-
Nice thread on the future of the Mac Pro on the Videoguy’s forum
-
[Walter Soyka] “If the huge innovation in the new Mac Pro successor is really a cross-platform Intel technology, then are we really all supposed to be excited about Apple’s big contribution of a new case design?”
Maybe. Maybe not. There’s nothing much about the chips in any other Mac that’s specific to Apple yet I suspect Apple believes their case design, along with the OS, are part of the attraction. It’s certainly profitable to Apple. In their price ranges, Apple laptops and all in ones sell well.
If the case design gives it a functional advantage, rather than just aesthetic, it might be. If it’s space efficient and/or rack mountable that will help. On the other hand some would consider it a serious draw back if, outside of a 16x PCIe slot or two, all PCIe and Storage would have to be external.
We’ll know when it’s released and sales happen or not. As I mentioned elsewhere Apple has had the Cube and hockey puck mouse, On the other hand I suspect people like the MBAir, MBPRetina and, seemingly, the new iMac.
Part of the key is whether Thunderbolt on a workstation has value. That would be whether higher end users really want the ability to move peripherals to other systems as needed or would prefer a traditional workstation with space for several Drives and PCIe slots. That hasn’t been done so there’s nothing to base it on.
Don’t understatement the importance of Thunderbolt. It’s probably a key reason why professionals find MBP and iMacs usable.
The slow uptake on Thunderbolt is whole ‘nother story which had to do, in part, with the lack of resources for developers. At this point there’s a reasonable number of Thunderbolt devices for a professional to do their work.
[Walter Soyka] “how no other developers in the post space are exploiting AV Foundation.”
I had mention that in Apple’s case this is sort of a catchup. Avid and Premiere weren’t entirely dependent on Quicktime for some time. That Apple finally can break that dependency themselves is a step forward for them. Others have already done that.
As to whether AV Foundation will be something worthwhile for other media software tools, probably wont begin to be apparent until OSX 10.9 I suspect. At that point I think the Quicktime legacy will fold and FCP7 will be broken amongst other things.
The odd thing is what’s going on with Apple on Windows. While Quicktime 7.6.6 was the end on the Mac, it’s up to 7.7.3. What will be handling ProRes on Windows? Does Apple have a plan?
-
Gary — great thread you’ve started here!
[Gary Bettan] “What if Apple released a scalable solution, perhaps built around optical Thunderbolt, that would allow users to add additional performance as needed. A modular system that would include the ability to add a GPU module or additional CPUs or huge banks of memory. Now that would be sweet! So you get a base iMac Pro, then you can attach via optical thunderbolt a Mac Mini like device that accepts multiple GPU. Or an external box filled with RAM or additional CPUs.”
Multiple GPUs via Thunderbolt could be very interesting.
I’m not saying never, but this I don’t think the RAM/CPU idea matches current computer architecture (geekery follows below). A lot would have to change about the way we design computers and write software before this would work.
Systems connected via expansion busses are more likely to work like clusters (multiple separate machines networked together) than multiprocessing systems (like current dual-socket Xeons).
Current Apple technology seems to be trending away from clusters. XGrid has been untouched for years. Grand Central Dispatch is designed for symmetric multiprocessing systems; it does not support clustering.
I’d think that supporting clusters would add significant development difficulty, as you have to build a whole client/server architecture to coordinate the multiple independent systems cooperating in the cluster.
There’s also the added wrinkle that third-party developers like Avid and Adobe may not choose to make their products dependent on Apple-only technologies because they would complicate cross-platform development.
On to the geekery. As I detailed elsewhere [link]:
Current Intel architecture puts the memory controllers directly on the CPUs (eliminating slower memory controllers), and uses QPI (QuickPath Interconnect) as the CPU interconnect (replacing the old front-side bus or FSB). The CPUs also connect via QPI to to an IOH (Input/Output Hub), which in turn connects to the PCIe controller (from whence Thunderbolt flows).
The separation between the system bus and the expansion busses means they do not need to be synchronized: the CPUs do not have to be slowed down to the speed of the expansion bus. This also allows the bus’s controller to coordinate the expansion devices; without an expansion bus controller, the CPUs themselves must expend cycles intermediating expansion devices.
Thunderbolt will get faster, no doubt — but remember that in terms of expansion, Thunderbolt (external packetized DisplayPort + PCIe) is itself a subset of PCIe with additional overhead. Also remember that while PCIe/TB speeds are improving, so too are CPU interconnect speeds improving. PCIe was never designed to connect CPUs and RAM, and it’s unlikely that it will ever be the best solution for doing so.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
[Craig Seeman] “There’s nothing much about the chips in any other Mac that’s specific to Apple yet I suspect Apple believes their case design, along with the OS, are part of the attraction.”
The industrial design on the Z-series is really quite good — arguably better than the cheese grater.
I really like Mac OS X, but it turns out that Windows 7 is pretty good, too. As always, it comes down to the software you need to run. Where the software is cross-platform, I’m continually amazed at the amount of OS prejudice in creative circles.
If Apple can solve the issues of heat and space that I bring up every time we discuss the Mac Mini Pro (in other words, if Apple can overcome physics), then they are pursuing the wrong industry. They should be going after the XServe market they gave up, where lowering heat heat output and increasing computational density really matter and where people are really willing to pay for those improvements.
[Craig Seeman] “Don’t understatement the importance of Thunderbolt. It’s probably a key reason why professionals find MBP and iMacs usable. “
I certainly don’t underestimate it, and you are absolutely right that Thunderbolt makes these machines usable for all kinds of work they couldn’t otherwise do.
I just think it’s less relevant on traditional workstations where it is not adding new capabilities (beyond a network effect with the smaller systems, which is considerable).
[Walter Soyka] “how no other developers in the post space are exploiting AV Foundation.”
[Craig Seeman] “I had mention that in Apple’s case this is sort of a catchup. Avid and Premiere weren’t entirely dependent on Quicktime for some time. That Apple finally can break that dependency themselves is a step forward for them. Others have already done that. As to whether AV Foundation will be something worthwhile for other media software tools, probably wont begin to be apparent until OSX 10.9 I suspect. “
I assume Adobe made the decision to develop MediaCore because their media applications were fundamentally dependent on someone else’s library (Apple’s QuickTime) that they couldn’t develop or control.
Seems like a pretty good move now that QuickTime is being deprecated on the Mac platform and is a zombie on Windows.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
[Walter Soyka] “I wouldn’t assume that Autodesk has inside knowledge of Apple’s long-term hardware plans.”
Neither would nor do I intend to imply that.That it’s a viable platform for professionals is important.
They moved Flame away from SGI and Irix when they felt it made sense. They could do that if they felt the same way about Apple. it’s obvious to me they don’t.[Walter Soyka] “Since there are so few serious GPUs available for the Mac Pro,”
While I don’t know if they’ll be more but I suspect they’ll be better options with the MacPro replacement. If they’re going to differentiate it from an iMac, I’d think there’s be a major improvement in the GPU options. Not that I expect a lot of options but I do believe the options will include competitively better cards. It’s going to depend on whether the new box is an attractive target for nVidia I suspect. Quadro K5000 might be interesting. Support for Radeon HD 7000 series has been spotted in OS builds. It’s all rumors but when I see things like this, while obviously not about the MP replacement, one might have some hope.
https://reviews.cnet.com/8301-13727_7-57508676-263/nvidia-quadro-k5000-coming-to-the-mac-pro/
and
https://www.nvidia.com/object/quadro-k5000-mac.html -
[Paul Harrison] “You may just be onto something. An MP with a Xeon in 2013 will be able to support the Phi coprocessor which has a Jan ’13 general availability date. This combo can be sealed Apple style into a unique “magical” product which outperforms all nVidia/AMD GPU’s and with t-bolt connectivity doesn’t rely on extra PCIe slots for anything. In my speculative fever tonight I predict it will be sold as a desktop “super computer” based on the Phi becoming the goto coprocessor for top clustered true supercomputers. The Phi is supposed to run OpenCl, and CUDA ports are supposedly quite manageable, and mainstream x86 code, too. This gets Apple out of the perennial angst over multiple, esoteric gpu programming models. AMD and nVidia can fight over orders for the next iMac/MBP on-board gpu socket space. And Intel can be held somewhat more accountable by Apple for the Phi working with the Intel Xeon main cpus. It wont be cheap, with the Phi going for 2k, but as a one-size-fits-all, sizzle-core beast Apple may be able to crank out good margins at a pro-affordable price. What BTO options will be left? It will come with SSD and user upgradeable memory. I am liking this fantasy. Why not, guys?”
1) Everything I have read on Phi suggests that it can work as a coprocessor (just like a GPU) or as a cluster within its host computer — so in either case, it’s not “bolt-on” power to the host computer that software can automatically exploit (see my geekery on expansion versus system busses elsewhere in this thread). You write for Phi as if it were a coprocessor (like writing for a GPU) or write for it as if it were a cluster. I am willing and eager to be educated here.
2) Phi is Intel’s tech, not Apple’s. What’s in it for Intel to make this Apple-only? Or what does Apple have to offer that makes this a compelling and platform-specific advantage?
3) If Apple can make this a compelling and platform-specific advantage, what sort of adoption will it see outside of Cupertino? Will cross-platform developers fork their development even further to support it?
4) Apple is out of the angst of of multiple esoteric GPU programming models. They are all about OpenCL, which is not just about GPGPU, but rather about parallel, heterogenous computing (dividing up many little tasks across all kinds of different hardware). OpenCL would fit very well with Phi here, but again, OpenCL is not Apple-specific.
I am not saying this won’t or can’t happen — I’m just asking if this would really change anything if it did.
If Apple uses the same commodity hardware as everyone else, then hardware cannot be their advantage.
If Apple uses radically different hardware than everyone else, then they risk isolating their platform.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
[Craig Seeman] “Neither would nor do I intend to imply that.That it’s a viable platform for professionals is important. They moved Flame away from SGI and Irix when they felt it made sense. They could do that if they felt the same way about Apple. it’s obvious to me they don’t.”
First, I am sorry for misinterpreting your points regarding Autodesk, Apple-flavored Kool Aid, and the iMac.
I don’t think Flame on SGI/Irix and Smoke on Mac are comparable. Smoke on Mac is Autodesk’s first move in this space as a broad-release, software-only product. The software is so heavily tied to UNIX that OS X was and is the only sensible choice for doing this on the desktop today.
I don’t think anyone in this thread is arguing that the Mac is not a viable platform for professionals, but let’s also note that the Mac is not the sole platform for any high-end, demanding finishing system today.
[Craig Seeman] “While I don’t know if they’ll be more but I suspect they’ll be better options with the MacPro replacement. If they’re going to differentiate it from an iMac, I’d think there’s be a major improvement in the GPU options. Not that I expect a lot of options but I do believe the options will include competitively better cards. It’s going to depend on whether the new box is an attractive target for nVidia I suspect. Quadro K5000 might be interesting. Support for Radeon HD 7000 series has been spotted in OS builds. It’s all rumors but when I see things like this, while obviously not about the MP replacement, one might have some hope.”
Agreed. I am very hopeful about the K5000 — it could really turn the GPU on Mac situation around. I’ll be even more hopeful when Apple and NVIDIA start playing nice about their driver releases.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
[Walter Soyka] “I don’t think anyone in this thread is arguing that the Mac is not a viable platform for professionals, but let’s also note that the Mac is not the sole platform for any high-end, demanding finishing system today.”
Personally I think a Xeon based Mac with properly capable GPU is part of the equation… and people are certainly arguing about that. I think Smoke might be a good example of something that would be that much more valuable given “workstation grade” CPUs/GPUs. Not that Autodesk knows what coming but I can’t believe they’d be entirely happy if the iMac was the best they could do with Smoke.
[Walter Soyka] “Agreed. I am very hopeful about the K5000 — it could really turn the GPU on Mac situation around. I’ll be even more hopeful when Apple and NVIDIA start playing nice about their driver releases.”
The politics between Apple and GPU manufacturers is probably one of the bigger problems with Apple computers. There’s even a recent batch of rumors saying Apple may move back to AMD in the next iMac.
Personally I think the real bellwether to MacPro replacement viability will be how they handle GPU issues between pressure from Intel to Apple’s ping ponging business relationships between nVidia and AMD.
-
[Walter Soyka] “I am not saying this won’t or can’t happen — I’m just asking if this would really change anything if it did. “
1) It’s not bolt on, but with x86 it’s pretty ‘easy’. The software all of us use today will have to be modified very little, unlike other GPU models, and it’s PCIe so with fast Thunderbolt (another intel product) this could be a reality. Maybe.
2) Nothing makes it a platform advantage to Apple exclusively, unless they do something exclusive with it which seems possible given the programing structure. The Phi is PCIe, so theoretically, since the programming is easier it could be sequestered in a TBolt external box. Of course, we’d like to have more than 4x speed, but it would still be faster than not having it, I would imagine. FCPX is aleady sending separeate functions to GPU and CPU, I’m sure a Phi instrcution wouldn’t be that hard. Maybe.
3) Isn’t it cross platform? If you know how to make x86 work, can’t you amke it work on Windows or Mac? Or Linux as the articles state?
4) Not much to react to here. 🙂
[Walter Soyka] “I am not saying this won’t or can’t happen — I’m just asking if this would really change anything if it did. “
Well, it would allow almost anyone to tap in to it with already available instruction sets and wouldn’t have to rely on a separate GPU architecture or host support for that architecture. If you don’t have it, things just run a bit slower, or thing don’t run as fast.
I’ve said it before, but this seems like it would be an awesome compliment to FCPX specifically as to what it allows you to do concurrently, but I am just theorizing.
Since Apple is already tied to intel for now, this seems like a logical extension of that bond.
-
[Walter Soyka] “I don’t think anyone in this thread is arguing that the Mac is not a viable platform for professionals, but let’s also note that the Mac is not the sole platform for any high-end, demanding finishing system today.”
I would argue that it barely was two years ago. In this regard, not much has changed. I know we all like to think that the Mac was the highest performance machine on the planet, but it simply wasn’t and never has been.
“High-end” and “demanding” and “finishing” are broad terms.
-
I didn’t realize they didn’t sell PCs. I never intended to besmirch anybody. I think I was clear in saying there is nothing wrong with selling things. I don’t think I have any more wishful thinking than those who dislike Apple. I have to say there’s quite a bit of wishful thinking on the other side also.
Reply to this Discussion! Login or Sign Up