Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

deconstruct60

macrumors G5
Mar 10, 2009
12,291
3,884
I still think you'll find that >1 is preferred by most "Pros" but I don't have any scientific data to back that up, just what I've read here and online.

One, the sampling here in these forums are highly skewed and exteremely likely non representative.

Two, there is almost nothing in the rumors/rumblings that says that the scaled down box is not >1. The mini isn't 1 (or zero). The iMac isn't 1 (or zero). The probability that the box more expensive than either one of those has less storage devices is pretty small. It doesn't even follow Apple design constraints for Macs.

The super handwaving on this front seems to be largely driven by mapping MBP evolution onto the Mac Pro. That is just looney-tunes. I don't think even think Apple is drinking that kind of Cupertino kool-aid. It is a completely different category. It would make almost as much sense to map the Mac Pro design constraints onto an iPad. Apple is trying to turn the Mac Pro into a laptop? Bull droppings.

The other set of hand waving is folks trying to mutate the Mac Pro into an iMac or Mini they like better ( warp the Mac Pro into "headless iMac" or "super mini"). Again why? Apple's got iMacs and mini class Macs in the line up. Those are not Apple motivations or design constraints.


Empty slots ... sure Apple may nuke those (at the very least the probably now vestigial 5.25" one(s)) . Wouldn't be surprising. Reduce the number to < 2.... sure has not done it on the other desktop Macs. In fact, Apple has made a big deal about "Fusion Drive". Why on Earth would that now want to incapacitate that?????? Where is the source for the motivation??????

Apple already has the rMBP, Mac mini, and iMac which depend heavily on TB.

LOL. Largely because they do not have PCI-e slots. That is where TB is a primary solution. As long as Apple leaves 1+ PCI-e slots on the Mac Pro, TB will play a secondary role.



but I hope they don't ruin the Mac Pro (for me, of course ;)).

I understand people "voting for me" but that is actually part of the FUD here. There are xMac , superMini folks "voting for me" by throwing out there Mac Pro mutations too.

None of those is likely to primarily guide Apple's choices because they aren't guided by aggregates of people (i.e., markets) and what those groups are doing.
 
Last edited:

ValSalva

macrumors 68040
Jun 26, 2009
3,783
259
Burpelson AFB
The super handwaving on this front seems to be largely driven by mapping MBP evolution onto the Mac Pro. That is just looney-tunes. I don't think even think Apple is drinking that kind of Cupertino kool-aid. It is a completely different category. It would make almost as much sense to map the Mac Pro design constraints onto an iPad. Apple is trying to turn the Mac Pro into a laptop? Bull droppings.

The other set of hand waving is folks trying to mutate the Mac Pro into an iMac or Mini they like better ( warp the Mac Pro into "headless iMac" or "super mini"). Again why? Apple's got iMacs and mini class Macs in the line up. Those are not Apple motivations or design constraints.

Voice of reason. I really like your posts. I still want at least one or two "extra" 2.5" empty drive bays and enough PCIe slots for PCIe SSD's though ;) Of course Apple will do what makes the most economic sense for them and I'm sure I don't fall into that category.
 

Shaggymax

macrumors regular
Jun 8, 2013
175
133
Lanc, Penn USA
The MacBook Pro is a portable workstation > over the years it has become more powerful, thinner and more portable.

Using the same logic, the Mac Pro is a powerful fixed workstation > it should become more powerful, flexible and expandable.

Fingers crossed for a good 2013 WWDC!
 

slughead

macrumors 68040
Apr 28, 2004
3,107
237
Thunderbolt's job is to deliver the SATA controller's data from the external box to the host. That is it. It is like complaining two PCI-e cards with similar controllers deliver similar performance are "bad".

Right, but there is no such thing as a "thunderbolt" hard drive, all external enclosures that use thunderbolt bridgeboards produce a significant amount of latency (just like FW and USB bridgeboards, perhaps just to a lesser extent).

If we're then relegated to stuff all our drives into a box and run it through a TBird controller, we're getting worse performance and ultimately a larger footprint on our desk at a much higher cost than having internal drives hooked directly to the on-board Sata Controller (thanks Apple!).

There is a $300 consumer ASUS motherboard with TEN Sata ports, all of which are 6gbps hot-swappable and can be converted to ESATA with a simple adapter, subtracting no performance at almost no additional cost. This is what professionals want. You can't get better performance than that. Oh, by the way, it has 2 thunderbolts too.... For $2500, why the heck not?
 
Last edited:

goMac

Contributor
Apr 15, 2004
7,662
1,694
My 2 cents: I'm starting to highly doubt we'll see an update at WWDC. But maybe I'll be surprised.
 

spunkgarLEWII

macrumors regular
Jun 1, 2013
100
0
What is the basis for your argument that you highly doubt we will see an update? Possibly we will, but will it even be a Mac Pro? Tim's email was not very specific at saying for sure it will be a new Mac Pro. For all we know it could be a piece of PRO software for the current machines and future machines(think MacBook Pro and iMac here), or it could be a machine that is not expandable and offers no upgrade path whatsoever.

Given Apple's current direction towards consumer and prosumer, I expect to see some kind of machine but the Mac Pro as we know it will cease to exist, replaced by some hybrid between iMac and Mac Mini..

I only hope I am wrong and so are you.. I am 50/50 undecided since little hardcore evidence is available to suggest a new machine.

My 2 cents: I'm starting to highly doubt we'll see an update at WWDC. But maybe I'll be surprised.
 

DisMyMac

macrumors 65816
Sep 30, 2009
1,087
11
I only hope I am wrong and so are you.. I am 50/50 undecided since little hardcore evidence is available to suggest a new machine.

I fear they will only give a presentation on the new Pro factory (USA). Specs and availability delayed, as always.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,291
3,884
all external enclosures that use thunderbolt bridgeboards produce a significant amount of latency (just like FW and USB bridgeboards, perhaps just to a lesser extent).

Not. Nothing there is a demonstration at all of Thunderbolt infrastructure latency. A demonstration of different implementations, but not of latency.

The LaCie device is being attached to another box with its own port-multiplier overhead. First, none of that is necessary. Thunderbolt doesn't demand that. So it can't possibly be a "tax" or onerous latency layered by TB. Thunderbolt isn't going to make any possible choice of equipment work faster.

If we're then relegated to stuff all our drives into a box and run it through a TBird controller, we're getting worse performance and ultimately a larger footprint on our desk at a much higher cost than having internal drives hooked directly to the on-board Sata Controller (thanks Apple!).

If you buy the wrong product or there isn't a "right" product in the TB market that isn't Apple's fault.


There is a $300 consumer ASUS motherboard with TEN Sata ports,

There is zero elements of Thunderbolt that prohibit a peripheral vendor of doing exactly the same thing inside of their external box. Nothing.

For better or worse no-one has build a JBOD Thunderbolt box yet. I'm not sure why it is taking so long. I guess because aimng for the higher end of the market were folks want to export a substantially more expensive SATA controller (e.g., RAID controller) out to the external box as opposed to the "motherboard breadth" controller like offerings.

It will cost a bit more becuse the 6 "free" ports from the chipset would have to be replicated by a discrete controller but that's isn't going to incur any latency problems. Perhaps bandwidth if stack enough SSDs on the resulting SATA network, but not latency.


We are now at a year after this article

http://www.anandtech.com/show/5956/qnaps-jtb400-a-byod-4bay-thunderbolt-enclosure

and still no JBOD box. Perhaps folks are waiting for TB pricing to settle down and chasing higher priced, lower volume products first during this initial phase of TB adoption. Eventually they will show up.

all of which are 6gbps hot-swappable

No hot swappable TB perherperials? Don't think so. Again this is a matter of vendor implementation; not Thunderbolt.

and can be converted to ESATA with a simple adapter,

Rather dubious when placed inside of a external TB peripheral since already outside the primary host box. Also, it is only going to lead to exactly the overhead shown in the benchmarks you used to poo-poo TB above.

subtracting no performance at almost no additional cost.

The almost no additional cost has only to do with the chipset ports which you have to buy anyway. The chipset has to be bought with the CPU. That has jack squat to do with with Thunderbolt latencies, speed, or suitability to transport SATA pci-e data.

If this was a Z77 chipset then it would be suck with just 6 Gb/s SATA ports. It is only 10 because the chipset moved up. If had to deliver 10 6b/s ports in previous version then would have paying substantially more ( either in a tradeoff in board features/space or discrete controller).

All of that is relatively independent of Thunderbolt speeds, costs, and latencies.


This is what professionals want.

As a whole group? Not really. Throwing digits number of drives into a box that isn't primarily a storage node is questionable. As the drive number goes up the likelihood of a failure goes up. That hot swap isn't useful unless can get to the drive. Making >10 drives all easily accessible for hot swap soaks up alot of room. Like these systems.

http://www.supermicro.com/products/nfo/Xeon_X9_E5.cfm?pg=SS&show=SELECT&type=SSG

Enough to probably exclude in a reasonably sized box multiple high TDP cards. The focus with interactive workstations is gong to be weighted more so on the high TDP cards.

Large data storage pro break out storage boxes into separate units specialized just for storage. This whole "everything and the kitchen sink has to fit inside of one box" isn't "pro". "If all you have is a hammer everything looks like a nail" isn't being a professional.

The tweaker/builder/'experiment in the basement' crowd... yes.
Spec porn chasers ... yes to those folks too. ( my box has more xxxx than yours.... it is more powerful. ).


You can't get better performance than that. Oh, by the way, it has 2 thunderbolts too.... For $2500, why the heck not?

Why not? It is capped at 4 x86 cores.
It has subsantially less I/O bandwidth than a Xeon E5 solutions would have ( capped at x16 PCI-e v3.0 lanes versus the x40+ a E5 system would have). The all that "mega" SATA capacity is going to get bottlenecked if actually try to walk and chew gum at the same time. There is a decent layer of PCI-e switches in there to "fake the funk" to hook up all those high bandwidth PCI-e v2.0 controllers.

Throw one or two PCI-e SSD cards at it, hook up and a couple SuperSeed USB 3.0 devices, 10GbE Thunderbolt device , and fire up those 10 drives and watch it wheeze. Juggle one or two of those? sure. Do all at the same time? no.
 

slughead

macrumors 68040
Apr 28, 2004
3,107
237
Not. Nothing there is a demonstration at all of Thunderbolt infrastructure latency. A demonstration of different implementations, but not of latency.

Why don't you provide a source? Show me some benchmarks that show that TB bridgeboards don't have latency and I will be quite amazed. I have more benchmarks below.

There is zero elements of Thunderbolt that prohibit a peripheral vendor of doing exactly the same thing inside of their external box. Nothing.

Nope, I'm sure you could have 10 hard drives hooked up via thunderbolt... though somehow Tbolt has significant bottleneck

Thunderbolt RAID 0 with 6Gb/s SSDs appears to run into a bottleneck when you compare it to the SAS RAID 0 with the same 6Gb/s SSDs. I guess the 1000+MB/s theoretical bandwidth is... theoretical. - Barefeats​


[performance of sata when put through an esata adapter is] Rather dubious when placed inside of a external TB peripheral since already outside the primary host box. Also, it is only going to lead to exactly the overhead shown in the benchmarks you used to poo-poo TB above.

Really? An adapter with no bridgeboard and 3 feet more of cable operating at the speed of light is going to be the same speed as a bridgeboard totally converting one signal to another? I find that claim highly dubious. It doesn't seem to slow down SAS (see above benchmarks).


If this was a Z77 chipset then it would be suck with just 6 Gb/s SATA ports. It is only 10 because the chipset moved up. If had to deliver 10 6b/s ports in previous version then would have paying substantially more ( either in a tradeoff in board features/space or discrete controller).

.... I'm not sure what your point is... this motherboard runs 10 ports, the previous one was unable to. I was simply pointing out that this rather inexpensive consumer motherboard has 4 more ports than the current MP each with twice the bandwidth, therefore I expect more out of the MP as ASUS was able to turn a profit selling this for $300.


[Question: Professionals want room for more hard drives with more bandwidth] As a whole group? Not really.

Really? You can't think of a reason to have more than 4 hard drives? Now that prices are getting ridiculously low again, you wouldn't want an extra backup drive?

Throwing digits number of drives into a box that isn't primarily a storage node is questionable.

My point was I hope they don't cut the number of hard drives down from 4 and force us to use thunderbolt... didn't I say that somewhere? I would prefer to have at least 6 internal HDs, I'll concede that may not be necessary for most professionals.

As the drive number goes up the likelihood of a failure goes up.

So you prefer to have fewer hard drives because you're afraid one might fail? That's a strange way of thinking about things...

There's this thing called "RAID." In many arrangements, the more drives you have, the more reliable the system. Not all pros want to have this useless "Time Machine" disk that can't be used for anything other than restoration in the event of failure. 4 hard drives is okay (my point was more that I hope they don't cut them down), 6 would be better.


Why not? It is capped at 4 x86 cores.

My point was if Asus can fit all that crap on an inexpensive consumer board, why can't Intel/Apple put this into a $2,500 computer. I'm not sure how pointing out how this motherboard is not a Xeon motherboard answers that question ??

----------

What is the basis for your argument that you highly doubt we will see an update?

He's probably basing it on the last 3 years :)
 
Last edited:

deconstruct60

macrumors G5
Mar 10, 2009
12,291
3,884
What is the basis for your argument that you highly doubt we will see an update?

He said update at WWDC ; not update. There is big gap there.

If Apple is moving the Mac Pro forward along the natural evolutionary infrastructure path this is the wrong time. Xeon E5 v2 is "a couple months away". 3 years "late" or 3 years and 3 months "late"... not much of a difference.

The time for Apple to "do something" in the first half of 2013 was in Jan-Feburary before they had to remove the product from the market in the EU. Now it is just more screwed than it was. So screwed up another 2-3 months is pretty much status quo.


Even if Apple was downshifting down onto Xeon E3. (dumping dual package set ups and capping the RAM expansion to iMac levels ). It is still an odd-ball time since 10.9 is due real-soon-now. It is yet-another system that would have to be short term validated on 10.8.4 (10.8.5 ?) and then flipped a couple months later to 10.9.

There is some Mac features that they'll need for WWDC demo but Mac Pro probably doesn't have any of them.


Finally, if completely changing contract manufacturors ( e.g., moving to a USA facility) then the initial product ramp probably has bugs. Fixed points in time like WWDC are bad things to coupled something like that too. ( e.g., Apple coupling the iMac to the iPad mini rollout .... not so great idea with bad outcome. I doubt Apple wants to repeat that again even if only the same general ballpark reasons. )


Most of this "it has to come at WWDC" is being driven by tying it to the speed bump last year at this time rather than any solid evidence. If look at this forums archives every year before NAB or WWDC there is a "rush" of 'Apple is going to do a big dog and pony for for Mac Pro at XXX this year". Most of the time it never happens. Really didn't happen last year either even with speed bump update.

Retail supplies hiccup from time to time on Apple products. There are not a good primary indicator of new models. ( there is no track record for that.... other than folks selectively cherry-picking the subset that do proceed. ) And even slowed doesn't make them come any faster. iMac slowed and they didn't show.
 

swester

macrumors 6502
Jul 26, 2010
259
904
Back to the original content of this thread, all those stores have Mac Pros in stock again.

It's pretty remarkable they'd even bother to restock - who in their right mind would buy one new at this point?
 

MacPoulet

macrumors 6502a
Dec 11, 2012
544
371
Canada
It's pretty remarkable they'd even bother to restock - who in their right mind would buy one new at this point?

Our building just got flooded, several labs will need to be replaced right away. Not sure how many Mac pros got soaked, but insurance claims would be one reason.

Lifecycle upgrades would be another.
 

spunkgarLEWII

macrumors regular
Jun 1, 2013
100
0
I guess your Mac Pros couldn't swim... :( Sorry to hear of your loss..

File insurance claims.

Our building just got flooded, several labs will need to be replaced right away. Not sure how many Mac pros got soaked, but insurance claims would be one reason.

Lifecycle upgrades would be another.


----------

I agree with everything you said.. these are not indicators of new models coming.

He said update at WWDC ; not update. There is big gap there.

If Apple is moving the Mac Pro forward along the natural evolutionary infrastructure path this is the wrong time. Xeon E5 v2 is "a couple months away". 3 years "late" or 3 years and 3 months "late"... not much of a difference.

The time for Apple to "do something" in the first half of 2013 was in Jan-Feburary before they had to remove the product from the market in the EU. Now it is just more screwed than it was. So screwed up another 2-3 months is pretty much status quo.


Even if Apple was downshifting down onto Xeon E3. (dumping dual package set ups and capping the RAM expansion to iMac levels ). It is still an odd-ball time since 10.9 is due real-soon-now. It is yet-another system that would have to be short term validated on 10.8.4 (10.8.5 ?) and then flipped a couple months later to 10.9.

There is some Mac features that they'll need for WWDC demo but Mac Pro probably doesn't have any of them.


Finally, if completely changing contract manufacturors ( e.g., moving to a USA facility) then the initial product ramp probably has bugs. Fixed points in time like WWDC are bad things to coupled something like that too. ( e.g., Apple coupling the iMac to the iPad mini rollout .... not so great idea with bad outcome. I doubt Apple wants to repeat that again even if only the same general ballpark reasons. )


Most of this "it has to come at WWDC" is being driven by tying it to the speed bump last year at this time rather than any solid evidence. If look at this forums archives every year before NAB or WWDC there is a "rush" of 'Apple is going to do a big dog and pony for for Mac Pro at XXX this year". Most of the time it never happens. Really didn't happen last year either even with speed bump update.

Retail supplies hiccup from time to time on Apple products. There are not a good primary indicator of new models. ( there is no track record for that.... other than folks selectively cherry-picking the subset that do proceed. ) And even slowed doesn't make them come any faster. iMac slowed and they didn't show.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,291
3,884
Why don't you provide a source?

Why don't you provide a source? Latency is measured in seconds. It only takes 1-2 seconds of looking at your so-called "proof" to distinguish that the units on the graph are wrong. They are not seconds. So therefore you have not measured latency.

What you have provided is smoke. I'm not going to do your homework for you if all you can do is blow smoke at me.


Show me some benchmarks that show that TB bridgeboards don't have latency and I will be quite amazed.

More of the same smoke. You can pile the smoke up all you want to it only clarifies you have no idea of what you are talking about.

The MB/s that is being measured isn't even salvageable here to make a decent guess at because the experiments involve sequential access. Frankly, I can't quickly think of a better way of suppressing the measurement of latency that in doing large sequential file benchmark. So not only are the units wrong the experiment is actually designed to suppress what you are waving your arms about.


In terms of measuring latency these are experiments with craptastic design.

These benchmarks measure throughput/bandwidth not latency. However, even in terms of that your full of it.


somehow Tbolt has significant bottleneck

ROTFLMAO. The TB set up there beats almost everything on the list. In fact from the article

"Insights
....
4. Thunderbolt just made Apple's Pro RAID card obsolete. ... "

Sure the card is older and has few fans here on these forums among folks who focus on RAID cards. But where is this onerous latency?

The only thing that beats the TB set up is a Highpoint RocketRaid 2770. Well that just happens to be a x8 PCI-e card. Thunderbolt is limited to x4. The Highpoint card is about 2x the throughput of the TB solution and delivers about twice the performance. I know, that is only coincidence. They can't possibly be largely driven by the difference in RAID controllers... no way. *cough*

If you put a slower SATA RAID controller on the other end of a TB connnection, TB isn't going to make it any faster. Period.

TB's relatively low latency largely is relative to the data of the much slower legacy protocols it transports. When the TB channel is 10Gb/s and FW and/or USB is less than 4Gb/s there isn't going to be a latency problem. The internals of the TB device have to respond to changes in signals and requests much faster.


Thunderbolt RAID 0 with 6Gb/s SSDs appears to run into a bottleneck when you compare it to the SAS RAID 0 with the same 6Gb/s SSDs.


All of which completely ignores the fact that the data from the disks has to pass through the respective RAID contronllers. If actually trying to measure latency then all of the elements would be the same. disks, raid controller , host box while making the only difference being inserting a TB controller or not.

None of these 10 Gb/s SATA channels are going to keep up with 10 6G/s SSD going full blast. All SATA controllers can only handle a subset of the number of lanes they support ( usually somewhere in the 50-70% zone depending upon price.)





Really? An adapter with no bridgeboard and 3 feet more of cable operating at the speed of light is going to be the same speed as a bridgeboard totally converting one signal to another?

Completely immaterial since no SATA controller is any faster. Travesing 3 feet of cable at 950,400,000 feet/sec ( ~180,000 miles/sec * 5280 feet/mile) ... yeah sure that is going to be significant. 3x10^-9 or about 3x10^-6 ms. SSDs are about 5-10 ms. Yeah those are really close. *cough*

The latency isn't even going to be driven by feet as much as the number of PCI-e swtiches have to propagate through. The real driver of latency would be putting the device as the end of 6 TB device chain. That might boost latency to perhaps being close to the same order of magnitude.






I find that claim highly dubious.

Dubious is not even having the units of measure correct.


So you prefer to have fewer hard drives because you're afraid one might fail?

As if I even said that. Professional/Enterprise environments have operational requirements. If aren't aware just keep waving your arms.


(my point was more that I hope they don't cut them down), 6 would be better.

Your point was look 10 ... oooh bigger number.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.