember that TCP/IP tunnels over Firewire quite nicely, no reason that it won't over Light Peak. Fiber optic cables are just waiting for commodity style demand to make pricing reasonable and Intel is most certainly making dedicated routing hardware, so it will be interesting to see if they have some sort of IP bridging capability. Imagine a $250 box with 12 Light Peak ports and a 10 GB ethernet jack. That could save companies a lot of money and force Cisco to rethink their strategy.

zenjabba

Jun 28, 2010, 12:35 AM

. That could save companies a lot of money and force Cisco to rethink their strategy.

Ain't that the truth... 24 Ports 10GB LC $9k and we needed 4 of them.

wisty

Jun 28, 2010, 02:21 AM

DoFoT9

Jun 28, 2010, 02:25 AM

Light Peak is holding things up I think. Imagine a $250 box with 12 Light Peak ports and a 10 GB ethernet jack. That could save companies a lot of money and force Cisco to rethink their strategy.

seriously? where have you come form? LP is at least another year away from PCs, let a lone Macs.

10GBoE is also a long way away, and <1% of people who buy a MP would need it! so it wont happen either

Bakafish

Jun 28, 2010, 03:53 AM

seriously? where have you come form? LP is at least another year away from PCs, let a lone Macs.

10GBoE is also a long way away, and <1% of people who buy a MP would need it! so it wont happen either

Okay, well I have 10GBE in my closet right now, so consider it a reality :) I come from a place where a lot of my budget gets spent on Cisco and Juniper hardware that frankly isn't much more sophisticated than the stuff I was buying 10 years ago. They have very slow, low power intel CPU's and very simple copper based interfaces. The optical solutions are equally overpriced and pathetically outdated (duplex fiber? Seriously?) There is an opportunity to completely leapfrog over what the current 'state of the art' is, and I believe that's what they are planning to do here. And as for need, of course they need it, it's just too expensive for consumers right now (artificially so if you ask me.) You want to switch back to 100 megabit ethernet? How about 10 megabit? You can copy the same stuff, it just takes longer right? You might argue that the difference between a second and ten is not very great, but given the choice I'm sure most will go with what is faster, and that's what LP is promising. Remember it's starting at 10GB, and it will scale from there. Moving around terabytes of data, especially locally is becoming really common even for 'normal' people. Using cheap fiber cables (yes there is such a thing!) and low cost high volume components LP could potentially change how data is moved.

Also, you may recall that LP was developed in partnership with Apple (and may have actually been a 100% Apple project.) And it was demoed on Apple hardware by Intel, so I'm willing to bet we will see it there first. And as I said before probably on the next generation Pro's.

In the kind of settings where Apple wants to get some penetration, say Architecture firms, animation studios and so forth, local storage is used for scratch disk and everything is going over fast, and quite expensive networks. I know ProSumers like the idea of fast onboard raid, but it's really an awful place for it for professional companies. We want centralized storage where security, reliability, access, allocation, replication and snapshotting can all be enforced. We want solid state storage on the nodes for scratch disks, but everything else should live on the SAN. LP potentially alleviates this issue in at least workgroup units, and the core frameworks are already set up to do well parralelized workloads across CPU's and GPU's. if you had a very fast low latency fiber network in the mix, suddenly you are able to increase the utilization of all your machines. They are doing this with distributed builds, but I imagine they want to abstract this into ad-hock render farms and supercomputers. Sony and IBM originally envisioned a similar usage of the Cell processors used in the PS3 and so forth, but they lacked a fast enough transport and any real market need.

PaulD-UK

Jun 28, 2010, 04:00 AM

Hi
Whatever Apple's Mac division's forward thinking may be it hasn't stopped the current i5/i7 iMacs being designed with an inferior ethernet chip that doesn't support jumbo frames (the 27" core 2 duo model still does, as do earlier iMacs) - thus ensuring that the fastest iMacs are unusable in a pro SAN/NAS environment.

As I understand it the current Mac Pro's motherboard had an Intel ethernet chip (Hartwell) that they've long since discontinued, so either Apple bought a huge EOL stockpile, or some sort of ethernet mobo re-engineering has to be undertaken. (Maybe it already has, I have no way of finding out what ethernet chip is fitted to current Mac Pros).

Basically until Apple completes a 64-bit cocoa/Grand Central Dispatch rewrite of Final Cut Suite there is not much need for further Mac Pro development, except in I/O enhancement - I would guess that professionals running FCS is a huge driver of sales for Mac Pros.

The biggest motivation for Apple to re-engineer the Mac Pro line is Intel's chip supply strategy, particularly when existing production is swapped over to new product thereby forcing change.

Any enclosure design decisions made now that don't anticipate the potentially revolutionary changes that a full implementation of Light Peak will unleash will be overtaken by progress...

Overall I remain optimistic - except for that crass i7 iMac ethernet choice - but that may be deliberate to force professionals to only buy the Mac Pro...

Bakafish

Jun 28, 2010, 04:37 AM

And I had no idea it was using that crippled Broadcom chip :-/ I have some Dell 1U servers that have that same POS chipset in them which was rectified by installing PCI based interfaces. No such luck for our iMac's though. I use it at home where I don't have any compelling need for Jumbo's but I can imagine people being miffed.

ValSalva

Jun 28, 2010, 04:43 AM

I think the Mac Mini announcement was just part of the 2010 update plan SJ was referring to. I was convinced of a June 29th update to the Mac Pro but given the pattern of updates to Macs this year being essentially monthly I have to admit that July 13th is more likely. IMHO something Mac will probably be released monthly with the Mac Pro, iMac, or Air receiving updates. The order - who knows?

Vylen

Jun 28, 2010, 05:17 AM

Just wait... the new Mac Pro is currently passing through Steve Jobs' colon...

I hope that gave you all some terrible terrible imagery.

DoFoT9

Jun 28, 2010, 05:30 AM

Okay, well I have 10GBE in my closet right now, so consider it a reality :)
well there you go, you are in the <1% of people who need it. what exactly do you do? and why do you have servers in the closet lol?

The optical solutions are equally overpriced and pathetically outdated (duplex fiber? Seriously?) There is an opportunity to completely leapfrog over what the current 'state of the art' is, and I believe that's what they are planning to do here. And as for need, of course they need it, it's just too expensive for consumers right now (artificially so if you ask me.)
and how much do you think LP will be to start off with? im not so sure apple would be keen to include it, it could add $500+ to the machine costs - does that seem a reasonable number?

You want to switch back to 100 megabit ethernet? How about 10 megabit? You can copy the same stuff, it just takes longer right? You might argue that the difference between a second and ten is not very great, but given the choice I'm sure most will go with what is faster, and that's what LP is promising. Remember it's starting at 10GB, and it will scale from there. Moving around terabytes of data, especially locally is becoming really common even for 'normal' people. Using cheap fiber cables (yes there is such a thing!) and low cost high volume components LP could potentially change how data is moved.
of course i dont want to go backwards - but i disagree that even "normal" people have a need for 10GBoE. will "normal" people have a RAID cluster that is fast enough to support that? no. all they have are external USB or internal HDDs that cant go any faster then 1GBoE at this point in time.

Also, you may recall that LP was developed in partnership with Apple (and may have actually been a 100% Apple project.) And it was demoed on Apple hardware by Intel, so I'm willing to bet we will see it there first. And as I said before probably on the next generation Pro's.
thats true, so you might be right - apple might start off the trend, but not this update. i dont believe so.

can i ask, what do you do that requires GB/s of data throughput? run a data centre lol?

Hellhammer

Jun 28, 2010, 05:35 AM

can i ask, what do you do that requires GB/s of data throughput? run a data centre lol?

He's from Japan, they already have USB 4.0 and 1Tb/s Ethernet there :p 1Pb/s Ethernet is already knocking the door. 1Gb/s is so 1998 :D

DoFoT9

Jun 28, 2010, 05:40 AM

He's from Japan, they already have USB 4.0 and 1Tb/s Ethernet there :p 1Pb/s Ethernet is already knocking the door. 1Gb/s is so 1998 :D

thats true. us aussies are always left behind :(

xgman

Jun 28, 2010, 09:40 AM

Light Peak is holding things up .

.

You better hope not. Think 2011 at the soonest.

GiantDolphin

Jun 28, 2010, 10:07 AM

Just wait... the new Mac Pro is currently passing through Steve Jobs' colon...

I hope that gave you all some terrible terrible imagery.

that was a good laugh. Thanks.

gpzjock

Jun 28, 2010, 11:01 AM

Those who are frothing at the mouth over the late arrival of their next PC smashing behemoth should stop for a minute and consider the past.
As a lowly iMac user back in the late 90's I had to wait 4 years for Apple to supersede my Bondi's G3/CRT architecture with a G4 chipset and LCD screen while the PC users around me revelled in their tech bonanza. I waited another 3 years for a G5 iMac to replace that too.
The path of progress in Cupertino is littered with speed humps and potholes much like the long delays we suffer today waiting for even a modern design of ATI card....
The shift to Intel architecture allowed Apple to accelerate the CPU development road map for Mac Pro and iMac ranges but the wait before a fully formed rewrite of any model can stretch to years still.
I waited 7 years for USB 2.0 FFS! While my PC using mates asked what a Firewire port was.....
I am pretty philosophical about upgrading now, I replaced the G5 iMac with a 2008 Mac Pro 8 core beastie and full intend to keep it till the latest OS won't run on it anymore, about 2014 I reckon.
My future proofing is 16 GB of RAM and a PCI slot waiting for the next GFX card upgrade when ATI/Nvidia and Apple finally get of their collective arses.
I could throw my toys about PCI improvements and USB 3.0 but when connecting a 2 year old Mac Pro to BT's ageing and decrepit phone system hobbles anything internet based, I really can't see any point in demanding the pinnacle of tech to go with it.

To Do List:
Buy the next earthshattering Mac Pro version for stupid money.
Move to Singapore to make the best use of it. :D

cohen777

Jun 28, 2010, 11:30 AM

Not refreshing the Mac Pro for over a year is inexcusable. Additionally, since the release of Snow Leopard and Adobe Photoshop CS5, it is now time for G5 tower owners to switch. Apple had a golden opportunity at WWDC to introduce a new hexacore Mac.

I know and I understand that the iPhone, iPad, and iPod side of Apple is big business. I also think it is great for the Mac to have people using these devices because they will buy iMacs and Macbooks, and Mac Pro's too.

But, Apple core business is the flagship Mac Pro and Mac OS. Hopefully, Apple is working on Snow Leopard's replacement (Lion?) to remain competitive with Windows 7, and the new Mac Pro.

johnnymg

Jun 28, 2010, 01:10 PM

dolz

Jun 28, 2010, 01:40 PM

Not refreshing the Mac Pro for over a year is inexcusable. Additionally, since the release of Snow Leopard and Adobe Photoshop CS5, it is now time for G5 tower owners to switch. Apple had a golden opportunity at WWDC to introduce a new hexacore Mac.

That's the position I'm in. If the Mac Pro isn't updated 6/29 I'm buying an used Intel based tower and holding off a few more years before getting a new one.

gpzjock

Jun 28, 2010, 01:59 PM

Not refreshing the Mac Pro for over a year is inexcusable. Additionally, since the release of Snow Leopard and Adobe Photoshop CS5, it is now time for G5 tower owners to switch. Apple had a golden opportunity at WWDC to introduce a new hexacore Mac.

G5 PowerPC based Mac towers were produced from 2003-5 the last model is 5 years old now, Early G5 Power Mac owners shoulda replaced in 2008 if they wanted to stay current! If they are bleating about the lack of a hexcore now, why didn't they drop a log at the sight of the 2009 model with 16 core virtualisation?
Apple leave between 12 and 24 months between OS refreshes so why follow a shorter cycle for the hardware? By that roadmap overdue is early 2011. By staggering the releases of OS and top-end hardware they excite sales of each top flight model twice, it's all about the Benjamins!

Apple don't have the insanely competitive PC market force of other vendors making a Mac Pro to hurry them up. Dell and HP slice cents off every refresh build cost, to beat the other on profit margin and persue each others spec upgrades as close as they can. Meanwhile Supermarkets lay out Medion media towers on the cheap to catch the non techie market. It's all driven by Windows compatible competition, who makes a Mac OS X machine to compete and chivvy Steve along? Oh yeah, they sued their asses....

The iPhone refreshes faster because of the mass of smart phones appearing to steal it's crown, Phone-using Joe Public isn't as brand loyal to a phone OS and Apple know it. If you want a Mac OS running tower you got 1 vendor or a Hackintosh, they won't hurry to break that market up anytime soon.

toke lahti

Jun 28, 2010, 02:46 PM

Current MP has worst memory design of them all and it has been longest in market.
What does this mean?

Hellhammer

Jun 28, 2010, 03:05 PM

Current MP has worst memory design of them all and it has been longest in market.
What does this mean?

It's so good that Apple can't find a reason to update it :p Seriously, there are couple Macs in need of an update, Mac Pro and MBA. There are always some updates in autumn so lets keep our thumbs up for that! All Macs except Mini and maybe MacBook can be updated some time during the autumn.

deconstruct60

Jun 28, 2010, 04:10 PM

Light Peak is holding things up I think. .... as NAS and SAN's are really coming down in price and there will be a big push from Apple and Intel to replace Firewire, FiberChannel, 10 Gigabit Ethernet and even Infiniband in a lot of applications.
...
.... That could save companies a lot of money and force Cisco to rethink their strategy.

Eh? There is little to indicate that Light peak doesn't suffer from the same fundamental networking flaw that USB has. Namely, that it is a hub based model oriented toward hooking peripherals to a computer. There is a one central device and the communication is managed through just one. The demos have been of pushing video from one place to another (e.g., multiple digital video streams ) which in most devices just goes in one direction ( from player to monitor). Don't get lost that they are pushing the demo video between two computers. One of them is just a hub ( as in this video http://www.youtube.com/watch?v=nfGevFIVKw4&feature=channel).
One of the computers in these demos is always headless and the images are coming out of another computer which is hooked to the displays.

Most of the LP descriptions contain a laundry list of connector/buses that are supports to be transported. PCI , USB , SCSI , SATA, HDMI are all generally hub-spoke based. Folks have thrown in stuff like Ethernet and Firewire in some lists which aren't. However, those are typically hooked to PCI connections to the computer so could be mimicked by transporting the backend PCI channel to the host computer. Intel isn't a fan of Firewire anyway so I'd be surprised if they bothered to deal with it.

You can't do NAS and SAN with a hub-spoke network constraint. You can do direct attached storage (DAS). In the DAS market, Fiber Channel, 10GB Ethernet , and Infiniband don't make economic sense. They cost far too much to hook to just one machine in most situations.

If there is support for multiple hosts networks then Intel is keeping awfully quiet about it. Neither is Light Peak going to replace everything. There is going to be a subset of protocols that it targets for encoding/decoding so that can be transported over the new, incompatible protocol and then decoded on the other side. There is lots of hand waving that it will do anything. There is no creditable explanation for that.

So, Cisco is probably not all that scared. I suppose there is a lost opportunity for folks with NAS boxes that with low host connection counts ( 1-2 computers) who may now switch to DAS, but that was already there anyway (with eSATA, SAS , etc. ).

Lightpeak has the potential for very small shops ( 1-4 workstations) to buy/use high speed storage boxes. However, they will still run into walls when comes time to push those files between the boxes.

Unmanaged, 24 port, 1GbE switches are in the sub $400 range. It isn't that expensive to deploy for a small group. Even is step up to one with link aggregation (to get closer to 2GbE bandwidth) still only in the $1,300 range which is still around a $100/port-pair.

10GbE still is high, but the costs for that will drop over time also. 40GbE ( and 100GbE ) just past this month as standards. So 10GbE is no longer top of the heap. The costs will start to drop faster now. Once it isn't the "fastest" and is mature ( passed in 2002 and mainstream stuff from 2005-6 era ) the prices will drop back into the reasonable range.

Infiniband and 10GbE will likely kill off Fibre Channel. If anything Apple needs a solution for one of those two since they historically offered Fibre Channel on the high end. (would help also if they figured out that the iSCSI protocol existed. )

All these expectations that Light Peak is going to cure all problems is off.

ValSalva

Jun 28, 2010, 04:46 PM

I think we'll still be waiting come tomorrow. I think there is a high probability that something will be updated next month, approximately July 13th. Whether it's the Mac Pro that gets the next update is anyone's guess. However it seems to me there is less evidence/rumor that the Air will be updated soon other than it has been almost 13 months. I was really hoping for June 29th but think now it will be another "just wait" tomorrow.

nanofrog

Jun 28, 2010, 04:52 PM

Light Peak is holding things up I think.
Actually, it's not, as LP parts aren't even due to ship until Q4 2010, which means if on time, it won't appear in systems until 2011.

The new CPU's are meant to work in existing LGA1366 boards with a microcode update. It's cheaper for Intel and vendors to take this approach (chipset and resulting boards can run for 2 years, with only the CPU changing to improve system performance). In most cases, new tech, such as LP, USB 3.0, FW1600, SATA 6.0 Gb/s,... are expected to be added to the non-compliant boards via add-on cards.

Newer board designs will incorporate them when determined financially feasible (i.e. is it a budget board that relys solely on bare minimum part count = fewer features, or a high-end unit that additional semis are added for features prior to their inclusion by Intel in the chipset). A good example is USB 3.0 not being added to Intel's chipsets until 2012 (formally announced not too long ago).

I don't know if it will really catch on and be able to replace everything that it potentially could, but I expect that that will be the intention.
It's definitely intended to be capable of replacing a fair few interfaces, but we'll have to see what "shakes out" (most will result over cost).

The cost of high end network infrastructures is a large percentage of the overall cost of the potential customer base of the Pro machines and typically limits the usefulness and scalability of the product.
I'm not so sure on this, as I've the strong impression most are owned/operated by small shops (i.e. graphics design), and independents more than larger entities.

In the former group, they're less likely to be on such a high speed network, and run more as an independent (stand alone) system. The network might be there for a backup solution, if there's not a dedicated means per system. But comparing price for small scale solutions, the independent means is going to be cheaper.

For the latter group, yes, there are significant investments in networking equipment for the desired requirement. But this is shrinking for workstations from what I've seen. In these cases, it's more common to use clusters and then attach the workstation via say FC or Infiniband.

LP would be desirable here though, and is a big selling point IMO (makes small clusters a practical solution for SMB or independents with a high computing requirement).

It is expensive to even support Gigabit speeds for any reasonable number of machines and I can picture Intel really looking hard at getting into the networking game by bypassing the current status quo of Ethernet.
I agree. When I saw the video of LP's first demonstration, that's one of the things that immediately crossed my mind. This would give Intel a serious "leg up" over the competition.

Assuming LP hits their target pricing, it will be cheaper too per system (i.e. card and cable ~$100 per, not including the switch; not sure what the switch costs will be yet). But I suspect they're well aware that it needs to come in under 10G Ethernet to make it attractive. Assuming this happens, the enterprise community will bite.

...(duplex fiber? Seriously?)...
LP is actually Duplex as well (Up and Down are each 10Gb/s). It's just a lot thinner = smaller cable.

And as for need, of course they need it, it's just too expensive for consumers right now (artificially so if you ask me.) You want to switch back to 100 megabit Ethernet? How about 10 megabit? You can copy the same stuff, it just takes longer right? You might argue that the difference between a second and ten is not very great, but given the choice I'm sure most will go with what is faster, and that's what LP is promising. Remember it's starting at 10GB, and it will scale from there. Moving around terabytes of data, especially locally is becoming really common even for 'normal' people. Using cheap fiber cables (yes there is such a thing!) and low cost high volume components LP could potentially change how data is moved.
It's not necessary for consumers to actually need 10G yet, but that doesn't mean they're not interested either.

For consumers, the data rate is usually governed by their ISP throughput anyway (i.e. ISP signal distributed via a router to the other systems in the household). That's not to say they don't transfer large files between systems, but it's not critical that it be done within a fixed period (fast enough to transfer say 1TB/s, as you might need in an enterprise environment for nightly runs such as backup).

Also, you may recall that LP was developed in partnership with Apple (and may have actually been a 100% Apple project.) And it was demoed on Apple hardware by Intel, so I'm willing to bet we will see it there first. And as I said before probably on the next generation Pro's.
The hardware used in the demo wasn't Apple's, but appear to be Intel's own products and commodity components (i.e. PSU, CPU cooler with an LED fan,...). Apple's boards use proprietary connectors, even for the PSU (think of how HDD's and graphics cards get power on an Apple board).

Apple may have played a part (and I think they did, but their contribution is with software, which is why the system was running OS X), Intel is the driving force behind LP in conjunction with other partners. Apple would benefit from this type of arrangement, as they get OS X developed in time, with most of the bugs worked out at the time of release.

and how much do you think LP will be to start off with? I'm not so sure apple would be keen to include it, it could add $500+ to the machine costs - does that seem a reasonable number?
It won't be that expensive, as the parts are only expected to be ~$50USD per LP connection. I'm not sure of switch costs yet, as fewer units sold mean a higher R&D figure per unit sold in order to recoup those costs.

But it would be attractive for Apple as users would be better able to set up small clusters for render farms. As it's a cheaper way to go, that means more independents and small shops can implement such a solution. That means more system sales for Apple (rather than just one workstation per user). Those additional systems may be XServes (assuming they're willing to have a rack rather than just pedestal systems). Either way, more systems sold would increase Apple's sales figures.

of course i don't want to go backwards - but i disagree that even "normal" people have a need for 10GBoE. will "normal" people have a RAID cluster that is fast enough to support that? no. all they have are external USB or internal HDDs that cant go any faster then 1GBoE at this point in time.
Good point, as that much throughput means there must be another part of the system that can push data at such a rate. Ultimately, there's additional costs over the network, and it's all expensive. This means it's out of reach for the average consumer at this time.

Consumer products are all made with a single driving compromise; low cost.

PeterQVenkman

Jun 28, 2010, 04:55 PM

Roman23

Jun 28, 2010, 05:05 PM

2010 mac pro :(

Best bet possibly July-September.

That's the position I'm in. If the Mac Pro isn't updated 6/29 I'm buying an used Intel based tower and holding off a few more years before getting a new one.

ValSalva

Jun 28, 2010, 05:07 PM

At least "next month" is closer than before. ;)

You're absolutely right. How Zen. Each day that passes brings us closer to the new Mac Pro. I'll try to keep that in mind tomorrow morning when I keep refreshing the Apple store web page. :D

vixducis

Jun 28, 2010, 05:20 PM

Also, you may recall that LP was developed in partnership with Apple (and may have actually been a 100% Apple project.) And it was demoed on Apple hardware by Intel, so I'm willing to bet we will see it there first. And as I said before probably on the next generation Pro's.

Actually... it wasn't apple hardware at all. It was apple software hower aka hackintosh.

TennisandMusic

Jun 28, 2010, 07:06 PM

"And it was demoed on Apple hardware by Intel"...

Oh, the irony. :p

exodius

Jun 28, 2010, 08:33 PM

G5 PowerPC based Mac towers were produced from 2003-5 the last model is 5 years old now, Early G5 Power Mac owners shoulda replaced in 2008 if they wanted to stay current! If they are bleating about the lack of a hexcore now, why didn't they drop a log at the sight of the 2009 model with 16 core virtualisation?

Oh, I guarentee you, I was salivating over the idea of that computer. Heck, I still am! I'd buy one in a heartbeat if, well, I'm paying top dollar for a 15 month old product. Why didn't I buy one? I had the money. My Dual 2GHz G5 has been running pretty strong on 2 GB of RAM and 1.5 TB of harddrive space. Even Snow Leopard, which effectively made the machine obsolete, didn't really kill it. Aperture still runs, though loading in 600+ 21MP RAW files basically takes all night, and I can edit them faster than the previews are updated.

It's been six years since I got the computer and it's only now not really current anymore. (My PC using friends tend to start feeling their computers' age after three years, and have literally dumped them in the trash after 6.) And that's the thing. I buy the Mac Pro because it LASTS. I get the best bang out of my buck with it. Buying a 15 month old computer, especially for the same price as back then, is basically shooting myself in the foot.

As much as I know that a longer wait more or less correlates with better tech in the box, I am starting to feel the need to upgrade. I can build a hackintosh NOW. And I think I'm not the only one in this boat. We'd buy, but the product just isn't there. Steve Jobs probably does have something big coming up, but I can only wait so long before it's time to give up.

Roman23

Jun 28, 2010, 08:54 PM

Nano,

Still don't think based on everything you told me that the Mac Pro is going to make use of light peak..the imac and the laptops will get it before the big behemoth will.. By then the xserve may very well be EOL.

I simply don't believe the pro market is Apple's bread and butter anymore, as out of the ashes of APPLE COMPUTER, INC(1984-2006) arose a brand new, pro-sumer(consumer) empire catering to the laymen and not the professionals out there.. Its happening.. and its happening right now!

Personally, I believe firmly that the Macbook/Macbook Pro and all mobile computing devices are the future.. while the desktop is dying. By 2020, I think even the strongest laptop will play the toughest high res games better than even a desktop could.. Am I right?

I walked into Best Buy and saw many many laptops, and very very few desktops - signs that the desktop market is FALLING.

Actually, it's not, as LP parts aren't even due to ship until Q4 2010, which means if on time, it won't appear in systems until 2011.

The new CPU's are meant to work in existing LGA1366 boards with a microcode update. It's cheaper for Intel and vendors to take this approach (chipset and resulting boards can run for 2 years, with only the CPU changing to improve system performance). In most cases, new tech, such as LP, USB 3.0, FW1600, SATA 6.0 Gb/s,... are expected to be added to the non-compliant boards via add-on cards.

Newer board designs will incorporate them when determined financially feasible (i.e. is it a budget board that relys solely on bare minimum part count = fewer features, or a high-end unit that additional semis are added for features prior to their inclusion by Intel in the chipset). A good example is USB 3.0 not being added to Intel's chipsets until 2012 (formally announced not too long ago).

It's definitely intended to be capable of replacing a fair few interfaces, but we'll have to see what "shakes out" (most will result over cost).

I'm not so sure on this, as I've the strong impression most are owned/operated by small shops (i.e. graphics design), and independents more than larger entities.

In the former group, they're less likely to be on such a high speed network, and run more as an independent (stand alone) system. The network might be there for a backup solution, if there's not a dedicated means per system. But comparing price for small scale solutions, the independent means is going to be cheaper.

For the latter group, yes, there are significant investments in networking equipment for the desired requirement. But this is shrinking for workstations from what I've seen. In these cases, it's more common to use clusters and then attach the workstation via say FC or Infiniband.

LP would be desirable here though, and is a big selling point IMO (makes small clusters a practical solution for SMB or independents with a high computing requirement).

I agree. When I saw the video of LP's first demonstration, that's one of the things that immediately crossed my mind. This would give Intel a serious "leg up" over the competition.

Assuming LP hits their target pricing, it will be cheaper too per system (i.e. card and cable ~$100 per, not including the switch; not sure what the switch costs will be yet). But I suspect they're well aware that it needs to come in under 10G Ethernet to make it attractive. Assuming this happens, the enterprise community will bite.

LP is actually Duplex as well (Up and Down are each 10Gb/s). It's just a lot thinner = smaller cable.

It's not necessary for consumers to actually need 10G yet, but that doesn't mean they're not interested either.

For consumers, the data rate is usually governed by their ISP throughput anyway (i.e. ISP signal distributed via a router to the other systems in the household). That's not to say they don't transfer large files between systems, but it's not critical that it be done within a fixed period (fast enough to transfer say 1TB/s, as you might need in an enterprise environment for nightly runs such as backup).

The hardware used in the demo wasn't Apple's, but appear to be Intel's own products and commodity components (i.e. PSU, CPU cooler with an LED fan,...). Apple's boards use proprietary connectors, even for the PSU (think of how HDD's and graphics cards get power on an Apple board).

Apple may have played a part (and I think they did, but their contribution is with software, which is why the system was running OS X), Intel is the driving force behind LP in conjunction with other partners. Apple would benefit from this type of arrangement, as they get OS X developed in time, with most of the bugs worked out at the time of release.

It won't be that expensive, as the parts are only expected to be ~$50USD per LP connection. I'm not sure of switch costs yet, as fewer units sold mean a higher R&D figure per unit sold in order to recoup those costs.

But it would be attractive for Apple as users would be better able to set up small clusters for render farms. As it's a cheaper way to go, that means more independents and small shops can implement such a solution. That means more system sales for Apple (rather than just one workstation per user). Those additional systems may be XServes (assuming they're willing to have a rack rather than just pedestal systems). Either way, more systems sold would increase Apple's sales figures.

Good point, as that much throughput means there must be another part of the system that can push data at such a rate. Ultimately, there's additional costs over the network, and it's all expensive. This means it's out of reach for the average consumer at this time.

Consumer products are all made with a single driving compromise; low cost.

dolz

Jun 28, 2010, 09:39 PM

2010 mac pro :(

Best bet possibly July-September.

Nope. If the great and powerful Jobs doesn't deliver on 6/29 I'm done waiting. My first computer was an Apple 2e and I'm sticking with Apple, just going with a used one this time.

nanofrog

Jun 28, 2010, 09:40 PM

Chilla Frilla

Jun 28, 2010, 10:31 PM

Just wait... the new Mac Pro is currently passing through Steve Jobs' colon...

I hope that gave you all some terrible terrible imagery.

Perfect imagery. We need more predictions from this guy.

Bakafish

Jun 28, 2010, 10:33 PM

Eh? There is little to indicate that Light peak doesn't suffer from the same fundamental networking flaw that USB has. Namely, that it is a hub based model oriented toward hooking peripherals to a computer. There is a one central device and the communication is managed through just one. The demos have been of pushing video from one place to another (e.g., multiple digital video streams ) which in most devices just goes in one direction ( from player to monitor). Don't get lost that they are pushing the demo video between two computers. One of them is just a hub ( as in this video http://www.youtube.com/watch?v=nfGevFIVKw4&feature=channel).
One of the computers in these demos is always headless and the images are coming out of another computer which is hooked to the displays.

There is no evidence that this is the case, it is claimed to be fully bidirectional and able to encapsulate the FireWire protocol specifically which would lead one to believe this is not an issue.

Intel isn't a fan of Firewire anyway so I'd be surprised if they bothered to deal with it.

They specifically refer to it.

You can't do NAS and SAN with a hub-spoke network constraint.

Of course you can, you mean to say Master Slave arraignment. USB's limitation is that there is only ever one master and everything else is a dumb slave to the master controller. Again, this is unsubstantiated fear on your part, and highly unlikely to be how it is actually implemented.

You can do direct attached storage (DAS). In the DAS market, Fiber Channel, 10GB Ethernet , and Infiniband don't make economic sense. They cost far too much to hook to just one machine in most situations.

Which is exactly why I think Apple and Intel want to solve this issue with low cost LP.

If there is support for multiple hosts networks then Intel is keeping awfully quiet about it. Neither is Light Peak going to replace everything. There is going to be a subset of protocols that it targets for encoding/decoding so that can be transported over the new, incompatible protocol and then decoded on the other side. There is lots of hand waving that it will do anything. There is no creditable explanation for that.

They clearly intend for this to be a low level transport encapsulating other protocols. Many things that once were done in hardware over proprietary busses can now be done in software over standard transports. iSCSI, Fiberchannel over Ethernet (FoE), and so forth. Where iSCSI was once done in hardware using expensive initiators, now we have open source software stacks that can outperform most of the dedicated hardware that was formally used.

So, Cisco is probably not all that scared. I suppose there is a lost opportunity for folks with NAS boxes that with low host connection counts ( 1-2 computers) who may now switch to DAS, but that was already there anyway (with eSATA, SAS , etc. ).

If low cost high speed switches come out that allow people to create networks it will fundamentally impact Cisco's core business. They are in the best position to capitalize on this, so I don't think they are scared, I'm just saying that people are willing to compromise a lot of functionality to get 90% of what they need at 15% of what it would normally cost. Building out a small 10G copper network right now is really, really expensive. There are a lot of professionals who would do it right now if they could afford it, even if it had some limitations. If TCP/IP can be encapsulated (and I can almost guarantee it can as there is already done over Firewire) then all bets are off.

Lightpeak has the potential for very small shops ( 1-4 workstations) to buy/use high speed storage boxes. However, they will still run into walls when comes time to push those files between the boxes.

That's exactly where it will start, and I think describes the target market for Pro's right?

Unmanaged, 24 port, 1GbE switches are in the sub $400 range. It isn't that expensive to deploy for a small group. Even is step up to one with link aggregation (to get closer to 2GbE bandwidth) still only in the $1,300 range which is still around a $100/port-pair.

Yeah, I wish aggregation worked that well. But regardless, that's still a lot of money, and it is much slower than 10G which is what we are talking about. As I said, I use 10G here at my office, I meant my equipment closet, not my house ;-) and it makes a big difference. But it is by no means plug and play. The Cisco gear I use needs lots of configuration and interconnectivity between hardware is finicky and there is lots of TCP stack tuning. It's not at all ProSumer friendly. And it's really really expensive.

10GbE still is high, but the costs for that will drop over time also. 40GbE ( and 100GbE ) just past this month as standards. So 10GbE is no longer top of the heap. The costs will start to drop faster now. Once it isn't the "fastest" and is mature ( passed in 2002 and mainstream stuff from 2005-6 era ) the prices will drop back into the reasonable range.

Actually this is what bugs me. The pricing has not dropped in any significant way. 10G has been around for a while and the companies have a lot of reasons to keep the pricing high. Smaller companies can't undercut pricing since there are a ton of patents around getting copper to do the unnatural things required to move that much data. And Ethernet itself is basically a stupid way to move this kind of data since it is inherently lossy (packets are droppable except with 10GbE!) The barrier to entry is high, the costs are high, the demand isn't there because it's not consumer friendly or priced right. Sound like an opportunity?

Infiniband and 10GbE will likely kill off Fibre Channel. If anything Apple needs a solution for one of those two since they historically offered Fibre Channel on the high end. (would help also if they figured out that the iSCSI protocol existed. )

FiberChannel is actually two things, a protocol and transport layer. They are easily separated, and if you can guarantee reliability (no dropping of packets) you can transport it happily over other layers like lossless 10GbE. Ethernet is an ancient transport designed for different times. LP has the opportunity to not replicate some of the design compromises they had to make.

All these expectations that Light Peak is going to cure all problems is off.

A man can dream can't he? :D

Assuming LP hits their target pricing, it will be cheaper too per system (i.e. card and cable ~$100 per, not including the switch; not sure what the switch costs will be yet). But I suspect they're well aware that it needs to come in under 10G Ethernet to make it attractive. Assuming this happens, the enterprise community will bite.

Yeah, the thing about Ethernet switches being so expensive is that the protocol itself is so freaking patent encumbered and complex. Speed solves lots of issues, and a simple addressing scheme, commodity priced interfaces and a naturally robust, low-noise, lossless and high speed transport all ad up to cheap switches.

LP is actually Duplex as well (Up and Down are each 10Gb/s). It's just a lot thinner = smaller cable.

Yeah, actually I missed the fact they are using two fibers, but at least it's in a single cable :-)

The hardware used in the demo wasn't Apple's, but appear to be Intel's own products and commodity components (i.e. PSU, CPU cooler with an LED fan,...). Apple's boards use proprietary connectors, even for the PSU (think of how HDD's and graphics cards get power on an Apple board).

Actually... it wasn't apple hardware at all. It was apple software hower aka hackintosh.

It was reported to be an Apple prototype board, so that's what I was saying. It certainly was running OS X and Intel often is contracted to do hardware designs for Apple, so I don't know. I think it's not really that interesting as there is very little unique about Apple hardware since they use standard chipsets and so forth. My main point was to refute an earlier post that Apple would somehow be at least year behind in deploying LP, I contend they will be first.

Nano,

Still don't think based on everything you told me that the Mac Pro is going to make use of light peak..the imac and the laptops will get it before the big behemoth will.. By then the xserve may very well be EOL.

I simply don't believe the pro market is Apple's bread and butter anymore, as out of the ashes of APPLE COMPUTER, INC(1984-2006) arose a brand new, pro-sumer(consumer) empire catering to the laymen and not the professionals out there.. Its happening.. and its happening right now!

Personally, I believe firmly that the Macbook/Macbook Pro and all mobile computing devices are the future.. while the desktop is dying. By 2020, I think even the strongest laptop will play the toughest high res games better than even a desktop could.. Am I right?

I walked into Best Buy and saw many many laptops, and very very few desktops - signs that the desktop market is FALLING.

Actually, I was the one predicting LP being the issue. And saying that desktops are dying based on what is stocked at best buy is like saying that whales are extinct on the basis of what you see in your bathtub. Okay, that was a little harsh :o Basically consumers think that they want laptops. They of course usually leave them on their desks and never actually run them off of the batteries, but that just consumers for you. They dream of blogging at the coffee shop, envy of everyone around. Then they try it once, realize they look like a dork, can't see the screen and can't actually get anything done or drink their coffee and never move the stupid thing again. Okay, rant over.

Desktops and more importantly "workstations" are bought be the corporate world and delivered from Dell and Apple by UPS. We don't go down to Best Buy...

nanofrog

Jun 28, 2010, 11:34 PM

If low cost high speed switches come out that allow people to create networks it will fundamentally impact Cisco's core business. They are in the best position to capitalize on this, so I don't think they are scared, I'm just saying that people are willing to compromise a lot of functionality to get 90% of what they need at 15% of what it would normally cost. Building out a small 10G copper network right now is really, really expensive. There are a lot of professionals who would do it right now if they could afford it, even if it had some limitations. If TCP/IP can be encapsulated (and I can almost guarantee it can as there is already done over Firewire) then all bets are off.
I agree. Cisco's business depends primarily on the enterprise market, which is all that can generally afford 10G Ethernet gear. It's still pricey, despite the length of time it's been available.

That's exactly where it will start, and I think describes the target market for Pro's right?
This market segment (small business/independents) is where a substantial amount of potential exists IMO; from inexpensive connections to high speed storage (DAS) to connecting nodes in a small cluster (makes it financially viable, as the networking gear can cost more than the systems used).

But it is by no means plug and play. The Cisco gear I use needs lots of configuration and interconnectivity between hardware is finicky and there is lots of TCP stack tuning. It's not at all ProSumer friendly. And it's really really expensive.
Cisco's interface is ancient (throwback to the '80's), and definitely not easy to deal with for most users.

Actually this is what bugs me. The pricing has not dropped in any significant way. 10G has been around for a while and the companies have a lot of reasons to keep the pricing high. Smaller companies can't undercut pricing since there are a ton of patents around getting copper to do the unnatural things required to move that much data. And Ethernet itself is basically a stupid way to move this kind of data since it is inherently lossy (packets are droppable except with 10GbE!) The barrier to entry is high, the costs are high, the demand isn't there because it's not consumer friendly or priced right. Sound like an opportunity?
Technology wise, we've reached the limit of copper. Optical is the only other way to go faster (and reliability will also improve, so long as the transceivers are designed properly). If the conversion (optical to electrical) is jittery for example, it won't work worth a crap (reliability suffers significantly, no matter the protocol used).

Yeah, the thing about Ethernet switches being so expensive is that the protocol itself is so freaking patent encumbered and complex. Speed solves lots of issues, and a simple addressing scheme, commodity priced interfaces and a naturally robust, low-noise, lossless and high speed transport all ad up to cheap switches.
If LP performs as advertised, Intel will have a "gold mine" on their hands. :D And they know it. :p

Yeah, actually I missed the fact they are using two fibers, but at least it's in a single cable :-)
It's also more flexible, allowing for tighter bends without signal loss. :) Despite it's small diameter, it's fairly robust as well as I understand it (not had my hands on one yet though).

It was reported to be an Apple prototype board, so that's what I was saying. It certainly was running OS X and Intel often is contracted to do hardware designs for Apple, so I don't know. I think it's not really that interesting as there is very little unique about Apple hardware since they use standard chipsets and so forth. My main point was to refute an earlier post that Apple would somehow be at least year behind in deploying LP, I contend they will be first.
The Evaluation Board (LP)?
Or the main board?

I ask, as the Evaluation board actually had a small bit of breadboard attached to it (small part count; looked like it might be a filter).

Basically consumers think that they want laptops. They of course usually leave them on their desks and never actually run them off of the batteries, but that just consumers for you.
I've had the same rant, so you're not alone. :p

VirtualRain

Jun 29, 2010, 12:47 AM

Nope. It the great and powerful Jobs doesn't deliver on 6/29 I'm done waiting. My first computer was an Apple 2e and I'm sticking with Apple, just going with a used one this time.

Another Apple IIe? :p That was a wicked machine in it's time... I still have mine in the garage... if they don't update the Mac Pro, I'm gonna go fire it up and do some rendering on it! :D

Just wait... the new Mac Pro is currently passing through Steve Jobs' colon...

I hope that gave you all some terrible terrible imagery.

:eek: I'm betting on a case redesign then... much more streamlined! :p :D

deconstruct60

Jun 29, 2010, 02:10 AM

There is no evidence that this is the case, it is claimed to be fully bidirectional and able to encapsulate the FireWire protocol specifically which would lead one to believe this is not an issue.

Bidirectional doesn't say anything about topology. Just means traffic flows in both directions along the arcs of whatever the topology is. For instance, USB 3.0 is full duplex over two signaling pairs..... just like LP. [ In fact, unclear just how much of LP is just recycled USB 3.0 optical. Not 100% same, but unclear just how different. ]

There is also nothing on intel site about encapsulating Firewire that I could find. Allusions to being the "one ring that rules them all" capability, but that is just hand waving. Most of the talk about it sweeping up Firewire are from the commentary not the publicly available technical docs. Conceptually, it could if added it to the protocol and incorporated the coder/decoders into the hardware. However, that just drives up costs. The notion that it "does everything for free" smells like somewhat selling "free lunches". I think Light Peak will do a strict subset of protocols. They just don't want to say what the subset is right now because folks will run out in fill in the blanks with all kinds of stuff. That adds a "magical" quality to Light Peak. (as oppose to being yet another new protocol which isn't compatible at the link level with anything right now. )

They specifically refer to it.

where? I'm talking official Intel docs/statements, not the interpretive articles written about it.

Of course you can, you mean to say Master Slave arraignment. USB's limitation is that there is only ever one master and everything else is a dumb slave to the master controller. Again, this is unsubstantiated fear on your part, and highly unlikely to be how it is actually implemented.

OK we're using different terms. USB and SATA are cheaper than Firewire and SAS in part because they dump the concept of each device on the network being an equal peer. Since LP is an attempt to do "more" for less money and on smaller devices, would not be surprising if it dumps elements that Ethernet/Firewire add topological flexibility. Historically, this is what has been done which doesn't make it highly unlikely. To put a fully capable master controller into every single network node will mean a network that has a higher cost.

They clearly intend for this to be a low level transport encapsulating other protocols. Many things that once were done in hardware over proprietary busses can now be done in software over standard transports. iSCSI, Fiberchannel over Ethernet (FoE), and so forth.

Not as efficiently. What 10GbE allows you to do is blow away lots of overhead to get back to similar speeds as the older protocols were delivering. The bigger bang for the buck is now only have to buy one set of networking gear.

Where iSCSI was once done in hardware using expensive initiators, now we have open source software stacks that can outperform most of the dedicated hardware that was formally used.

Again... what is the CPU utilization with this 100% software stack ?

Building out a small 10G copper network right now is really, really expensive.

So go fiber. Most of the same optical tech for LP could be applied to 10G/40G/100G Ethernet too without any of the Light Peak overhead. (For example 10GBASE-SR uses VCSEL lasers just like light peak on slightly narrower fiber. ) Going over 5Gbs is generally going to turn most devices from being FCC Class B ( ok for home use) into Class A ( not generally OK for home use... noisy clanky radios in part).

Light Peak is useful because many of the legacy (as of now) protocols for connection are stuck at a speed limit. Therefore, can use increases in speed to do cross box transport without introducing noticeable latency if can multiplex/demultiplex with very low overhead.

Actually this is what bugs me. The pricing has not dropped in any significant way.

There has been no faster standard so no pressure. 10G is toooo slow for you ... go to buy 40G. Oh there isn't any 40G. The cycles are about the same as for 10 -> 100 -> 1000 -> 10000 MBps 5-7 years to full maturation.

And Ethernet itself is basically a stupid way to move this kind of data since it is inherently lossy (packets are droppable except with 10GbE!)

Ethernet & TCP/IP is designed as a long distance transmission mechanism. Of course it has to deal with lossy.

The barrier to entry is high, the costs are high, the demand isn't there because it's not consumer friendly or priced right. Sound like an opportunity?

The barrier to entry is going down. The costs are coming down, just apply observation to what has happened to the 1G connector/switch market over last 4 years. Who can't ship a 1Gb switch right now if they want to. Can get a Netgear home router with 1Gb on it for $99. When 1G was top of the heap? No way.

Unmanaged switches are plug-and-play. Much of tweaking and fussing with switches is because want to do something custom. That won't change with light peak if want to do custom and fault tolerant topologies. In fact, won't be very surprising is light peak is simpler because you only get a simple single root tree, daisy chain network. Won't be any switches. Just repeater nodes (hubs) and some lightweight routing.

Ethernet is an ancient transport designed for different times.

Ethernet means you don't have to have a switch. There has not been any switches in the light peak demos either.

It was reported to be an Apple prototype board,

There were no Apple markings on the board. People assumed it was an Apple board primarily because it was running Mac OS X. I'd bet Intel can legally run an R&D hackintosh if they want to through an R&D agreement between the companies. That is the easiest way to allow Intel to boot up the OS on early CPU prototypes if they want to.

Vylen

Jun 29, 2010, 04:28 AM

:eek: I'm betting on a case redesign then... much more streamlined! :p :D

Or lumpy and irregular... he hasn't been healthy of recent.

Let the imagery continue!

Icaras

Jun 29, 2010, 04:43 AM

I must say, this is a thread of epic multi-quote proportions.

toke lahti

Jun 29, 2010, 04:51 AM

Lightpeak has the potential for very small shops ( 1-4 workstations) to buy/use high speed storage boxes. However, they will still run into walls when comes time to push those files between the boxes.
Not to mention that NASes that fit small shop's budget, can't even saturate 1G-ethernet in raid5. With link aggregation 1G will be fast enough for many years still...

ValSalva

Jun 29, 2010, 04:59 AM

Eh, it's about 6:00 AM EST and the store is still open. I don't think it will be this Tuesday :( More waiting.

toke lahti

Jun 29, 2010, 05:09 AM

In most cases, new tech, such as ...FW1600...
Is there any chipset supporting fw1600 even announced to be planned?

Consumer products are all made with a single driving compromise; low cost.
This is why usb3 will be the most cost effective (speed/money). Only thing Apple can do for pushing LP to consumer market is by not putting usb3 to new Macs. This will be dead end because the price of LP hubs & peripherals would scare consumers away from Macs.

For "pro" Macs there has to be reason why Apple won't give eSata. This might be, that thtey don't want pros to buy a lot of eSata hardware when Apple introduces the next fast connection. But will that be usb3, fw1600 or LP?
At least when most of new peripherals are usb3 next year, they'll have to add that to Macs or start loosing customers. Will there be only usb3 or usb3+fw1600 or usb3+LP?
Or will they neglect usb3 like blu-ray?
Majority of Mac users can't buy hd-movies from iTunes or don't want because the better PQ in BD, but Apple don't care. They just sell more iPhones & iPads. All others will use usb3 & LP is too expensive for consumers. Apple will just sell more iPhones & iPads?

Hopefully, AFAIK, crossgrading CS is still free...

DoFoT9

Jun 29, 2010, 06:02 AM

It won't be that expensive, as the parts are only expected to be ~$50USD per LP connection. I'm not sure of switch costs yet, as fewer units sold mean a higher R&D figure per unit sold in order to recoup those costs.
ok thats interesting to note. that seems relatively cheap and very reasonable given the amount of R&D that would have been invested! though it has been a joint-development, more money to throw around.

But it would be attractive for Apple as users would be better able to set up small clusters for render farms. As it's a cheaper way to go, that means more independents and small shops can implement such a solution. That means more system sales for Apple (rather than just one workstation per user). Those additional systems may be XServes (assuming they're willing to have a rack rather than just pedestal systems). Either way, more systems sold would increase Apple's sales figures.
so LP can be utilised for things such as networking and clustering as well? what will happen to technologies such as ethernet (cables) and Infiniband? for XServer i can definitely see LP as an early entrant, but for the MP and other consumables from Apple i just dont think they will introduce it very early on - in saying that, FW was introduced early!

Good point, as that much throughput means there must be another part of the system that can push data at such a rate. Ultimately, there's additional costs over the network, and it's all expensive. This means it's out of reach for the average consumer at this time.
yes without doubt. the bottlenecks are the hard drives - SSDs are slowly bridging the gap and enterprise markets are slowly converting to them for energy savings etc, but the storage capabilities are just too limiting!! sure, you can buy a 2TB SSD but at $~1,500USD a pop who can afford that!?? :rolleyes:

Consumer products are all made with a single driving compromise; low cost.
which = crap/limited performance. LP has virtually unlimited potential for bandwidth, but given the very nature of consumables they may be forced to limit it to 10Gb/s to begin with.

"And it was demoed on Apple hardware by Intel"...

Oh, the irony. :p
they also demoed it using USB as the peripheral connectors :rolleyes:

It will be able to benefit the iMac line, and the inclusion of LP could allow them to consider it as a replacement for the MP at the juncture described above (Apple does determine to EOL the MP).
how could it benefit the iMac line? for things such as video importing or something similar?

There has been no faster standard so no pressure. 10G is toooo slow for you ... go to buy 40G. Oh there isn't any 40G.
actually, i believe there to be 100Gb/s ethernet adaptors capable of fitting in laptop consumable devices - they came at a price of around $30k of course. but they do exist. cannot find the link i used to have :(

Ethernet & TCP/IP is designed as a long distance transmission mechanism. Of course it has to deal with lossy.
thank god for message integrity eh? ;)

For "pro" Macs there has to be reason why Apple won't give eSata. This might be, that thtey don't want pros to buy a lot of eSata hardware when Apple introduces the next fast connection. But will that be usb3, fw1600 or LP?
that would make sense. weaning away the competition as per usual.

Or will they neglect usb3 like blu-ray?
i wouldnt say neglect. apple supports BD discs, we can burn them with DVD SP, copy them if we want - its just the reading of commercial encrypted discs that isnt supported.

Majority of Mac users can't buy hd-movies from iTunes or don't want because the better PQ in BD, but Apple don't care. They just sell more iPhones & iPads. All others will use usb3 & LP is too expensive for consumers. Apple will just sell more iPhones & iPads?
idk about you - but the majority of people that i have talked to consider the Apple "HD" movies to be very clear and crisp! they dont know any better and frankly i dont think 90% of them are fussed. us pro-sumers might be, but we are different ;)

dolz

Jun 29, 2010, 06:59 AM

Another Apple IIe? :p That was a wicked machine in it's time... I still have mine in the garage... if they don't update the Mac Pro, I'm gonna go fire it up and do some rendering on it! :D

Yeah, mine screamed with 48K of memory. Bet you're done with the rendering before the next Mac Pro is released.:D

Mr. Anderson

Jun 29, 2010, 09:24 AM

I'm off on vacation at the end of the month of July with no internet - I'm betting it will go on sale that Tuesday just because that has been my luck on almost all my purchases of Macs in the past.

I have really, really bad luck - shipped MacPro stolen off the docks at FedEx (been tracking it all the way from Asia and then it just goes out for delivery and never arrives), Apple shipping me the wrong computer and I had to send it back before I get the one I asked for and of course then its back ordered... - all resulting in a month or more delay.

D

Vylen

Jun 29, 2010, 09:42 AM

Mr. Anderson

Jun 29, 2010, 10:06 AM

Clearly, you need to do more where things go wrong for you, but right for us :P

I'm doing what I can - sorry I didn't plan a vacation for next Tuesday :p

D

xgman

Jun 29, 2010, 10:07 AM

Clearly we are desperate. Hanging on to a vague "just wait"? I feel like an idiot. All hail the two word master. Sheesh, I wish there was another viable path here.

Vylen

Jun 29, 2010, 10:11 AM

Clearly we are desperate. Hanging on to a vague "just wait"? I feel like an idiot. All hail the two word master. Sheesh, I wish there was another viable path here.

Kidnap Steve Jobs?

Icaras

Jun 29, 2010, 12:13 PM

Clearly we are desperate. Hanging on to a vague "just wait"? I feel like an idiot. All hail the two word master. Sheesh, I wish there was another viable path here.

It's game over, man. Game over! :(

No but seriously, I would at least wait until the middle of July. It makes most sense if you consider the monthly Mac updates, and the iPhone 4 fever should finally take care of itself by then.

I say hang in there. It HAS to be just around the corner now. I'm going to keep waiting myself...big sigh.

deconstruct60

Jun 29, 2010, 02:20 PM

ok thats interesting to note. that seems relatively cheap

Should wait to find out how much CPU time these controllers burn up. If they need a core to drive them then the effective cost is higher.

how could it benefit the iMac line? for things such as video importing or something similar?

Currently, if need a 3-4 drive set up to get sufficient speed + storage space combination you are likely considering a Mac Pro since it can go 4-wide with no problems. if had 2 LP connectors on iMac could buy an external box to go 4 wide with negligible speed hit.

actually, i believe there to be 100Gb/s ethernet adaptors capable of fitting in laptop consumable devices - they came at a price of around $30k of course. but they do exist. cannot find the link i used to have :(

There might be a demo around but.
1. The standard wasn't finalized until June 17th. $30K for some "pre standard" card... only if you're desperate. That card probably only works best with the matching switch which would be as equally crazy high priced.

2. Would have to be a built in adapter. Need minimally a PCI-e 8x connection if not a 16x one. How many laptops have 2 16x (or 16x + 8x ) internal ? ExpressCard ... LOL. would be completely blown away by a 10G card let alone a 100G one.

However, an optical controller similar to the LP one could be used for a short range controller. Could get something the size of a drive into a laptop size case.

toke lahti

Jun 29, 2010, 03:08 PM

idk about you - but the majority of people that i have talked to consider the Apple "HD" movies to be very clear and crisp! they dont know any better and frankly i dont think 90% of them are fussed. us pro-sumers might be, but we are different ;)
Still majority of Mac users can't buy hd-movies from iTunes. They are offered in just few countries.

HurryKayne

Jun 29, 2010, 03:12 PM

"just wait"?
"Stay tuned" ( very "...funny")
Mr.Jobs..please....we are not sons of a telegraph.....
Make evolve Apple as you know,,,and as we love...but
please....don't lose us out sight,,, .
Don't kill the real Mac,,for the mobile ones.
:rolleyes:

DoFoT9

Jun 29, 2010, 04:03 PM

Should wait to find out how much CPU time these controllers burn up. If they need a core to drive them then the effective cost is higher.
they use up CPU time? i thought they would have their own controller (like FW) - if not, what an outrage.

Currently, if need a 3-4 drive set up to get sufficient speed + storage space combination you are likely considering a Mac Pro since it can go 4-wide with no problems. if had 2 LP connectors on iMac could buy an external box to go 4 wide with negligible speed hit.
fair point. seems logical that somebody might be wanting quad bays for their iMac - i know i do! too expensive for me, even the bad ones! but how on earth would you get a RAID controller? youd have to use the chip that comes with the box - which wouldnt be great from what ive seen. i guess anything can be had for a price.

There might be a demo around but.
it was for sale :)
1. The standard wasn't finalized until June 17th. $30K for some "pre standard" card... only if you're desperate. That card probably only works best with the matching switch which would be as equally crazy high priced.
i've no idea at this stage. i cant even find the bloody link :(

2. Would have to be a built in adapter. Need minimally a PCI-e 8x connection if not a 16x one. How many laptops have 2 16x (or 16x + 8x ) internal ? ExpressCard ... LOL. would be completely blown away by a 10G card let alone a 100G one.
lol@expresscards.. from what i remember they had a 100Gb adaptor that could fit into any standard laptop, the size - im not entirely sure dimension wise.

However, an optical controller similar to the LP one could be used for a short range controller. Could get something the size of a drive into a laptop size case.
thats true. wouldnt be overly cheap either by the sounds of it. i wonder how much latency is associated with the controllers themselves.

Still majority of Mac users can't buy hd-movies from iTunes. They are offered in just few countries.
thats a good point, and probably another topic ;) most countries have access to some sort of movie downloads that would be equivalent to what the iTunes store is (in quality/price).

nanofrog

Jun 29, 2010, 04:53 PM

Is there any chipset supporting fw1600 even announced to be planned?
Last I've seen, there's been no part production scheduled for S1600, but evaluation parts have been made (Symwave (http://www.symwave.com/), and I didn't even see mention of S1600 at all on the site). I think they may have abandoned it, but can't be 100% positive. S3200 also has evaluation parts out, but are scheduled for production late this year last I knew (Dap Technology (http://www.daptechnology.com/); there is an article on the site, but it's not been updated with newer information).

This is why usb3 will be the most cost effective (speed/money). Only thing Apple can do for pushing LP to consumer market is by not putting usb3 to new Macs. This will be dead end because the price of LP hubs & peripherals would scare consumers away from Macs.
LP has a basic topology in common with USB 3.0 (USB over fibre). As per the switch cost, we don't know yet, but I'd think it's not going to be that bad, or the sales will suffer, and adoption could actually stall.

Intel's pushed USB 3.0 support back in their chipsets to 2012, which is what most boards use for peripheral connections (lower cost). That would only leave higher-end boards that add separate semi's to support standards/features not available in the chipset (Intel based boards). This is meant to give LP a better chance at adoption IMO (makes sense, as it's up to Intel to determine if/when standards are supported in their own products), and from a business POV, they're going to give their own products the chance before others.

AMD products will be a bit different (i.e. chipset out earlier IIRC), as well as separate semi's to even beat that in terms of release date.

For "pro" Macs there has to be reason why Apple won't give eSata.
There is.

It's not available in the chipset (6 available, 4x used for the HDD bays, and 2x for the optical bays). So the only way to provide it, is to use a separate SATA controller chip from say Marvell for example.

Now the question is why don't they do this?
Answer = additional part cost and added complexity (further increases cost).

Remember, Apple has their MP boards ODM'ed (and all their gear is made by another company at a bare minimum, as some are designed by Apple, namely the iDevices). Intel's done most of it for the MP ('06 - '08 for sure), but Hon Hai Precision may have taken over on the 2009 systems (they make a lot of other products for Apple, as they're the primary ODM used from what I can tell).

This might be, that they don't want pros to buy a lot of eSata hardware when Apple introduces the next fast connection. But will that be usb3, fw1600 or LP?
LP has the greatest potential as it's not a fixed protocol. Meaning, you can run multiple protocols over it (including simultaneously).

CPU utilization may be it's Achilles Heel, depending on the specifics (i.e. protocol used and data rate).

At least when most of new peripherals are usb3 next year, they'll have to add that to Macs or start loosing customers. Will there be only usb3 or usb3+fw1600 or usb3+LP?
Or will they neglect usb3 like blu-ray?
In terms of USB 3.0, it won't be out next year in the MP, as Apple doesn't like to add in additional semiconductors for features. It's simpler and more importantly, cheaper to rely on the chipset, which Intel won't support until 2012.

Will we see FW1600/3200?
I don't know, as it will depend primarily on part cost. If it's cheap enough, they may update the FW chip to one that's compliant with the newer standard, as Apple's invested a lot of time and marketing in FW. But the cheapest way to continue FW support is with S800.

Majority of Mac users can't buy hd-movies from iTunes or don't want because the better PQ in BD, but Apple don't care. They just sell more iPhones & iPads. All others will use usb3 & LP is too expensive for consumers. Apple will just sell more iPhones & iPads?
1080P streamed isn't yet possible for many (vast majority aren't able to get 40Mb/s, which is the bare minimum for uncompressed signals), as their ISP connection is just too slow. When it happens, a compression scheme will almost certainly be used (I'd have a coronary if it isn't :p), but it won't the ISP connection requirement by all that much, as you need additional band to cover latency/retransmitted packets to prevent stuttering (buffering can help too, but it only goes so far, as most may not want to wait for 50% or so to download before viewing is started. So a 40Mb/s connection would still be where we need to be IMO to prevent issues in view-ability (downstream rate BTW).

ok thats interesting to note. that seems relatively cheap and very reasonable given the amount of R&D that would have been invested! though it has been a joint-development, more money to throw around.
LP's not an insignificant design in terms of funds required to develop it, but more importantly, Intel's not equipped to do it all on their own, as they don't concentrate in each area necessary to create it (i.e. fall short in areas of fibre optic cable design, lasers, optical transceivers).

so LP can be utilised for things such as networking and clustering as well? what will happen to technologies such as Ethernet (cables) and Infiniband?
Yes, LP could be used to connect up nodes and other high speed networks. Ethernet and Infiniband are already ingrained though, and will stick around for some time yet. Tech replacement never happens overnight, and LP may show some limits in areas, such as CPU utilization needed for translating the Ethernet protocol if the band is full. I see a bigger problem with multiple protocols used simultaneously though (LP is capable of this as well). Such limitations could cause older standards to remain, if they're:
1. cheaper
2. better suited to a dedicated task (there can be stipulations here too, as cost will always be a major factor)

the bottlenecks are the hard drives - SSDs are slowly bridging the gap and enterprise markets are slowly converting to them for energy savings etc, but the storage capabilities are just too limiting!!
I was thinking along the lines of large scale RAID systems and Flash drives (PCIe based) for massive drive throughputs, as it's way out of consumer budgets (or need) in the vast majority of cases.

Those that actually do, are workstation users or those that would be able to benefit from a small cluster for example (i.e. can saturate 10G Ethernet, FC, or even InfiniBand).

which = crap/limited performance. LP has virtually unlimited potential for bandwidth, but given the very nature of consumables they may be forced to limit it to 10Gb/s to begin with.
LP will initially release at 10Gb/s, and is stated that it can go to 100Gb/s. Past that, more work would have to be done (i.e. materials development).

how could it benefit the iMac line? for things such as video importing or something similar?
DAS (i.e. external RAID system that can push up to 1GB/s or so) and multiple monitor solutions come to mind. ;)

i wouldn't say neglect. apple supports BD discs, we can burn them with DVD SP, copy them if we want - its just the reading of commercial encrypted discs that isn't supported.

idk about you - but the majority of people that i have talked to consider the Apple "HD" movies to be very clear and crisp! they dont know any better and frankly i dont think 90% of them are fussed. us pro-sumers might be, but we are different ;)
720p is better than SD, especially on a larger set/monitor. Some of the earlier HD sets are only capable of 720p anyway (US anyway, and IIRC, it won't be much different Down Under). The only advantage we have in the US, is we can get the equipment cheaper from the price links you've shown in the past. :rolleyes:

Should wait to find out how much CPU time these controllers burn up. If they need a core to drive them then the effective cost is higher.
This is what I'm concerned over as well, especially if the band is near capacity or if multiple protocols are used simultaneously. I keep thinking of it as the proverbial "Other Shoe waiting to drop".

if had 2 LP connectors on iMac could buy an external box to go 4 wide with negligible speed hit.
Exactly.

This is why I'm wondering about the MP reaching EOL in the near future, say 2013 (i.e. if sales numbers are too low to continue, as LP in an iMac could allow it's usage in areas previously only possible in the MP).

DoFoT9

Jun 29, 2010, 05:20 PM

1080P streamed isn't yet possible for many (vast majority aren't able to get 40Mb/s, which is the bare minimum for uncompressed signals), as their ISP connection is just too slow. When it happens, a compression scheme will almost certainly be used (I'd have a coronary if it isn't :p), but it won't the ISP connection requirement by all that much, as you need additional band to cover latency/retransmitted packets to prevent stuttering (buffering can help too, but it only goes so far, as most may not want to wait for 50% or so to download before viewing is started. So a 40Mb/s connection would still be where we need to be IMO to prevent issues in view-ability (downstream rate BTW).
40Mb/s shall be possible for pretty much all aussies in a few years (if the government doesnt change anyway). $40bn invested to upgrade everybody to fibre (and those on copper will get VDSL2). so it could be possible if its hosted in australia, outside of that i dont think our underwater lines can cope.

LP's not an insignificant design in terms of funds required to develop it, but more importantly, Intel's not equipped to do it all on their own, as they don't concentrate in each area necessary to create it (i.e. fall short in areas of fibre optic cable design, lasers, optical transceivers).
oh i see. group effort then.

Yes, LP could be used to connect up nodes and other high speed networks. Ethernet and Infiniband are already ingrained though, and will stick around for some time yet. Tech replacement never happens overnight, and LP may show some limits in areas, such as CPU utilization needed for translating the Ethernet protocol if the band is full. I see a bigger problem with multiple protocols used simultaneously though (LP is capable of this as well). Such limitations could cause older standards to remain, if they're:
1. cheaper
2. better suited to a dedicated task (there can be stipulations here too, as cost will always be a major factor)
speaking of which - how will that implementation actually work using many protocols etc on the same band? i dont think its feasible to have a whole lot of protocols running over the same interface - unless they allow x amount of streams to be transmitting over parallel connections, especially where latency comes into it.

I was thinking along the lines of large scale RAID systems and Flash drives (PCIe based) for massive drive throughputs, as it's way out of consumer budgets (or need) in the vast majority of cases.
ahh true. 2,200MB/s for those PCIe cards isnt it? very fast indeed. given the first revision of LP it would still seem too limiting for 1 LP connection, so i take it that its possible to combine LP connections so that you have quad band LP, octo band LP - etc?

Those that actually do, are workstation users or those that would be able to benefit from a small cluster for example (i.e. can saturate 10G Ethernet, FC, or even InfiniBand).
and with ease. what is the insentive for users to invest in switching?

LP will initially release at 10Gb/s, and is stated that it can go to 100Gb/s. Past that, more work would have to be done (i.e. materials development).
yup i read 10Gb/s would be the starting point - which seems reasonably low, however given the "newness" of the technology its understandable that it will take some time to properly utilise all development technologies. i only hope that it is popular enough to be commerically available for us end users.

DAS (i.e. external RAID system that can push up to 1GB/s or so) and multiple monitor solutions come to mind. ;)
true - there would be many users wanting that for even the iMac or a MBP. as for DAS + multiple monitors on the same LP connector, how exactly does that work? or will there be one connection per interface port?

720p is better than SD, especially on a larger set/monitor. Some of the earlier HD sets are only capable of 720p anyway (US anyway, and IIRC, it won't be much different Down Under). The only advantage we have in the US, is we can get the equipment cheaper from the price links you've shown in the past. :rolleyes:
it all comes down to bitrate i.e. bandwidth. for hardware playback for TV equiptment and whatnot the prices have come down a lot here. im not entirely sure of the competitiveness compared to the US, given our import taxes and whatnot, but they are a lot cheaper. a resonable BD ROM can be had for $125Aus ($110 US?), and a medium speed BD writer for ~$200Aus. not sure about the players of late.

This is what I'm concerned over as well, especially if the band is near capacity or if multiple protocols are used simultaneously. I keep thinking of it as the proverbial "Other Shoe waiting to drop".
hmm. does this mean there are similar limitations that mirror that of USB? :\

nanofrog

Jun 29, 2010, 06:28 PM

speaking of which - how will that implementation actually work using many protocols etc on the same band? i dont think its feasible to have a whole lot of protocols running over the same interface - unless they allow x amount of streams to be transmitting over parallel connections, especially where latency comes into it.
It's a Quality of Service (QoS) method. You can look it up on wiki if you want.

what is the insentive for users to invest in switching?
Low cost performance is the primary reason, especially for those who've not been able to afford 10G Ethernet solutions. Less on switching, and more on bringing new users that can benefit to higher speed connections within their budget.

Think of an independent that's willing to build a small render farm (has need of it). They can afford the computers and software licensing involved, but they fall short on the connectivity for the nodes. LP can change that, and actually make it feasible for that person (or smaller shop).

Even for those that actually have 10G Ethernet (or other connections, such as FC or InfiniBand), it might be more cost effective to switch to LP when there's a major system/equipment upgrade planned. Think less $$$ per system for connectivity (10G E is expensive, and if a lower cost solution can be found, it will be persued when the opportunity arises).

yup i read 10Gb/s would be the starting point - which seems reasonably low, however given the "newness" of the technology its understandable that it will take some time to properly utilise all development technologies. i only hope that it is popular enough to be commerically available for us end users.
The biggest problem IMO for consumer users, will be with the availability of bridge chips for end-use devices (i.e. LP to SATA, LP to USB, LP to FireWire,...), as such devices are already using a fixed protocol. Without it, such devices can't be used.

as for DAS + multiple monitors on the same LP connector, how exactly does that work? or will there be one connection per interface port?
It's possible they can share a single LP port, but given that both of these uses are higher bandwidth connections, it could throttle (throughput of DAS and multiple monitors, could exceed 10Gb/s).

Ideally, there should be more than one LP port for this reason (say 2x, as it requires PCIe lanes to connect to the system), as that's the most attractive items to run on it IMO. Keep the USB and FW connections separate in the iMac, and the laptops as well if there's sufficient PCB space (especially the features included in the accompanying chipset for whatever CPU is used).

Anything requiring a separate chip could be sacrificed over cost constraints (lose it in favor of LP, such as FW, to help balance out the costs).

it all comes down to bitrate i.e. bandwidth. for hardware playback for TV equiptment and whatnot the prices have come down a lot here. im not entirely sure of the competitiveness compared to the US, given our import taxes and whatnot, but they are a lot cheaper. a resonable BD ROM can be had for $125Aus ($110 US?), and a medium speed BD writer for ~$200Aus. not sure about the players of late.
I was thinking only of separate equipment (stand-alone players and HD sets) to keep ISP issues out of it.

Prices are affected by taxes (which are higher for you), but also by population. According to Wiki, the estimated 2010 population is 22,401,525. For the US, it's 309,612,000 (also an estimate for 2010). Makes a big difference IMO (ye olde supply and demand bit).

hmm. does this mean there are similar limitations that mirror that of USB? :\
Yes and no. Topology wise, it does have it's basics in common (dedicated up and downstream signals). But where USB has a fixed protocol, LP does not. This can complicate matters in some respects, and is why Quality of Service was the approach taken. But to keep the costs low, there will be CPU utilization involved with translating protocols that doesn't exist with USB (that's the price involved for that degree of flexibility). And the necessitation of bridge chips for end-use devices as well, but that's added to that product, not the LP card or system that has it included on the main board.

deconstruct60

Jun 29, 2010, 10:48 PM

Intel's pushed USB 3.0 support back in their chipsets to 2012, which is what most boards use for peripheral connections (lower cost). That would only leave higher-end boards that add separate semi's to support standards/features not available in the chipset (Intel based boards). This is meant to give LP a better chance at adoption IMO (makes sense, as it's up to Intel to determine if/when standards are supported in their own products), and from a business POV, they're going to give their own products the chance before others.

But LP isn't a standard unless they get a qurom of folks to buy in. That means getting peripheral and other semi folks to buy in also. So far there are lots of Apple, Sony, Intel but not seeing much of a standardization effort. If Intel is the sole source supplier not sure this is going to get wide spread buy in. Especially if Intel is jacking up other stuff that is standardized in order to make head room for their proprietary stuff.

In terms of USB 3.0, it won't be out next year in the MP, as Apple doesn't like to add in additional semiconductors for features. It's simpler and more importantly, cheaper to rely on the chipset, which Intel won't support until 2012.

Firewire is sitting on a PCI-e connection. Apple could declare FW dead (like did with Flash ... old tech that had its time) and swap one of the USB 3.0 pci-e controllers. If do a swap there isn't much of an increase in board space or parts count. It isn't like Apple hasn't tried to kill off FW before on other Macs. Folks can also plug in a FW card into a PCI-e slot if essentially new one. Apple could sniff out that there were a couple of quality boards out there to fill the gap if that introduce it.

LP is likely even less likely than USB 3.0 to be put into the core chipsets before 2012. In 2011 there will be far more USB 3.0 devices to plug into than there will be LP devices. In fact have any devices (not computers ) been demo'ed? All seen so far are PCI-e cards and giant dongles to the standard interfaces.

Will we see FW1600/3200?
I don't know, as it will depend primarily on part cost. If it's cheap enough, they may update the FW chip to one that's compliant with the newer standard,

Again, a more telling sign is how many vendors have demo'ed peripheral prototypes with newer stuff. Can certainly see FW3200 for aerospace and embedded apps but that doesn't necessarly translate over to computers peripherals.

Apple slow rolled the adoption of FW800 on the Mac platform. It wasn't till recently that all macs went 800. The vast majority of PCs , if have it, are still stuck at FW400. Likewise cameras , etc. There is a smaller subset of devices that went 800 but there is little market pressure to push FW faster; otherwise all these 400 sockets wouldn't survive.

USB 3.0 has many of the historically differentiating features that FW offered (channel like, bidirectional connections, isochronous , speed , etc. ). That's got to put FW on Jobs' "old tech" hit list sooner or later. If throwing FW under the bus (so intel can run over it) gets Apple LP then I suspect Apple would go for that deal.

Intel's not equipped to do it all on their own, as they don't concentrate in each area necessary to create it (i.e. fall short in areas of fibre optic cable design, lasers, optical transceivers).

Errrr.... intel makes lots of Ethernet connector tech. Not sure how can be in the Ethernet connector business and not be looped into fiber optic cable, lasers , optical transceivers. If seriously going to be in the 10G , 40G, 100G Ethernet market have to have all that covered.

Yes, LP could be used to connect up nodes and other high speed networks. Ethernet and Infiniband are already ingrained though, and will stick around for some time yet. Tech replacement never happens overnight, and LP may show some limits in areas, such as CPU utilization needed for translating the Ethernet protocol if the band is full.

Not sure how you do Inifiband, IB, with some translation layer. The RMA latencies have to be low or you loose one of the major advantages of of IB. Not sure how going to bounce from LP transciever , trap up to CPU for decoding translating, then push to memory address block without hitting more latency than if natively doing IB. IB almost requires non blocking switch paths between the end points too. Again not so compatible with typical USB topologies.

Both Ethernet and IB have multiport reliability features too which are difficult to implement virtually.

I see a bigger problem with multiple protocols used simultaneously though (LP is capable of this as well). Such limitations could cause older standards to remain, if they're:
1. cheaper
2. better suited to a dedicated task (there can be stipulations here too, as cost will always be a major factor)

3. Get faster.

LP's advantage is that is substantially faster than what it is being aim at carrying/transporting. You can hide latency gaps and protocol overhead in that speed gap. As soon as you want to push to 80+% of LP bandwidth going to run into problems. A couple of PCI-e 1x links, some USB 2.0 , and a 720p video stream. Sure, basically a docking station for a laptop or mobile device.

If trying to run multiple higher speed, isochronous protocols ... much more likely going to run into trouble. You're also not going to get 10GbE , let alone 100GbE, out of it.

LP is much better at hunting older, slower game.

This is why I'm wondering about the MP reaching EOL in the near future, say 2013 (i.e. if sales numbers are too low to continue, as LP in an iMac could allow it's usage in areas previously only possible in the MP).

I think it much more depends upon the applications that come to the machine. The iMac is always going to be a more unbalanced box. If there is more software which scales up with resources then market should do OK. MP is more doom if it has been propped up by "high status symbol" and gaming users than need it to run a business users.

The XServe is probably in a much more precarious state. If it falls then the MP is next in line. If XServe stays and MBA falls then perhaps will have more cycles to put more value into the smaller line up.

Vylen

Jun 29, 2010, 11:08 PM

Wikipedia

According to Intel, the companies that will produce Light Peak technology include Foxconn, Foxlink, IPtronics[6][7], SAE Magnetics[8], FOCI Fiber Optics Communications Inc, Avago, Corning, Elaser, Oclaro, Ensphere Solutions[9], and Enablence.[10]

http://en.wikipedia.org/wiki/Lightpeak

Personally, with such long-winded discussions such as this, I take things people say with a grain of salt.

Cite sources please - even if it is something as dubious as Wikipeda ;)

nanofrog

Jun 30, 2010, 12:50 AM

But LP isn't a standard unless they get a quorum of folks to buy in. That means getting peripheral and other semi folks to buy in also. So far there are lots of Apple, Sony, Intel but not seeing much of a standardization effort. If Intel is the sole source supplier not sure this is going to get wide spread buy in. Especially if Intel is jacking up other stuff that is standardized in order to make head room for their proprietary stuff.
Actually, there's a fair number of partners, notably Manufacturing Partners listed by Vylen.

This much of an entwined infrastructure, LP definitely has the manufacturing (and design specification) set. Given the end-user product manufacturing from the likes of Hon Hai Precision, there's a serious chance it will show up. Ultimately setting the stage for a full-fledged adoption = specification is accepted as a standard (since LP's not been created by an organization like IEEE).

Afterall, Sony finally won out with BR, and is the "last man standing" in the HD disk format war. Nor was it purely accepted as the technologically superior specification. Their business planning and partnerships played a significant role (more the deciding factor from what I recall).

But even this much backing/support can fail if something falls short, notably the financial aspect (too expensive) or if the actual products can't deliver. If this ends up the case, it's just another attempt that ends up in the waste bin of history. There's definitely a few others sitting in that can somewhere. :eek: :p

But this one seems to have been fairly well planned out in terms of it becoming adopted and a recognized standard IMO. Now we need to see if it will actually deliver as expected.

Firewire is sitting on a PCI-e connection. Apple could declare FW dead (like did with Flash ... old tech that had its time) and swap one of the USB 3.0 pci-e controllers.
Additional parts can be cut out of a design over cost or "falling off" in terms of usability for what it's designed for compared to other specifications. Happens rather often in the world of computers, as I'm certain you're well aware. ;)

If do a swap there isn't much of an increase in board space or parts count. It isn't like Apple hasn't tried to kill off FW before on other Macs. Folks can also plug in a FW card into a PCI-e slot if essentially new one. Apple could sniff out that there were a couple of quality boards out there to fill the gap if that introduce it.
Where it can get tricky though, is in the Tick part of the cycle, as the PCB's have already been designed and in production for about a year or so. To fit newer parts, it's easy if the component package is identical, but in the case of USB 3.0, it's not (not even the same pin count). It would actually require a PCB redesign, which means more $$$. This makes it far less attractive for all but all out high-end boards that users would be willing to pay for. Apple's not in this type of mentality, going by their history with Intel parts.

It's far easier to include (add, or swap out) newer components in a Tock cycle, as you have to design a new board anyway.

Assuming LP is adopted by Apple, I would expect to see FW be eliminated, assuming the bridge chips aren't horrible in price, making LP to FW adapters prohibitive (I wouldn't think so, as it could be detrimental to adoption, but it is possible).

LP is likely even less likely than USB 3.0 to be put into the core chipsets before 2012. In 2011 there will be far more USB 3.0 devices to plug into than there will be LP devices. In fact have any devices (not computers ) been demo'ed? All seen so far are PCI-e cards and giant dongles to the standard interfaces.
No, the initial parts will be additional components attached to the PCIe lanes, not within the chipset. As per devices, I've not seen them, and it's one of the issues that concerns me (i.e. lack of bridge chips currently to demo products with, as well as the potential of unwanted/detrimental CPU utilization).

I'm not trying to indicate LP is an absolute perfect solution, but it certainly has promise so long as it actually delivers what's been promised. What's attractive to me, is the fact there's not a dedicated protocol allowing it to be used for a multitude of busses. But this is one fact that can make or break it.

Assuming it actually does work, I do expect full blown adoption/proliferation will take time, with adoption starting in the workstation market moreso than the server (i.e. small clusters).

That could help out USB 3.0 in the short term, especially for consumer systems, as existing peripheral devices will already work with it. But it has it's limitations too.

Again, a more telling sign is how many vendors have demo'ed peripheral prototypes with newer stuff. Can certainly see FW3200 for aerospace and embedded apps but that doesn't necessarily translate over to computers peripherals.

Apple slow rolled the adoption of FW800 on the Mac platform. It wasn't till recently that all macs went 800. The vast majority of PCs , if have it, are still stuck at FW400. Likewise cameras , etc. There is a smaller subset of devices that went 800 but there is little market pressure to push FW faster; otherwise all these 400 sockets wouldn't survive.

USB 3.0 has many of the historically differentiating features that FW offered (channel like, bidirectional connections, isochronous , speed , etc. ). That's got to put FW on Jobs' "old tech" hit list sooner or later. If throwing FW under the bus (so intel can run over it) gets Apple LP then I suspect Apple would go for that deal.
FW parts have historically been more expensive than others, and is another part of it's demise (i.e. FW HDD disk enclosures are more expensive than their USB counterparts). Performance does cost, but other interface technologies have caught up, and are cheaper. It's really nothing more than simple economics (and what the accountants look hard at). If FW is offered (generally speaking), the S400 spec is chosen because it's cheaper.

I wouldn't be surprised at all that Apple wants to dump it in favor of an economically viable replacement, especially if it also includes things like far faster throughputs and the ability to minimize connections (can allow for other interface chips to be tossed out, actually rendering LP the cheapest solution). The latter is more of an issue with laptops, but Apple does well in that particular market.

Errrr.... intel makes lots of Ethernet connector tech.
I was talking about the optical aspects necessary for LP (why I mentioned the lasers, transcievers,... = optical aspects of the standard), not other standards such as Ethernet.

Intel's great at making chips, but they would take a financial beating if they tried to create LP completely on their own. There's too much R&D in areas they've never worked in before. Which would translate into too much time and money spent on the project, and likely result in it's being scrapped before completion.

Not sure how you do InfiniBand, IB, with some translation layer. The RMA latencies have to be low or you loose one of the major advantages of of IB. Not sure how going to bounce from LP transceiver , trap up to CPU for decoding translating, then push to memory address block without hitting more latency than if natively doing IB. IB almost requires non blocking switch paths between the end points too. Again not so compatible with typical USB topologies.
What I meant with InfiniBand, wasn't via translation, but rather as a replacement (bonded LP ports) to be able to reach sufficient speeds.

Now I don't see it as a full replacement (unless LP's actually capable of being run for such a use without a protocol translation, which I haven't seen anything on), but it's viable for smaller clusters due to the cost IMO (i.e. those that would be sufficient if run with 10G E or FC for example).

3. Get faster.
I figured that was implied in #2. :eek: Guess not. :p

LP's advantage is that is substantially faster than what it is being aim at carrying/transporting. You can hide latency gaps and protocol overhead in that speed gap. As soon as you want to push to 80+% of LP bandwidth going to run into problems. A couple of PCI-e 1x links, some USB 2.0 , and a 720p video stream. Sure, basically a docking station for a laptop or mobile device.

If trying to run multiple higher speed, isochronous protocols ... much more likely going to run into trouble. You're also not going to get 10GbE , let alone 100GbE, out of it.
For a single line, no. Latency will prevent. But if you bond it, you can still exceed the band required (static, before accounting for latency), and still be able to achieve a target requirement (i.e. can get 10G E if bonded 2x 10Gb/s LP ports).

This is dependent on whether or not LP can actually be bonded, but I'd think that they thought this one through. If not, then it's really only suited to the consumer market afterall right now.

I think it much more depends upon the applications that come to the machine. The iMac is always going to be a more unbalanced box. If there is more software which scales up with resources then market should do OK. MP is more doom if it has been propped up by "high status symbol" and gaming users than need it to run a business users.

The XServe is probably in a much more precarious state. If it falls then the MP is next in line. If XServe stays and MBA falls then perhaps will have more cycles to put more value into the smaller line up.
Software will definitely have an impact in regard to the MP surviving (and XServe).

Unfortunately, the software always falls behind the software, and it's not helped by the necessitation for backwards compatibility. Worse, is that Apple hasn't finished the Cocoa/Carbon translation work yet, as developers like to wait for others to do as much of the work as possible (i.e. graphics related application suites). Slower cycles for professional applications also tends to push things back (i.e. 3 - 5yr rather than consumer software that may update annually). Then there's of course, software that can't benefit from SMP at all, such as a word processor.

I definitely agree the XServe is in more danger, and could negatively affect the MP if it goes (i.e. shared R&D only carried by the MP, pushing up prices, further reducing sales to the point of unsustainability).

Personally, I wouldn't be offended to see the MBA go away, as to free up time that can be used to shorten current development cycles (since they seem to be adding in new products, but not gone on a hiring binge to alleviate the current product refresh cycle). It just seems like too many projects, and too few people.

gpzjock

Jun 30, 2010, 03:53 AM

Current MP has worst memory design of them all and it has been longest in market.
What does this mean?

I'm not sure how the rest of the Tech world reacts to Global Recession, but I would assume Apple won't be that bothered about busting a gut to hurry out a new model that the majority of the public think is way beyond their means even when they are feeling flush.

As the world's governments make austerity budgets and promise to halve the national deficits I really don't see a market for a new über Mac being that buoyant, do you?

A Mac Pro has a productive lifespan of twice that of a PC possibly because Apple update their product range at half the speed, you ain't 4 generations of Mac out of date 5 years later because they don't bring a new one out every year.....

The 2009 Mac Pro may be a lash up for memory but it still tears all previous models a new one on benchmarks and real world performance. It also costs far more than any "normal" desktop PC so attracts a very small niche market by comparison. If Apple think the sales volumes are not there they will drag their heels on development till the market looks better. "Stay tuned" is as good an answer to "Why don't you blow millions on developing a machine that will not sell very many units ?" as they can give.

Hence the massive development of cheaper portable devices they think will sell at a million units a month while the tiny market share of financial hernia inducing towers stays in it's current guise. Dollars and cents win over hearts and minds every time on the bottom line.

"Deutschmark, Deutschmark über alles" as we used to sing in the EU way back in the 90's. :D

sOid

Jun 30, 2010, 04:01 AM

ValSalva

Jun 30, 2010, 04:45 AM

Well, the good thing of all this waiting is that the amount of interest on my savings account is increasing each day. So thank you Apple, for actually making me money in stead of spending it!

/sarcasm

:D and the price of the SSD I am going to install as the OS drive has decreased too. Thanks Apple.

Ravich

Jun 30, 2010, 11:18 AM

If the next "just wait" ends up being the macbook air.....

I just might lose it.

strausd

Jun 30, 2010, 04:41 PM

If the next "just wait" ends up being the macbook air.....

I just might lose it.

Hahaha that would be awful. The MP is about 100 days more overdue for an update than the MBA, so yes, that would be terrible. Honestly, with the iPad, I wouldn't be surprised if they got rid of the MBA soon.

Sin

Jun 30, 2010, 05:28 PM

UK Store seems to be down - clearly this means we're getting the new Mac Pro first.

deconstruct60

Jun 30, 2010, 09:21 PM

Actually, there's a fair number of partners, notably Manufacturing Partners listed by Vylen.

Most of those are parts suppliers not peripheral vendors. Namely mostly optical transciever vendors. Frankly the laser needed for 10G ethernet or infinband isn't all that different than light peak. electrical 1's and 0's come in and light goes out or light comes in and 1's and 0's go out. That is not a protocol controller. Where are the multiple implementors of the protocol controller ? ( this was a dust up with USB 3.0 also where Intel wanted to ship before anyone else had a chance to implement the standard)

Likewise vendors who are actually going to put parts in their devices that end users will buy. Unless end users buy complete devices to plug in it is not going anywhere. If end users are going to buy dongles that's going to limit adoption. I know Apple likes selling dongles but mini-display isn't exactly taking the market by storm.

Ultimately setting the stage for a full-fledged adoption = specification is accepted as a standard (since LP's not been created by an organization like IEEE).

Flash is a standard by that critieria. A real open standard is where there are multiple implementors. As long as there is just one implementor of the controller it isn't very standard. One of the short term limitations on USB 3.0 was that only NEC managed to get something working at first. There are other implementors coming online now.

Afterall, Sony finally won out with BR, and is the "last man standing" in the HD disk format war.

Sony isn't the only implementor of BR. Nor is there just one controller for the whole market.

Likewise with BR there were "end users" ( content providers ) on the standard committee. The analogous situation would be vendors who are going to put LP on their peripherals appear to be missing here. Not the parts folks.

Where it can get tricky though, is in the Tick part of the cycle, as the PCB's have already been designed and in production for about a year or so.

USB 3.0 parts have been available for at least a year at this point. Whereas LP parts are coming toward end of this year. Dropping USB 3.0 onto next year's board would be a far more conservative move. Could drop LP onto cards which drop into the PCI x16 slot in BTO.

Second, with Apple's decoupling the PCI planes from the CPU package sockets and high speed PCI controller means they can change PCB with zero impact on the sockets. Still will save money to run for two years with minor tweaks. However, money not socket implementation is driving that.

It's far easier to include (add, or swap out) newer components in a Tock cycle, as you have to design a new board anyway.

Eh? Tock is the "shrink with same socket" phase. Tick is where you upturn the socket and do big tweaks on architecture without doing any shrinks. Nehalem/Westmere -- tick/tock . SandyBridge/IvyBridge tick/tock. Next year's board is likely getting new socket anyway. I was talking about 2011 boards; not this year's.

To me it makes more sense to build into the motherboard a tech that has the more widespread set of devices. The number of USB 3.0 devices is going to greatly outnumber the number of LP ones. The exception would be two boxes need to snap together ( laptop with docking station). Unless there is going to be some lego block change to the MP, just don't see it as pressing.

What's attractive to me, is the fact there's not a dedicated protocol allowing it to be used for a multitude of busses. But this is one fact that can make or break it.

I don't buy that. There has to be a protocol. It may be a simple one just oriented to transport data packets from one machine to another with some simple QoS/isochronus abilities, but there has to be something. You can't have multiple protocols floating about at he same time. The telephone network has one for local delivery but there is another when mux multiple calls onto one line.

Big difference between LP being a "controller that just takes arbitrary 1's and 0's and flashes the laser on and off over a given optical fiber with a standard connector " and "a controller that can mux/demux multiple protocols onto a single wire".

I wouldn't be surprised at all that Apple wants to dump it in favor of an economically viable replacement, especially if it also includes things like far faster throughputs and the ability to minimize connections (can allow for other interface chips to be tossed out, actually rendering LP the cheapest solution). The latter is more of an issue with laptops, but Apple does well in that particular market.

I was talking about the optical aspects necessary for LP (why I mentioned the lasers, transcievers,... = optical aspects of the standard), not other standards such as Ethernet.

I meant Intel makes actual Ethernet NICs ("Ethernet connectors"). I wasn't talking about making RJ-45 jacks or PHY implementations. They should have labs with lasers , fiber cabling , etc. to test those devices since part of the products. The R&D is is primarily in the controller tech ( the protocol , etc.). The lasers transcievers , fiber cable , etc. was already invented. It would exist without light peak. If it exists without LP it is a big stretch to label that a LP R&D cost.

What I meant with InfiniBand, wasn't via translation, but rather as a replacement (bonded LP ports) to be able to reach sufficient speeds.

So now the "so lightweight you can't notice it" protocol supports bonded ports. I'm not holding my breath. In fact, in the demo two ports were attached to one controller. Bet you'd need two controllers to actually get 2x the bandwidth from box to box.

If not, then it's really only suited to the consumer market afterall right now.

Not sure where this went off the track that is wasn't primarily aimed at consumer market. The demos have consisted of showing digital video streams from one box to another box. They mention connectors and bus implementations common on generic whitebox PCs. Likewise, being harder to get to Class B when doing >5Gb is also somewhat a consumer oriented problem.

I think there is amble opportunity for Ethernet and Infiniband to adopt the laser transceivers and wider fiber cable to lower adapter costs. Optical connectors don't have to be crazy high in price.

Unfortunately, the software always falls behind the software, and it's not helped by the necessitation for backwards compatibility.

But it is increasing an old, tired excuse at this point. The Mac Pro in 2006 had 4 cores. Any app that can't go 4-way in some sections now are slackers in the Pro space. It has been 4 fraking years. That's going to include a major upgrade cycle for all but most glacially slow development. Some iMacs can go 4-way. A decent chance all iMacs could be 4-way by end of this Fall ( unless Apple sticks 2core/IGP into the lower end.)

Leveraging OpenCL or GCD ... yeah I can understand. That stuff has barely been in production release for a year. But scaling to more than 2 threads in your app is an old issue at this point.

Not being able to deal with 8 cores is on a slippery slope. It was extremely predictable given the transition from 2 to 4. It has been out for at least 2 years.

Then there's of course, software that can't benefit from SMP at all, such as a word processor.

It is relatively easy to start more than one program at a time. Even more so where there are ones that "do something" and don't require user interaction. For Apps that primarily sit and wait for the user actively do something .... often don't really need a MP. That isn't the MP's problem, nor should it particularly be a constraint on MP's design criteria.

nanofrog

Jul 1, 2010, 03:52 AM

Most of those are parts suppliers not peripheral vendors. Namely mostly optical transceiver vendors. Frankly the laser needed for 10G ethernet or infinband isn't all that different than light peak. electrical 1's and 0's come in and light goes out or light comes in and 1's and 0's go out. That is not a protocol controller. Where are the multiple implementors of the protocol controller ? ( this was a dust up with USB 3.0 also where Intel wanted to ship before anyone else had a chance to implement the standard)
I know most are suppliers on the component end, but the one that got my attention is Hon Hai Precision, which is a major end-product manufacturer.

Specifically the volume of OEM/ODM work they do for other major vendors (Apple and HP for example). The Foxconn label they produce as on their own, will only get Intel so far. I see the ability to use the OEM/ODM supplier as a means of persuading other vendors to take the plunge, whether installed on the main board products they sell to these vendors, or as an add-on card they'd supply as an optional product for those systems.

Likewise vendors who are actually going to put parts in their devices that end users will buy. Unless end users buy complete devices to plug in it is not going anywhere. If end users are going to buy dongles that's going to limit adoption. I know Apple likes selling dongles but mini-display isn't exactly taking the market by storm.
I figured this aspect was covered before in a previous post. At any rate, there are potential problems (that I do think will surface, but not sure to the degree yet, as there's still voids in the available information). Namely the supporting semiconductors necessary to make anything work with it becoming available, save connecting 2x LP equipped systems together, as was done in the demo.

USB 3.0 does have the advantage here, as it's already backwards compatible with devices users already own. For the newer standard, the necessary parts already exist, so it's not a situation of Part A is there, but B is vapor = useless for most, if not all practical purposes.

Flash is a standard by that criteria. A real open standard is where there are multiple implementors. As long as there is just one implementor of the controller it isn't very standard. One of the short term limitations on USB 3.0 was that only NEC managed to get something working at first. There are other implementors coming online now.
There's definitely a difference between a standard that began as an open standard (or became that shortly after it was created), and one that has it's IP held by a company or group of companies.

LP's not based on an open standard, so the companies involved would have to agree to do so before any other parts suppliers could get involved without a licensing fee. Unlike NEC, where it was, and they were just the first out with a working part. Others are following, but no license fees are involved.

Sony isn't the only implementor of BR. Nor is there just one controller for the whole market.
No, but they were pushing it hard, and are what most consumers associate with it (I was trying to keep the example simple, as the posts are getting long). Few may actually realize that others such as Panasonic were ever involved with it from the beginning.

Second, with Apple's decoupling the PCI planes from the CPU package sockets and high speed PCI controller means they can change PCB with zero impact on the sockets. Still will save money to run for two years with minor tweaks. However, money not socket implementation is driving that.
Money indeed. It was a good move, as there's fewer separate assemblies between the SP and DP systems. Shortens the parts bin requirements for easier production, and more importantly, saves costs.

Eh? Tock is the "shrink with same socket" phase.
Typing too fast. Not much different than typing MB/s instead of Mb/s mistake.

To me it makes more sense to build into the motherboard a tech that has the more widespread set of devices. The number of USB 3.0 devices is going to greatly outnumber the number of LP ones. The exception would be two boxes need to snap together ( laptop with docking station). Unless there is going to be some lego block change to the MP, just don't see it as pressing.
Other standards are entrenched, and won't be easily moved. I actually see LP making the largest impact on laptop systems as a means of simplification combined with a cost reduction.

For desktop systems, not as much. There would need to be some other reason, namely inexpensive speed, to cause them to go get the necessary pieces to make it all work (assuming they actually show up, as this late in the game, as none of the peripheral components have been announced that I've seen).

Workstations I hope will benefit by being able to exceed 1G E for example, for at most, similar funds. It would assist in the creation of small lower cost clusters IMO if this proves itself out (i.e makes clusters more available to independents or SMB markets that could utilize that much performance).

I don't buy that. There has to be a protocol. It may be a simple one just oriented to transport data packets from one machine to another with some simple QoS/isochronous abilities, but there has to be something.
I actually agree here, and do suspect it's that simple, as they've been silent on those details.

What I expect, is that they differentiate this from other standards (much more complex).

I meant Intel makes actual Ethernet NICs ("Ethernet connectors"). I wasn't talking about making RJ-45 jacks or PHY implementations. They should have labs with lasers , fiber cabling , etc. to test those devices since part of the products.
I know that they make NIC's, and they'd have labs with the necessary equipment for validation testing.

But that can be bought. I think of it this way. Why build a scope, frequency generator, .... when companies like Tektronix, Agilent and Lecroy already product it?

Lasers, fibre optics, ... are areas their partners specialize in. IIRC, there's a significant amount of work put into the lasers,..., namely to get the materials and manufacturing in at a low cost.

So now the "so lightweight you can't notice it" protocol supports bonded ports. I'm not holding my breath.
I previously mentioned that the ability to bond is a hoped for speculation that could allow it be used for more than just a way to simplify laptop connectors.

I think there is amble opportunity for Ethernet and Infiniband to adopt the laser transceivers and wider fiber cable to lower adapter costs. Optical connectors don't have to be crazy high in price.
Assuming Intel and their partners are willing to license them the technology.

But it is increasing an old, tired excuse at this point. The Mac Pro in 2006 had 4 cores. Any app that can't go 4-way in some sections now are slackers in the Pro space. It has been 4 fraking years. That's going to include a major upgrade cycle for all but most glacially slow development. Some iMacs can go 4-way. A decent chance all iMacs could be 4-way by end of this Fall ( unless Apple sticks 2core/IGP into the lower end.)
But they are slow at it, as it's expensive (time consuming as it's almost certainly a major re-write). Those that haven't done it seem to be waiting for someone else to do most of the work for them (i.e. waiting for tools, OS support,...). Basically just adding new features to the same old, tired base code that's been recycled for years (finding ways of milking the product with as little development time as possible).

It sucks, but I don't see anything motivating most software developers that've been sitting on the fence, to get off and go to work. :(

It is relatively easy to start more than one program at a time. Even more so where there are ones that "do something" and don't require user interaction. For Apps that primarily sit and wait for the user actively do something .... often don't really need a MP. That isn't the MP's problem, nor should it particularly be a constraint on MP's design criteria.
It was just to illustrate a point in general. Yes, multiple instances of the same application or simultaneous use of multiple applications can make better use of a system. But that won't change the fact if it can't benefit from SMP, which is what the MP and XServe are really meant for (where the performance will really become apparent).

toke lahti

Jul 2, 2010, 10:49 AM

Answer = additional part cost and added complexity (further increases cost).
This makes no sense.
Most laptops which are cheaper than cheapest Mac have eSata and adding that one chip to MP would raise the costs & "complexity" for about 0.1%.
Apple is almost infinite in greed, but I don't think that even they are so picky.

Anyway, when usb3 isn't enough, this will walk all over LP before LP is even in the market:
http://hdbaset.org/
All you need in computer is one RJ-45 and it works with existing cat5e/6 cabling and speeds are enough for homes and 99% of business...
If LP ever comes to market it will be as common as fc is now, which means many times more expensive than mass adopted options.

toke lahti

Jul 2, 2010, 10:55 AM

...vast majority aren't able to get 40Mb/s, which is the bare minimum for uncompressed signals...
You should know that uncompressed hd(-sdi) is 1.485Gb/s.
And 2.97 Gb/s for 4:4:4...

Heilage

Jul 2, 2010, 11:00 AM

You should know that uncompressed hd(-sdi) is 1.485Gb/s.
And 2.97 Gb/s for 4:4:4...

SDI is kind of a different league. I wouldn't compare that to anything for a consumer, really.

nanofrog

Jul 2, 2010, 12:37 PM

This makes no sense.
Most laptops which are cheaper than cheapest Mac have eSata and adding that one chip to MP would raise the costs & "complexity" for about 0.1%.
Apple is almost infinite in greed, but I don't think that even they are so picky.

Anyway, when usb3 isn't enough, this will walk all over LP before LP is even in the market:
http://hdbaset.org/
All you need in computer is one RJ-45 and it works with existing cat5e/6 cabling and speeds are enough for homes and 99% of business...
If LP ever comes to market it will be as common as fc is now, which means many times more expensive than mass adopted options.
Apple has a lot in common with budget box systems (desktop, laptop,...), where they try to rely on the chipset for connectivity as much as possible. Despite what they actually charge for the system.

Where the complexity comes in, is the fact that a USB 3.0 chip isn't a direct drop-in replacement (different package = not the same pin-out). That would force a board revision (new PCB's) to make it work. That's really where the cost comes in.

As for HDBaseT, it's aimed specifically at video (not a replacement for Ethernet, though it can run packetized data), and I'm not sure if that will become ubiquitous either. SDI isn't for mainstream use as stated by Heilage, and has licensing issues involved.

PaulD-UK

Jul 2, 2010, 02:14 PM

All you need in computer is one RJ-45 and it works with existing cat5e/6 cabling and speeds are enough for homes and 99% of business...Hi
Except...
That Apple has chosen to fit a half-cock Broadcom 5764 ethernet controller chip in the 27" i5/i7 iMac that doesn't support advanced features such as jumbo frames to enhance bandwidth...

And Apple chose to fit a half-cock ethernet controller chip (Intel Hartwell) in the 2009 Nehalem Mac Pro which totally craps out (= disconnects the network) at over 60MB/s, thereby requiring a PCI-e slot solution for high-bandwidth (= video) pro usage.

Short-termism or what?
A lot more than 1% of Mac Pros are sold to professional video-edit facilities :(
And that's why a replacement Mac Pro for 2010+ needs a re-engineered motherboard revision.

toke lahti

Jul 2, 2010, 02:28 PM

Where the complexity comes in, is the fact that a USB 3.0 chip isn't a direct drop-in replacement (different package = not the same pin-out). That would force a board revision (new PCB's) to make it work. That's really where the cost comes in.
They'll have to make new board for new MP anyway, so where's the additional cost?
Btw, does X58 have fw integrated?

nanofrog

Jul 2, 2010, 02:45 PM

They'll have to make new board for new MP anyway, so where's the additional cost?
Btw, does X58 have fw integrated?
Not with the 2010 systems. Intel designed the socket and chipset to work with both halves of the Tick Tock cycle, as it's cheaper. So the existing boards and other parts (non CPU), will work with the newer parts with microcode update. Very cost effective this way, as R&D and manufacturing costs (i.e. tooling) are spread out over 2 years instead of one.

Apple had to put the FW chip in, as that's not built-into an Intel chipset (they've a history with FW, and there's been need of something faster than USB 2.0, though they took their time making the entire line S800 compliant IIRC).

deconstruct60

Jul 2, 2010, 03:00 PM

As for HDBaseT, it's aimed specifically at video (not a replacement for Ethernet, though it can run packetized data), and I'm not sure if that will become ubiquitous either. SDI isn't for mainstream use as stated by Heilage, and has licensing issues involved.

No, HDBaseT is aimed at video and Ethernet. A significant number of TVs are already getting Internet jack already. This is in part driven by

http://en.wikipedia.org/wiki/Digital_Living_Network_Alliance

HDBaseT can be used to provide a 100BaseT along the exact same capabilities that DLNA enables. This connector takes it a step further in that can also transfer large Video/Audio files on same cable ( if bother to buy Cate6 cable in first place. Shouldn't be that hard to start to tell folks to use those with their new TVs. )

The significant factor is that it is the major TV vendors are pushing this.
Whether they are peripherals or not depends upon as being the hub in the system, the data or the display.

I'm kind of curious how they ship high power over RJ-45 and still get the high throughput. Will probably need a special "video Ethernet" to regular "Ethernet" bridge so that TVs can still get to the internet and interact with the other home network computers, but that's a reasonable price to pay if now can transfer large files between media appliance boxes easily.

It has similar "but the plug looks the same" problem that LP will have if it reuses USB like connector. Colored cables and sockets can help with that.

This solution runs into a brick wall when need even faster Internet connection (although could just put in second RJ-45 for that) and even larger video/audio files. However, in a world where most folks only ship 1080p content over short distances it works. It doesn't depend upon some large, widely distributed, infrastructure to be deployed to work.

So can easily build a smart connection between TV and DVR box and pump mainstream internet content through the same connection. The TV focuses on display and decoding streams while the external box focuses on storage and serving the data. Plug it in and will work together if the all the vendors make them play nice together.

It doesn't have to replace LP. LP is the one with the big anchor around its next, because it has to replace everything to live up to the hype.

toke lahti

Jul 2, 2010, 03:25 PM

Apple had to put the FW chip in, as that's not built-into an Intel chipset (they've a history with FW, and there's been need of something faster than USB 2.0, though they took their time making the entire line S800 compliant IIRC).
So it wouldn't be any more complicated to add another chip for usb3. And this is why many manufacturers are already doing this. Don't know how long Apple will keep Macs away from anything faster than fw800. Still 2 years before LP that will flop?

toke lahti

Jul 2, 2010, 03:29 PM

LP is the one with the big anchor around its next, because it has to replace everything to live up to the hype.
This is why LP will fail in mass market. Nothing can replace everything anymore. Too much infrastructure already installed. And new things will need mass adoption from consumer market to succeed. Otherwise price point will be too high.

nanofrog

Jul 2, 2010, 03:44 PM

So it wouldn't be any more complicated to add another chip for usb3. And this is why many manufacturers are already doing this. Don't know how long Apple will keep Macs away from anything faster than fw800. Still 2 years before LP that will flop?
The FW chip is already installed on the existing boards. Adding anything that's not a direct drop-in replacement, means the PCB's have to be reworked. This is the point you seem to either be missing, or underestimating the time and money involved.

The newer CPU's don't need new boards at all, just a change to the firmware so they'll work.

If this were new architecture, then it makes sense to add anything new, assuming it's obtained financial approval.

This is why LP will fail in mass market. Nothing can replace everything anymore. Too much infrastructure already installed. And new things will need mass adoption from consumer market to succeed. Otherwise price point will be too high.
It's intended to be cheap, which will help. The biggest issue IMO will be the peripheral chips needed to make other standards work over it. User's can't adopt if it doesn't exist, no matter how low the cost is on the system end.

I see potential, but will wait and see what actually materializes, as it's easier to screw up than get it right (missing parts, high costs,...). There's also a fair bit of missing information, and that will determine what it's really going to be usable for (i.e. consumer use only, or will it extend to lower cost solutions for independent and SMB sized organizations).

deconstruct60

Jul 2, 2010, 04:04 PM

This is why LP will fail in mass market. Nothing can replace everything anymore.

I honestly don't think the technical folks constructing LP are trying to make it suck in everything. The everything spin is more a marketing ploy to get folks to buy into yet another new standard. Few folks want a new one with all new connectors and dongles to keep things to connected. However, if hold out the illusion that everything will just hook up and magically work folks will play along. If it can devour everything it must be great, right? Just let people fill in the blanks for themselves.

There does need to be something in the home and over 5Gb space. If Intel can get enough LP deployed so that it is a viable player they could own that. In the mean time between now and when that is a more pressing need, they can transport slower legacy stuff over one wire. In the short term, aggregating lots of different streams together is the only way to drive the high bandwidth requirement in the generic commercial marketplace.

It is also a different approach than USB 3.0 which only tackles one protocol. I don't think Intel is trying to kill off USB 3.0 with LP as much as trying to get USB 4.0 moving so that can be adopted. One way of doing that is to make USB 4.0 tackle a different set of problems initially. Or at least divert USB 4.0 into making quality improvements that don't involve speed.
That would open the door so that there were two "universal" domains. One over and one under 5Gb. Whether they both live under "USB" umbrella is open story.

Note that USB never was going to kill off Ethernet or other more long distance connections. There is an implied "universe" of devices it was going to be applied to that didn't actually include everything.

toke lahti

Jul 2, 2010, 05:01 PM

Adding anything that's not a direct drop-in replacement, means the PCB's have to be reworked.

The newer CPU's don't need new boards at all, just a change to the firmware so they'll work.

Well, MP's memory design needs to be worked anyway.
Or it will be interesting to see what kind of welcoming new MP will have if it still has the worst design of all new workstations. Surely Apple could afford new board design more often than every 3 years...

deconstruct60

Jul 2, 2010, 05:11 PM

This is the point you seem to either be missing, or underestimating the time and money involved.

Some folks may think the run rate is high enough you can distribute the costs. For example 1M Mac Pros even if the PCB adjustment is $2M that is only $2 a device. Apple's margins on a MacPro are at least around 25-30+%. 30% of $2,500 is $750 and 25% is $625. $2 is about a 0.26% change in margin (or 0.3% in 25% case). That isn't a big difference on the corporate balance sheet ( it still going to register as overall 30+% corporate margin) versus putting something of additional value on the box to justify the higher than average price.

I don't think the run rate is anywhere near that high. It is probably closer to a 10th of that. So the percentage change would be high given those conservatively low margin estimates. The other benefit though is that low run rates won't effect the corporate or mac numbers since relatively small.

If Intel came to apple over a year ago and said USB 3.0 is not coming for 3-4 years instead of 1-2 years, then it would demonstrate that Apple is more responsive that other large workstation vendors if they made shift now rather than later. If going to be several years before USB 3.0 is in the chipset then far more discrete solutions are going to get deployed than were envisioned when core chipsets were going to get them "real soon".
Going with the "we only implement what Intel gives us" approach makes it much more difficult for Apple to justify their price premiums for "thoughtful" design.

If Apple sticks to about 12 month cycle for renewal then at this point there will be other workstations released with USB 3.0 onboard before the 2011 rev of the Mac Pro comes out. Additionally, since the CPU/chipset is on another PCB board could make a new one that would take both current and new daughter cards with this one update. (i.e., this part of the board is not hard synched to Inel CPU socket updates.). In that case just moved cost would have incurred from one year to another. There would be no impact on amortization.

In short, it is somewhat of a myopic viewpoint to just look at what happens over 3-4 quarters when perhaps making a change now would be more beneficial over 8-24 quarters. Apple has enough money they don't have to optimize for each quarter. They can make long term strategic moves that may temporarily cost more money, but has far more long term upside. For example, putting the right set of parts on a board given all the of the advance information they are given.

Personally if Intel told me well in advance they were going to delivery the 3620 and 3640 late, push USB 3.0 back multiple years (this on top of previous meetings with Intel where there date kept sliding backwards) , and stick me with IGP processors in my other Macs that was significantly slower than what as shipping now.... I would have rejuggled the rollout for this and next year also.

However, Apple could take the Scrooge McDuck route. It is only going to give the folks with the "lack of value" just more fuel to add to the fire. Apple is going to be last system vendor by years to support USB 3.0 if wait till it is in the Intel chipset (Linux already in production. Windows 7 should get it next service pack if recall correctly). Shoveling $5-10 more in their overflowing pockets is worth loosing customers over time.

bzollinger

Jul 2, 2010, 05:37 PM

Originally Posted by nanofrog
Adding anything that's not a direct drop-in replacement, means the PCB's have to be reworked.

The newer CPU's don't need new boards at all, just a change to the firmware so they'll work.

Well, MP's memory design needs to be worked anyway.
Or it will be interesting to see what kind of welcoming new MP will have if it still has the worst design of all new workstations. Surely Apple could afford new board design more often than every 3 years...

This is interesting. So what nanofrog is saying is that the 2010 single proc MPs will still only have 4 RAM slots??:eek:

Apple better not fill three of them up with 1GB chips for the base model!

nanofrog

Jul 2, 2010, 06:49 PM

Well, MP's memory design needs to be worked anyway.
Or it will be interesting to see what kind of welcoming new MP will have if it still has the worst design of all new workstations. Surely Apple could afford new board design more often than every 3 years...
Technically speaking, they could do a PCB rework for things like DIMM slots and add features such as USB 3.0 (personally, I'd like to see a SAS RAID chip added, as it would better compete with other offerings = adding value to the system from previous MP's). But given the financial aspects (want the highest margins possible = lowest production costs possible for the target MSRP), I just don't see any of that happening. :(

To keep the margins static (same fixed %), the prices would increase even further to recoup the additional costs. And I'm not so sure how much more buyers would tolerate, given they're using the same basic technology as systems produced by other vendors (same Intel CPU family and chipsets in the workstation segment). And that's just to keep up with other choices in the market, not necessarily exceed them (6x DIMM slots per CPU certainly comes to mind).

Personally, I think there's better odds on a new case, as the existing boards can be stuffed in. Then recycle it for the next model (internals reworked as necessary to fit new board configurations), as they've done for years.

I don't think the run rate is anywhere near that high. It is probably closer to a 10th of that. So the percentage change would be high given those conservatively low margin estimates. The other benefit though is that low run rates won't effect the corporate or mac numbers since relatively small.
I'm figuring things on ~100K units. Unfortunately, I'm not sure if the PCB rework will actually come in in the $2M range though, and there's contract issues to take into account as well (i.e. did they negotiate for a 1 or 2 yr production run?).

Either higher costs or contract reasons could be major sticking points that result in unjustifiable due to financial impact.

For example, if the rework and production increase (based on 100K) is $20M, the cost per system increases $200 per unit (design for both 6x DIMM's per CPU and USB 3.0, validation testing, and component acquisition). That would get their attention, as it's a substantial rework (mainly due to the DIMM slot portion within the space constraints).

If Intel came to apple over a year ago and said USB 3.0 is not coming for 3-4 years instead of 1-2 years, then it would demonstrate that Apple is more responsive that other large workstation vendors if they made shift now rather than later. If going to be several years before USB 3.0 is in the chipset then far more discrete solutions are going to get deployed than were envisioned when core chipsets were going to get them "real soon".
Going with the "we only implement what Intel gives us" approach makes it much more difficult for Apple to justify their price premiums for "thoughtful" design.
I'm not sure on the Apple - Intel relationship situation, but I've the impression it's more of the latter (base on the chipset methodology to minimize costs).

Additionally, since the CPU/chipset is on another PCB board could make a new one that would take both current and new daughter cards with this one update. (i.e., this part of the board is not hard synced to Intel CPU socket updates.). In that case just moved cost would have incurred from one year to another. There would be no impact on amortization.
This was a smart move IMO for reduction in SP and DP system cost, and as you indicate, could prevent the necessitation of a complete redesign with each new architecture (assuming of course, the main connector between the daughterboard and main board doesn't end up short on pin count). Not enough traces for PCIe lanes for example.

In short, it is somewhat of a myopic viewpoint to just look at what happens over 3-4 quarters when perhaps making a change now would be more beneficial over 8-24 quarters. Apple has enough money they don't have to optimize for each quarter. They can make long term strategic moves that may temporarily cost more money, but has far more long term upside. For example, putting the right set of parts on a board given all the of the advance information they are given.
I agree with this, but with the changes that occur, there's limitations as to how far projections can be made.

Personally if Intel told me well in advance they were going to delivery the 3620 and 3640 late, push USB 3.0 back multiple years (this on top of previous meetings with Intel where there date kept sliding backwards) , and stick me with IGP processors in my other Macs that was significantly slower than what as shipping now.... I would have rejuggled the rollout for this and next year also.
Understandable, and I agree.

Unfortunately, I don't think Intel came completely clean on this (i.e. chipset support delay for USB 3.0 not mentioned at that time).

It's another reason that could explain Apple's negotiations with AMD, as they can get consumer chips cheaper, and continue using discrete graphics for improved performance. But I don't see it working out as well in the workstation segment (though not impossible it could happen).

However, Apple could take the Scrooge McDuck route. It is only going to give the folks with the "lack of value" just more fuel to add to the fire. Apple is going to be last system vendor by years to support USB 3.0 if wait till it is in the Intel chipset (Linux already in production. Windows 7 should get it next service pack if recall correctly). Shoveling $5-10 more in their overflowing pockets is worth loosing customers over time.
This seems to be the approach they pursue though, going by what's happened with the Intel based systems (i.e. no support for things like SAS controllers/built-in hardware RAID). They did offer a much better value however, as the prices were lower (better cost/performance).

Perhaps the market's smaller than even we're estimating, and they're just milking what they can out of it rather than investing a single additional cent past bare minimum to retain customers, let alone increase them. Further, software issues such as the lack of FCP being updated to what users really want/need, and the tumultuous relations with Adobe, are causing users to look to other solutions for their needs.

Hard to say for sure, but there are increasing alternatives from what I understand from various posts in terms of software, which opens up the possible use of other systems.

This is interesting. So what nanofrog is saying is that the 2010 single proc MPs will still only have 4 RAM slots??:eek:
Unfortunately, I do think they'll only have 4x DIMM's per CPU (see above). :rolleyes: :(

I'm not sure if they will continue with 1GB sticks in the base configurations or not, but I wouldn't totally discount that being the case though.