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

All Taken

macrumors 6502a
Dec 28, 2009
780
1
UK
You can't have a dongle for everything. Desktops will still have enet connectors. Hell, that's the one thing I miss on my Air and rMBP. I use a dongle every day as I have to connect to lots of stuff that is not wifi.

I see Apple's vision for the future being more cluster oriented. I think the replacement for the MacPro will be a means to cluster Mini's together using thunderbolt connections and peripherals. I just don't see them going with a big hunk of hardware at this point in their vision of the future of computing.

They've shown at every step in the past several years that their train is headed their way. You either get on or get left. More power to them. I have learned at every turn that the faster I move on and ditch old legacy stuff, the more efficient I am and the less headaches I have. (ethernet connections aside! :) )

I haven't seen anything in OS X in recent memory that points to support for clustered computing i.e A stack of Mini's. It is very very unlikely to be next year if at all in the next decade. Surely we would have seen a change in the OS?
 

GermanyChris

macrumors 601
Jul 3, 2011
4,185
5
Here
You lost me when you said 1.98 x the power. No chance. It'll be a 25% gain at most.

Where people pull these figure from is astounding.



They got them early once. When they half updated the 06 model.

He said he pulled it from passmark whether that is credible is up to you. Sandy i7's are much more powerful that the first generation core (about 60% by my calculations) I see no reason that it wouldn't apply to their xeon cousins.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
I also know of one major vendor that firmly believes Apple will be releasing an updated Thunderbolt next year that will be "at least 3x faster than the current generation".

Seem very unlikely given Intel open stated no new Thunderbolt updates until 2014.

http://www.anandtech.com/show/5405/the-first-thunderbolt-speed-bump-likely-in-2014

At this tentative stage of adoption it is unlikely that Thunderbolt would "change" the standard before it ever reached widespread adoption by the industry. A standard which is relatively constantly changing isn't going to be adopted by system and peripheral vendors.

Until the I/O hub (PCH) controllers from Intel start shipping with PCI-e v3.0 links and/or Intel adds 4 more PCI-e v3.0 links out of its CPU packages, it is unlikely that Thunderbolt will move to x4 PCI-e v3.0 input links. Without that change there is no speed change coming. Moving to v3.0 would enable a 2x bump; not 3x

Most likely Intel will demo a new Thunderbolt in 2014 and products will show in the 2015 time frame (when Intel chipsets/CPUs better support it ).


Which leads me to my pie-in-the-sky prediction: With a Thunderbolt update that is that fast, the Mac Pro can be turned into an large external 'module' for the iMac.

Quite pie in the sky even if the handwaving about 3x increase were true (which it isn't )

Interprocessor QPI (or Hypertransport links are around 150-300 Gb/s.

Thunderbolt is around 10 Gb/s

Even if TB increased 4x it will would be 26-10% of what would need for a high performance interlink for modular processors.

Thunderbolt hooks to a x4 PCI-e v2.0 link. That is not enough bandwidth for things like Processors , RAM, or even high performance GPGPU computations. It may work for graphics traffic that is optimized for relatively limited connections (e.g., x8 PCI-e links ) but that would be highly handicapped system.

This "modular Mac Pro" vibe where processors and higher end GPUs are outside the box is not motivated by sound computer organization design at all. There is I/O elements ( SATA , Ethernet , etc ) can trim off to the exterior relatively painlessly but things like CPU/RAM/GPU require much more substantive bandwidth that want Thunderbolt provides.

It's pretty clear Apple has a general disliking of the big towers

No. Apple doesn't like Macs that people don't buy in sufficiently large and growing quantities. If Mac Pros were flying off the shelves Apple would have more interest. Instead 1/5 of the threads in this Mac Pro forum are indicative of folks complaining about how newest OS X upgrades for their 6-8 year old Mac Pro. (i.e., no buyers just complainers ). And another 1/5 of posts are about how they can built less expensive Hackintosh with their trusty screwdriver.
 

Umbongo

macrumors 601
Sep 14, 2006
4,934
55
England
He said he pulled it from passmark whether that is credible is up to you. Sandy i7's are much more powerful that the first generation core (about 60% by my calculations) I see no reason that it wouldn't apply to their xeon cousins.

That's just Passmark being useless. They have the W3565 at 4,731 and the i7 960 at 6,050 and the i7 965 at 5,542.

This is a much better, although still not perfect, comparison: http://www.anandtech.com/bench/Product/45?vs=523
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
You lost me when you said 1.98 x the power. No chance. It'll be a 25% gain at most.

While a 98% jump for average increase is unlikely at the entry level, the 25% is also off the mark. Going from the Nehalem to Sand Bridge is jump up of a whole tick/tock cycle. It is usually a 11-20% jump between "tick" or "tock" improvements. 25% is likely low as being a max gain for a jump of two.

The memory will likely move from 1066 to 1600 (~ 50% bump). The instruction caching is substantially better with micro ops caching in Sandy Bridge. The L3 cache sizes are substantially bigger. The internal bandwidth is higher with new ring bus and the sustained Turbo abilities are higher.

Throw in the SATA III (6.0Gbps ) basic support in the I/O hub chipset along with a decent modern SSD and it would quite easy to leave a 2009 era based Mac Pro behind in the dust with relatively "standard" 2013 parts.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
Logicboard + ports
The major upgrades will be USB 3.0 support and thunderbolt ports. I also wouldn't be surprised if we saw more USB ports. My bet is that we see:
Front: 2x USB 3.0, 1x Firewire 800, 1x Thunderbolt, 1x headphone
Back: 4x USB 3.0, 2x Firewire 800, 1x Thunderbolt, optical in and out,

Extremely unlikely. You can't separate Thunderbolt ports that far. The external port has to be inches from the TB controller. To have "front" and "back" TB ports you'd need two TB controllers. Again extremely unlikely because it is huge waste of x8 PCI-e lanes to have two ( x4 for each controller).

If present the two Thunderbolt ports will be as close together as they are on the new retina MBP's and the iMac. That isn't going to vary on the Mac Pro.

Similarly, the 6 USB 3.0 ports are quite unlikely. In the other Mac designs flipped to USB 3.0 Apple simply bumped them all to 3.0. The current Mac Pro has 5. So the 2013 Mac Pro will at most have 5. It wouldn't be surprising if it were just three USB 3.0 ports (perhaps two front and one back) and two USB 2.0 ports.

At best 4 USB 3.0 ports. The Renesus controller only has 4 ports.

http://am.renesas.com/products/soc/usb_assp/product/upd720201/index.jsp

Using just three has better chance of not oversubscribing the bandwidth and leaves the two USB 2.0 ports on the Core chipset's I/O controller ( for mundane stuff like keyboard, mice, and security DRM keys ).

[ Sure it is possible for Apple, like some other PC motherboard vendors, to put USB 3.0 hub inside the Mac Pro to "expand" the number of USB 3.0 ports out to more than 4. I seriously don't see them choosing to add that additional complexity and cost to the design. ]


If Apple went OCD and insisted that all USB ports had to be USB 3.0 then the number of USB ports would drop to just four; even further away from six. Hopefully, they don't go OCD. It would be kind of ridiculous for the chip set's USB controller not to be used at all and keyboards/etc are a waste on USB 3.0.


It is doubtful that Apple is going to increase the amount of front/back space for built-in connectors. If get new stuff then some old stuff will go. A net increase in sockets probably won't happen.


analog in and out, 2x Gigabit ethernet, Dual-link DVI, 2x Display-port. Maybe HDMI since the MacBook Pro saw such an addition depending on videocard. Kepler will be here in full force, so HDMI is definitely likely.

HDMI only on a PCI-e GPU card. Presuming the Thunderbolt precipitates an embedded GPU to simply things it would be like the iMac where the two Display Port outputs only are hooked to the Thunderbolt controller. Sure, the Mac Mini and retina MBPs have HDMI that's a different context.


GPUs:
Kepler is the only choice and is already supported in ML.

Not the only choice. It wouldn't hurt to through AMD a bone to keep them looped into the Mac market.

However, of there is an embedded GPU (driven by Thunderbolt) then that is more likely to be Nvidia Kepler based. The same GT 650M in one of the iMacs and the MBP 15" variants would be a very likely candidate. It is already being bought in large volume for the incremental costs of adding it to the Mac Pro would be lower. Lots of software is going to get tuned to leverage a GT 650M as a GPGPU processor because it is relatively prevalent in Macs.

The discrete PCI-e GPU card is different story. Apple has had Mac Pro class boxes in the past that had both AMD and Nvidia options. It doesn't have to be 'either/or'. The only 'either/or' context would be for the entry level default configuration PCI-e card. If there is an embedded GPU then an option BTO configuration could be to ship with both x16 PCI-e sockets empty. That way folks could choose AMD or Nvidia without so much angst over their favorite GPU vendor being unfairly excluded from the Mac Pro. Folks would buy their own card.
 

MacMilligan

macrumors 6502
Original poster
Aug 2, 2012
255
8
Extremely unlikely. You can't separate Thunderbolt ports that far. The external port has to be inches from the TB controller. To have "front" and "back" TB ports you'd need two TB controllers. Again extremely unlikely because it is huge waste of x8 PCI-e lanes to have two ( x4 for each controller).

Good insight. After considering purpose and hardware requirements, TB doesn't seem to have a place on the front of the new Mac Pro.

Similarly, the 6 USB 3.0 ports are quite unlikely. In the other Mac designs flipped to USB 3.0 Apple simply bumped them all to 3.0. The current Mac Pro has 5. So the 2013 Mac Pro will at most have 5. It wouldn't be surprising if it were just three USB 3.0 ports (perhaps two front and one back) and two USB 2.0 ports.

At best 4 USB 3.0 ports. The Renesus controller only has 4 ports.

What has bugged me with Apple in the past is they say they like products that "just work," yet seem soft on USB inputs. 2+3 is not typical on a workstation. I know I've had to go through the hassle before of buying a hub to get all my devices online with a Mac and it isn't pretty. I'd think they would serve the community well by having 2+6, but it probably won't happen. I know developing on XCode, I have a keyboard, mouse, time capsule, and numerous iOS devices all plugged in at the same time to test on.


However, my stance is quite subjective, yours is the most likely outcome.

HDMI only on a PCI-e GPU card. Presuming the Thunderbolt precipitates an embedded GPU to simply things it would be like the iMac where the two Display Port outputs only are hooked to the Thunderbolt controller. Sure, the Mac Mini and retina MBPs have HDMI that's a different context.

I would hate to see a proprietary GPU card design wrapped around TB. It is a strange choice either way they go. Off-the-shelf GPUs use DL-DVI, HDMI, and DP. TB in Z77 motherboards use the integrated graphics package of the CPU to pass video, but Xeon's do not have integrated graphics. One manufacturer makes a DP-In port that the discrete GPU can be connected to that passes discrete graphics to the controller and outputs that, but it looks messy. I'm unsure how Apple will use a Xeon and discrete graphics to power a TB port.

Not the only choice. It wouldn't hurt to through AMD a bone to keep them looped into the Mac market.

However, of there is an embedded GPU (driven by Thunderbolt) then that is more likely to be Nvidia Kepler based. The same GT 650M in one of the iMacs and the MBP 15" variants would be a very likely candidate. It is already being bought in large volume for the incremental costs of adding it to the Mac Pro would be lower. Lots of software is going to get tuned to leverage a GT 650M as a GPGPU processor because it is relatively prevalent in Macs.

The discrete PCI-e GPU card is different story. Apple has had Mac Pro class boxes in the past that had both AMD and Nvidia options. It doesn't have to be 'either/or'. The only 'either/or' context would be for the entry level default configuration PCI-e card. If there is an embedded GPU then an option BTO configuration could be to ship with both x16 PCI-e sockets empty. That way folks could choose AMD or Nvidia without so much angst over their favorite GPU vendor being unfairly excluded from the Mac Pro. Folks would buy their own card.

Apple has never given much consideration to what people have already adopted. Radeon may indeed show up. Choice is a good thing, but this is Apple we are talking about haha. I'd hate to see a mobile graphics card embedded in a Mac Pro. No upgradability...

----------

You lost me when you said 1.98 x the power. No chance. It'll be a 25% gain at most.

Where people pull these figure from is astounding.



They got them early once. When they half updated the 06 model.

Its from passmark, it says that in the post. There are very few places to pull performance data from that have all of those processors. I had to find a creditable source that had both single and dual CPU charts.

----------

That's just Passmark being useless. They have the W3565 at 4,731 and the i7 960 at 6,050 and the i7 965 at 5,542.

This is a much better, although still not perfect, comparison: http://www.anandtech.com/bench/Product/45?vs=523

Hmmm, I don't even see Xeon's listed there let alone dual CPU charts.
 

Umbongo

macrumors 601
Sep 14, 2006
4,934
55
England
Hmmm, I don't even see Xeon's listed there let alone dual CPU charts.

The two processors that come up in the link are consumer versions of the Xeons you compared, there is no real performance difference between consumer and Xeon so they make for a valid comparison. Passmark is a poor thing for comparing, as you can see by the varied results for three processors with the same specifications. It also doesn't really mean anything, you know? It's a score. Not actual real world performance measurement. Like what does 5,000 points mean compared to 7,000. Time and FPS are what we measure performance in as a user.

It's okay for saying "this CPU is a bit better than this one", or "this CPU is a lot better than this one", but not much more.
 

xgman

macrumors 603
Aug 6, 2007
5,671
1,378
I think it is difficult to assume much till we know whether something in the ballpark of the current form factor will live on or not.
 

-hh

macrumors 68030
Jul 17, 2001
2,550
336
NJ Highlands, Earth
I think it is difficult to assume much till we know whether something in the ballpark of the current form factor will live on or not.

I suspect that the current form factor is going to end. The real question is how much of the existing architecture could be retained by cutting out the "right" stuff?

For example, consider slicing off the 5.25" optical bays and also two of the PCI card slots - - that would do quite a bit to cut down the size of the box. Ditto for cutting out two PCI slots .. and Apple's argument is that they get replaced with Thunderbolt ports - - probably only two (2) total, even though its IMO likely that the FW800 ports will be cut too, which means that they really should be four (4) Thunderbolt ports to be "equivalent" to such a loss of both PCI + FW800.

Next, get rid of using one of the 3.5" bays for the boot drive. Replace it with one of the laptops' SSD solutions and put it on the motherboard. This also provides the opportunity for Apple to shrink even further by removing the entire row of the four 3.5" bays under the rationale that storage expansions belong in a Promise R4 Thunderbolt cabinet. At this point, we should look to see if the remaining bits can be repackaged into a 1U form factor....

...and the last step would be to see if TB could possibly in any way be used to network multiple 1U Mac Pros together in a rack for a power cluster. If so, then the Mac Pro only needs to be sold in a single configuration: a single CPU box, since you'll add CPUs by racking on another Mac Pro...

-hh
 

Joe The Dragon

macrumors 65816
Jul 26, 2006
1,023
474
Guaranteed, FW800 will not be on the next revision of the Mac Pro. And with Thunderbolt, all you need is this cable and Hasselblads, old external drives, and whatever else will still be available:

http://store.apple.com/us/product/MD464ZM/A/apple-thunderbolt-to-firewire-adapter

Keep in mind Thunderbolt is a path to a lot of accessories that Apple can then nix from the box... Lending further possibilities to my "Pro Module" concept. And if the next revision of Thunderbolt really is 3x the throughput of the current one... That would blow everything wide open and change the idea of what a professional computer tower is for all of us.

Tim Cook played coy in his email from August (where he said Apple was working on something really great). A 'simple' form factor change with updated memory and processors doesn't fit this statement very well. Either he's full of it, or Apple really does have a major Fusion Drive + Next Generation Thunderbolt card up its sleeve.

Signs are pointing to yes.

----------



I don't think they'll drop the Ethernet ports either, but if they did, this would work just fine:

http://store.apple.com/us/product/MD463ZM/A/thunderbolt-to-gigabit-ethernet-adapter

Will need a lot more TB ports or a Ethernet box with TB pass though and even then TB is only like pci-e X4 linked.

A video card will need a Full X8 / X16 setup.
 

theSeb

macrumors 604
Aug 10, 2010
7,466
1,893
none
I wouldn't use Passmark as a comparison tool for tomatoes, never mind CPUs. Also, those that are trying to use Geekbench, don't forget that it's a measure of how fast the CPU and memory can sprint. Not how fast it can perform a real world task where considerations like Turbo Boost not always being available due to thermal deficiencies do not come into the picture.
 

DJenkins

macrumors 6502
Apr 22, 2012
274
9
Sydney, Australia
I don't know why people keep going with the 'stack' concept, at least for this next release. To me it seems like computers get more efficient by having everything closer together, not splitting it apart into separate boxes. The limitations of TB have also been pointed out as an unrealistic interface for this design.

People really are pinning a lot of responsibility on TB for replacing port functionality. But there is no reason for this on a mac pro. The space saving challenge doesn't apply here like it does a laptop. There is more than enough room for Ethernet and many USB ports, the only limitation should be bandwidth.

I also hate the idea of having external thunderbolt hubs and expansion boxes, replacing everything that could easily be supported in the tower. What an ugly idea. It's a freaking tower for goodness sake, they could put 12x USB, 4x firewire, 4 x TB and 8 x PCI slots and I would only then think they were actually taking the form factor seriously.

Although being apple they will probably remove everything as a 'cool' point of difference, and try and convince people to buy new wireless keyboards and mice to save USB ports.

I know deconstruct60 said earlier about people making hackintoshes with their trusty screwdriver (not sure what the tone was on that comment.. good/bad?) but it will be disappointing if whatever gets released can still be beat by these home built machines running on ~6 year old technology.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
What has bugged me with Apple in the past is they say they like products that "just work," yet seem soft on USB inputs. ... I'd think they would serve the community well by having 2+6, but it probably won't happen. I know developing on XCode, I have a keyboard, mouse, time capsule, and numerous iOS devices all plugged in at the same time to test on.

It isn't soft but taking recognition of what the users actually use. Use of that many USB sockets is unusual even in workstation market segment.

Most folks with a large collection of USB devices have one that is a also a hub. For example, Monitors and keyboards typically are also USB house. Mouse -> keyboard -> computer is a far more sensible configuration. Likewise for folks that move the workstation off the desk the monitor is closer and hence mouse+keyboard -> monitor -> computer.

Time capsule? That's likely a HDD. FW or eSATA/SAS deliver higher performance. USB Thumb Flash drives I can see. That's one reason why USB 3.0 on the front would be useful.

Numerous iOS all plugged in at the same time is a corner case.


I would hate to see a proprietary GPU card design wrapped around TB. It is a strange choice either way they go.

It is not so much proprietary as not removable; hence embedded. The essential components from the standard reference card design are just moved to the motherboard. There is not particularly proprietary about it.


Off-the-shelf GPUs use DL-DVI, HDMI, and DP.

Not all of them. Lower end card typically don't have all three. The mid-high price range ones that Apple has used as of late on the Mac Pro it is more common.

TB in Z77 motherboards use the integrated graphics package of the CPU to pass video, but Xeon's do not have integrated graphics.

Xeon E3's do. Apple isn't likely to use them. However, in a couple of years it would not be surprising for the Xeon E5 1600 series to get integrated graphics also (probably after the Haswell/Broadwell tick/tock ).
As OpenCL matures and is adopted, the transistor budget gets bigger , and memory bandwidth grows it makes to add a iGPU as a faster augment for "embarrassing parallel" computation tasks more so then the bumps you can get by adding SIMD stuff like AVX or 'yet another x86 core' ( which at 8 cores is quite substantial for a very large variety of workloads).





I'm unsure how Apple will use a Xeon and discrete graphics to power a TB port.

Just like they currently do on the iMac and MBP 15" models. It is an already solved problem that is deployed. It is amazing how folks in the Mac Pro forum turn this into some "land a man on the moon and safely return" or "Manhattan project" endeavor.


I'd hate to see a mobile graphics card embedded in a Mac Pro. No upgradability...

If there are two (or more) empty x16 PCI-e slots upgradability would still be extremely present in the overall system.

The MBP 15" variants have two GPUs. The iMac has two GPUs. It would actually be in alignment if the Mac Pro (priced above both of those) also had two GPUs.

This isn't about functionality, usefulness, or upgradability. It is really about control. The desire to rip out a component Apple put in because the owner comes up with something better. The notion that the Mac Pro has to be some erector set skeleton is likely at odds with Apple's view of the Mac Pro.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
I suspect that the current form factor is going to end.

Change from the general form factor ( rectangular box with removable side, PCI-e slots, and drive slots ) is not likely. Change from approximately the same dimensions and "holes"/sockets on the box; yes.

For example, consider slicing off the 5.25" optical bays

That is pretty tough to do if want to keep a symmetrical design since the power supply is the largest principal member of that top thermal zone. Even if you remove the ODDs, it is still there.

Apple could re-purpose the 5.25" bays. Users somewhat commonly already do that for 2.5" and 3.5" drives. If the Mac Pro mutated to just simply afford what some folks are already doing that would be simply adapting to the market. The Mac Pro needs more 2.5" bays than it has now. It keep it locked just on four 3.5" bays is to keep the design firmly rooted in the past; not the future.


and also two of the PCI card slots - - that would do quite a bit to cut down the size of the box.

Why nuke two PCI-e card slots? It really going to save all that much space if leave two very power hungry x16 PCI-e slots behind.

First, The fan(s) required to cool those are going to soak up about as much space as the fan to cool the 4 slots if go with large diameter , low noise options.

Second, if punting x4 slots for Thunderbolt the likely defacto addition of a embedded GPU ( GPU , VRAM , etc. ) likely consumes just as much space (if not more ) as the two PCI-e slots took.


Third, current Mac Pro splits x4 PCI-e v2.0 bandwidth between the two x4 PCI-e slots with a switch. The vast majority of mainstream PC designs use switches to split x16 bandwidth into two x16 physical but x8 electrical slots as part of the basic board design.


Ditto for cutting out two PCI slots .. and Apple's argument is that they get replaced with Thunderbolt ports - - probably only two (2) total,

since one x4 Thunderbolt controller corresponds to one x4 PCI-e slot it is unlikely that Apple woud claim that two slots was equal to two TB ports. Those two ports only represent one slot.


even though its IMO likely that the FW800 ports will be cut too,

FW800 is probably on a slippery slope but so are the audio sockets. The number of FW800 sockets may go down but still may be above zero. If adding Thunderbolt and attaching it to the PCH's x8 PCI-e v2.0 limited bandwidth then the x1 PCI-e link for FW800 would probably get swapped for a x1 PCI-e link for USB 3.0 unless some other x1 link got squeezed off.

Even if they did for the hard core "can only buy Apple branded stuff" folks would buy the Thunderbolt Display/Docking Station that Apple sells. It puts just as many, if not more, ports back on the combined system as Apple would take away.

A one-for-one correlation between TB and the Ethernet/Firewire/USB sockets is just deeply flawed. The whole "one socket to rule them all" was just marketing kool-aid that Intel (and some others) initially floated to get traction. TB is not the universal socket replacement. It never has been outside of the kool-aid spin-meisters zone.



which means that they really should be four (4) Thunderbolt ports to be "equivalent" to such a loss of both PCI + FW800.

Four TB ports requires two TB controllers. One that is a huge waste of x8 PCI-e lanes. Second, it complicates and limits the display flexibility for the both controllers ( likely at least one of the controllers would be limited to just one Display Port input ).


Next, get rid of using one of the 3.5" bays for the boot drive. Replace it with one of the laptops' SSD solutions and put it on the motherboard.

The 3.5" bays could be moved into the 5.25" bays. Or a subset of the 3.5" bays could be moved one of those (e.g., 1-2 into bay 1 ) and the other bay used for 2.5" drives (e.g. 4 into bay 2 ).

I would suspect at least two if not all four 3.5" bays get moved out of the PCI-e card thermal zone. If Apple is adding an embedded GPU and higher TDP cooling for PCI-e cards into that zone those devices should be moved out to help balance the increase.

This also provides the opportunity for Apple to shrink even further by removing the entire row of the four 3.5" bays under the rationale that storage expansions belong in a Promise R4 Thunderbolt cabinet.

Low to medium workstation storage requirements can be done with 2.5" drives which require less space. There is zero need to appeal to external Thunderbolt boxes.

However, for very large data storage that is where folks go already.

At this point, we should look to see if the remaining bits can be repackaged into a 1U form factor....

Apple had a 1U form factor Mac. If that was a magically highly successive market property they would have kept it. It isn't.

As long as there is a 900-1000W power supply feeding 200-400W of CPU/RAM demands and 300-500W of PCI-e card demands the primary drivers of the Mac Pro form factor still exist if Apple continues to target the same relatively low noise production design constraints.

Waving at trading PCI-e slots for an every larger number of TB ports does absolutely nothing to alleviate that core design issue.

The Mac Pro should be a couple inches shorts for it is not gratuitously horizontal rack hostile. But there is no "thin"

If Apple threw high end GPU cards and Xeon E5's and high end GPUs out the window then yeah sure they could do a 1-2 slot , at least 1/3 smaller , Mac Pro like derivative with a Xeon E3 .


...and the last step would be to see if TB could possibly in any way be used to network multiple 1U Mac Pros together in a rack for a power cluster.

A workstation trend is for the cluster to move inside the box. Not to push things outside. The Xeon Phi , Nvidia K20 cards, and AMD Firepower 9000 cards can all bring approximately 1TFLOP DP performance inside the box. High Triple digit GFLOP DP performance is doable with more mainstream GPU cards that would likely land in a updated Mac Pro progression anyway. If the Mac Pro cannot participate in that then it will be whipped in the workstation market going forward. If Apple is moving the case design that they will stick with for the next 5-10 years then it needs to be one that tracking the current trends. Not hiding in the flawed "1U pizza boxes will rule the world" past that is out of step with current trends.
 

d-m-a-x

macrumors 6502a
Aug 13, 2011
510
0
Fire Wire 800

I hope they at least leave us with the pci option. The thunderbolt adaptor will definitely slow things down.

from phase one

"The adaptor has been tested and found to work adequetly for tethered capture with certain Phase One and Leaf digital backs. However, there are some workflow concerns to consider when using the Adaptor;

Due to the additional "translation" of the Firewire communication through the Thunderbolt Terminal, capture speed will suffer some performance.

Due to the loss of capture speed performance, it is possible to fill the buffer of the IQ, Credo and P+ backs much faster than with a standard Firewire Connection. Discretion when shooting quickly is suggested.

Due to Power Consumption with this setup it is recommended to have additional batteries as well as have the Apple system connected to a power source."
 

goMac

Contributor
Apr 15, 2004
7,662
1,693
I hope they at least leave us with the pci option. The thunderbolt adaptor will definitely slow things down.

from phase one

"The adaptor has been tested and found to work adequetly for tethered capture with certain Phase One and Leaf digital backs. However, there are some workflow concerns to consider when using the Adaptor;

Due to the additional "translation" of the Firewire communication through the Thunderbolt Terminal, capture speed will suffer some performance.

Due to the loss of capture speed performance, it is possible to fill the buffer of the IQ, Credo and P+ backs much faster than with a standard Firewire Connection. Discretion when shooting quickly is suggested.

Due to Power Consumption with this setup it is recommended to have additional batteries as well as have the Apple system connected to a power source."

I see a new Mac Pro having at least 3 PCI slots, likely the full four already present.

Thunderbolt is not fast enough to replace a lot of cards.
 

Phrygian

macrumors regular
Nov 26, 2011
196
0
This is Apple.... If they change the colour of a device it's described as magical. Tim will exaggerate any new Apple product as expected, I hope it is something major but don't think for a second that keeping the form factor and adding thunderbolt isn't something Tim/Apple would describe as really great. I'm pretty sure it would be magical.

So much this. IO6 and Iphone 5 are a perfect example.

"we made it taller and ****ed you your maps, its a revolution up in here!"
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
I hope they at least leave us with the pci option. The thunderbolt adaptor will definitely slow things down.

from phase one

"...
Due to the additional "translation" of the Firewire communication through the Thunderbolt Terminal, capture speed will suffer some performance.
...."

A bit dubious root cause analysis since Thunderbolt never sees the Firewire protocol data. The FW controller passes back PCI-e encoded data to Thunderbolt for transport. Wouldn't be surprising if that "slow down" appeared for a PCI-e card with a previously unseen FW controller on it.


The FW driver to the controller in the adapter by be a bit under (or over ) optimized ( should be the same controller used inside macs though, but if different that is a more likely root cause issue). Similarly if stretching power serial power draw through TB and then through FW may also push some glitches into the mix.

But "translating" to Thunderbolt. That basically opaque and hardcoded in the TB hardware.
 

d-m-a-x

macrumors 6502a
Aug 13, 2011
510
0
A bit dubious root cause analysis since Thunderbolt never sees the Firewire protocol data. The FW controller passes back PCI-e encoded data to Thunderbolt for transport. Wouldn't be surprising if that "slow down" appeared for a PCI-e card with a previously unseen FW controller on it.


The FW driver to the controller in the adapter by be a bit under (or over ) optimized ( should be the same controller used inside macs though, but if different that is a more likely root cause issue). Similarly if stretching power serial power draw through TB and then through FW may also push some glitches into the mix.

But "translating" to Thunderbolt. That basically opaque and hardcoded in the TB hardware.

Ever since the mbp and imac came out, a lot of people have unreliable connections with the tb adaptor. There are countless threads
 

deconstruct60

macrumors G5
Mar 10, 2009
12,227
3,827
Ever since the mbp and imac came out, a lot of people have unreliable connections with the tb adaptor. There are countless threads

Number of posts don't point to "translation" as being a root cause. In fact many threads point to limited bus power as being an issue to medium format camera backs. The adapter is limited to 7W in part because TB maxes out at 10W ( but also has to power the FW controller itself).

For better or worse a subset of devices were designed around maximum power draw from FW ( which could actually go quite high in Apple set ups). As a relatively high power distribution network none of these connector schemes are really a good idea. That doesn't mean some designer didn't leap onto that slippery slope.

If find references to FW devices with their own AC power that are having bandwidth or latency problems that would lend some credibility to "translation" problems. But as long as this is confined to devices with large power draws on the bus it will stand as a dubious root cause classification.
 

d-m-a-x

macrumors 6502a
Aug 13, 2011
510
0
Number of posts don't point to "translation" as being a root cause. In fact many threads point to limited bus power as being an issue to medium format camera backs. The adapter is limited to 7W in part because TB maxes out at 10W ( but also has to power the FW controller itself).

For better or worse a subset of devices were designed around maximum power draw from FW ( which could actually go quite high in Apple set ups). As a relatively high power distribution network none of these connector schemes are really a good idea. That doesn't mean some designer didn't leap onto that slippery slope.

If find references to FW devices with their own AC power that are having bandwidth or latency problems that would lend some credibility to "translation" problems. But as long as this is confined to devices with large power draws on the bus it will stand as a dubious root cause classification.

The backs have their own power source... The only mistake the designers made was to drink the kool aid made by apple
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.