Be Tim Cook for a day, which GPU brand for new Mac Pro?

[tuxon86]

And Nvidia is doing what with Tegra ? AMD has a x86 license so sticking mainly with it. Nvidia doesn't. They are both betting a large fraction of the farm on CPU+GPU more SoC solutions.

So is Intel.

Why are you so defensive?

AMD as repeatedly said they are moving to an APU/SoC solution based for the future. NVidia hasn't made such statement. NVidia hasn't scrapped half of the business that made them great to go instead with APU like AMD did by ending their desktop CPU line. AMD has abdicated to Intel on that front so what is to say they won't abdicate on the dGPU front as well and instead target the APU/SoC market only....

 

[liquid stereo]

Other than you and a few geeks,...

Who would you sell it to? The 200 people per month who currently buy towers?

Tim for a day?

I would cancel the trashcan and order development of the two following Mac Pros;

Full-sized Mac Pro. Same case, new mainboard, TB2, USB3, eSATA, FW800, SATA3, PCIe 3, single or dual processors, additional SATA ports in the upper/lower bays, choice of nVIDIA or AMD. Single Xeon starts at $2500.00.

Mini-tower Mac Pro. Smaller case with Mac Pro look, all modern internals, two full-sized PCIe 3 slots, one optical drive bay (optional), one internal 2.5 inch bay and one 3.5 bay., choice of nVIDIA or AMD. Fast quad i5/Iris Pro starts at $1450.00. Single Xeon starts at $1850.

That's what I would do.

 

[goMac]

AMD as repeatedly said they are moving to an APU/SoC solution based for the future. NVidia hasn't made such statement. NVidia hasn't scrapped half of the business that made them great to go instead with APU like AMD did by ending their desktop CPU line. AMD has abdicated to Intel on that front so what is to say they won't abdicate on the dGPU front as well and instead target the APU/SoC market only....

NVidia hasn't made such a statement on the PC because they aren't an x86 vendor (and you'd be foolish if you didn't think they were eagerly looking for a way to fix that right now.)

On handheld, NVidia is putting everything into APU. They just released a handheld console without a dGPU. Again, the only reason they haven't done this on PC is because they don't have an x86 license.

But the entire reason they're shoveling as much as they can into the SoC market is because they're thinking the dGPU market is going to die and they're going to be left holding the bag.

 

[deconstruct60]

Thunderbolt doesn't really work with internal PCIe GPUs. Can't have it both ways.

It doesn't work well if try to tightly couple them. Don't really have to do tight coupling any more than two standard PCI-e socket GPU cards have to be coupled.

Some folks will which generally leads to proprietary solutions and some folks won't.

It isn't so much Thunderbolt as an additional design criteria that practically all video flows through TB that is the hang up. Attach that to a system and minimize Rube Goldberg aspects and the GPUs go embedded.

 

[goMac]

It doesn't work well if try to tightly couple them. Don't really have to do tight coupling any more than two standard PCI-e socket GPU cards have to be coupled.

Some folks will which generally leads to proprietary solutions and some folks won't.

It isn't so much Thunderbolt as an additional design criteria that practically all video flows through TB that is the hang up. Attach that to a system and minimize Rube Goldberg aspects and the GPUs go embedded.

Right. That's kind of implicit in the new Mac Pro's design. The cards are still technically PCI-E cards, they just lack the video output and are rerouting it somewhere else. Either directly to the Thunderbolt controller or back across the PCI-E bus.

 

[tuxon86]

NVidia hasn't made such a statement on the PC because they aren't an x86 vendor (and you'd be foolish if you didn't think they were eagerly looking for a way to fix that right now.)

On handheld, NVidia is putting everything into APU. They just released a handheld console without a dGPU. Again, the only reason they haven't done this on PC is because they don't have an x86 license.

But the entire reason they're shoveling as much as they can into the SoC market is because they're thinking the dGPU market is going to die and they're going to be left holding the bag.

Tegra's future is linked with Cuda. Tegra isn't there to replace the dGPU but to complement it.

http://www.brightsideofnews.com/news/2013/3/20/new-tegra-roadmap-reveals-logan2c-parker-and-kayla-cuda-strategy.aspx

AMD on the other hand as already dropped part of their business for APU.

 

[thekev]

That is partially because OpenCL isn't as complete as it should be. OpenCL 1.2 and SPIR should help close the gap on some of this pre-test to see which preformed package to load.

http://www.anandtech.com/show/7161/...-opengl-44-opencl-20-opencl-12-spir-announced

SPIR should reduce that need except for extremely hand optimized stuff.

I was unaware that OpenCL 1.2 addressed that.

The entry level 21" perhaps. ( to drop back down to the $1,199 level ). But the problem Apple has with the iMac though is all the substantially non integrated graphics at lower price points in the Overal desktop PC market. Higher than average cost host to get supported by higher than average value.

Really depends upon how cost effective the dGPU + VRAM vs. Intel's integrated with eDRAM get and the relative $/Performance ratios.

I don't think the desktop cpus have the eDRAM option. It seems to be limited to relatively expensive cpus designed for notebooks. I don't expect it on the mini either due to cost. With the 21" imac, I didn't mean in the next revision. I was thinking further out, but I suspect it will be a case where they move discrete graphics to 27" only at some point. The 21" is probably the highest volume desktop unit, so I don't think they would be too aggressive on that.

Well they tend to dominate in the decreasing pro-gamer market for the moment. Of course the bad economy that we are presently going through means that people tend to buy the cheaper system or gadget which make AMD decision to go deep in the APU/SoC market a good one. Of course NVidia is also pushing is Tegra 3 and 4 in the smartphone/tablet/device also.

I didn't forget about the Tegra. I just didn't comment on it, because I don't know how well that research integrates with the Quadro/Tesla lines where they derive their higher margins. It's certainly a long term bet on the direction of technology.

 

[scottsjack]

Who would you sell it to? The 200 people per month who currently buy towers?

Better than the 25 people per month who might by the trashcan!

Also the towers have to be sell-able with modern, competitive features that users want.

 

[deconstruct60]

Why are you so defensive?

Just pointing out the truth. If that makes you feel defensive that says more about you than me.

AMD as repeatedly said they are moving to an APU/SoC solution based for the future. NVidia hasn't made such statement.

Chuckle. Nvidia's Q1 press release...

Quote from CEO

"... "We grew our GPU and Tegra Processor businesses. We are sampling production silicon of the Tegra 4 platform which includes our 4G LTE modem. And we created new pillars for long term growth with Project SHIELD and NVIDIA GRID -- first-of-their-kind devices that will extend our leadership in visual computing into mobile and the cloud." ... "
http://nvidianews.nvidia.com/Releas...nnual-and-Fourth-Quarter-Fiscal-2013-91d.aspx

Yeah that sure sounds like desktop GPU is the only primary focus.

and later in same press release"

"...
Fourth Quarter Fiscal 2013 Highlights:

NVIDIA's customers brought three Windows RT devices to market -- Asus VivoTab RT, IdeaPad Yoga 11 from Lenovo, and Microsoft Surface RT
NVIDIA announced Project SHIELD™, a unique Android gaming device that will ship in the second quarter of fiscal 2014
NVIDIA launched Tegra® 4, the world's fastest mobile SOC and the first quad-core A15 SOC
NVIDIA continued to drive the streaming of gaming from the cloud by signing deals with six middleware providers that will supply GRID™ gaming technology to service operators worldwide
NVIDIA officially launched the Tesla® K20 family of GPU accelerators, making the technology behind the world's fastest supercomputer, Titan, available to all ... "

Yep.... desktop discrete GPUs right there in the major highlights Nvidia wanted folks to walk away with .... not.

Win RT ... that is ARM SoC devices.
Project Shield .. integrated implementation gaming platform.
Tegra 4 ... again SoC devices ( note it incorporate a LTE so this is expansion out of CPU+GPU into digital radio IP also. )
Cloud gaming ( because you have to super duper local desktop GPU so you can simply get the bits from the cloud. )
K20 ... Sure K20 derivatives arriving now in mainstream desktop products.

So what? 1 out of 5 highlights for desktop GPUs.

NVidia hasn't scrapped half of the business that made them great to go instead with APU like AMD did by ending their desktop CPU line.

AMD hasn't scrapped their desktop line up any more than Intel has.
Next generation Intel mainstream desktop line up going BGA. Yeah AMD is out there by their lonesome. Major changes are coming in desktop CPU line ups from all directions. Spinning that it is primarily AMD is just smoke.

AMD has abdicated to Intel on that front

They haven't. They have not executed a evolutionary path on their x86 cores or their foundry relationships very well. The latter has adversely impacted their APU path as much as the x86 cores.

They gotten beaten by better execution and focus in terms of competitive products. That's happened before. Given the revolving door at the leadership helm at AMD hasn't exactly helped either.

There are areas where AMD feels they have more of a competitive edge and they are putting more focus on those. Unlike Intel they don't have the bank account to pursue 4-5 major development areas at once.

so what is to say they won't abdicate on the dGPU front as well and instead target the APU/SoC market only....

Primarily because there is shrinking differences between dGPU and the integrated ones. The dGPU allows bigger implementations, but 1-2 process shrink generations later largely the same stuff in the dGPU is in the embedded/integrated ones.

Trends like OpenCL and PCIe v3 supporting uniform memory addressing and eDRAM/stack-DRAM on chip packages only all the more so.

 

[tuxon86]

Just pointing out the truth. If that makes you feel defensive that says more about you than me.

Your truth, not the whole truth.

Chuckle. Nvidia's Q1 press release...

Quote from CEO

"... "We grew our GPU and Tegra Processor businesses. We are sampling production silicon of the Tegra 4 platform which includes our 4G LTE modem. And we created new pillars for long term growth with Project SHIELD and NVIDIA GRID -- first-of-their-kind devices that will extend our leadership in visual computing into mobile and the cloud." ... "
http://nvidianews.nvidia.com/Releas...nnual-and-Fourth-Quarter-Fiscal-2013-91d.aspx

Yeah that sure sounds like desktop GPU is the only primary focus.

and later in same press release"

"...
Fourth Quarter Fiscal 2013 Highlights:

NVIDIA's customers brought three Windows RT devices to market -- Asus VivoTab RT, IdeaPad Yoga 11 from Lenovo, and Microsoft Surface RT
NVIDIA announced Project SHIELD™, a unique Android gaming device that will ship in the second quarter of fiscal 2014
NVIDIA launched Tegra® 4, the world's fastest mobile SOC and the first quad-core A15 SOC
NVIDIA continued to drive the streaming of gaming from the cloud by signing deals with six middleware providers that will supply GRID™ gaming technology to service operators worldwide
NVIDIA officially launched the Tesla® K20 family of GPU accelerators, making the technology behind the world's fastest supercomputer, Titan, available to all ... "

Yep.... desktop discrete GPUs right there in the major highlights Nvidia wanted folks to walk away with .... not.

Win RT ... that is ARM SoC devices.
Project Shield .. integrated implementation gaming platform.
Tegra 4 ... again SoC devices ( note it incorporate a LTE so this is expansion out of CPU+GPU into digital radio IP also. )
Cloud gaming ( because you have to super duper local desktop GPU so you can simply get the bits from the cloud. )
K20 ... Sure K20 derivatives arriving now in mainstream desktop products.

So what? 1 out of 5 highlights for desktop GPUs.

Nowhere in this does it says that NVidia is droping out of the dGPU market anytime soon. On the contrary, the fact that Tegra is moving more and more toward Cuda integration means that their SoC solution will work in tandem with their more powerful brothers.

AMD hasn't scrapped their desktop line up any more than Intel has.
Next generation Intel mainstream desktop line up going BGA. Yeah AMD is out there by their lonesome. Major changes are coming in desktop CPU line ups from all directions. Spinning that it is primarily AMD is just smoke.

Oh you mean you missed the whole droping of dedicated mainstream desktop CPU and going to crappy under performing APU.

http://www.tomshardware.com/news/amd-intel-cpu-apu-processors,15741.html

What is still being released are the leftover design. They'll be completely out in 2015.

Trends like OpenCL and PCIe v3 supporting uniform memory addressing and eDRAM/stack-DRAM on chip packages only all the more so.

People don't buy computer or hardware for what they could be in the future. They buy them for what they can do today! And as of today OpenCL is lagging behind Cuda.

 

[RoastingPig]

780 3gb would be sick.

 

[deconstruct60]

Better than the 25 people per month who might by the trashcan!

While it is unclear just how well or badly the new cylinder form factor will do long term and after pricing is introduced, it probably will not be a 87% decrease in Mac Pros sold.

There are a substantial number of defectors, but even without pricing information the largest single category here is "buy".

https://forums.macrumors.com/poll.php?do=showresults&pollid=10273

Sure, the "no's" are fragmented but the collectively still don't add up to 50%. Apple probably get a double digit drop in volume. Long term whether that is successful or not hinges on if the drops continue or starts to grow at about (or better) than other Mac products do a couple years out.

 

[goMac]

Tegra's future is linked with Cuda. Tegra isn't there to replace the dGPU but to complement it.

http://www.brightsideofnews.com/news/2013/3/20/new-tegra-roadmap-reveals-logan2c-parker-and-kayla-cuda-strategy.aspx

AMD on the other hand as already dropped part of their business for APU.

Sure... I guess you could use Tegra as a co-processor. That doesn't change that Tegra is mostly intended as a tablet or phone GPU. NVidia is also kind of betting Windows RT will work out so they can push PC people onto Tegra (so far looking like that's not going to work out.)

Nowhere in this does it says that NVidia is droping out of the dGPU market anytime soon. On the contrary, the fact that Tegra is moving more and more toward Cuda integration means that their SoC solution will work in tandem with their more powerful brothers.

Again, the only reason why this is the case is because they are the only GPU maker without an x86 design. It's not because they want to be in this position, it's because they are stuck. Their options are either they pick up an x86 design somewhere or they really push ARM, and so far they are going with really pushing ARM.

Tegra working with other chips isn't a well founded business plan, it's desperation because they have no other choice. And that plan isn't even around dGPUs.

 

[tuxon86]

Sure... I guess you could use Tegra as a co-processor. That doesn't change that Tegra is mostly intended as a tablet or phone GPU. NVidia is also kind of betting Windows RT will work out so they can push PC people onto Tegra (so far looking like that's not going to work out.)



Again, the only reason why this is the case is because they are the only GPU maker without an x86 design. It's not because they want to be in this position, it's because they are stuck. Their options are either they pick up an x86 design somewhere or they really push ARM, and so far they are going with really pushing ARM.

Tegra working with other chips isn't a well founded business plan, it's desperation because they have no other choice. And that plan isn't even around dGPUs.

But as demonstrated, AMD are getting out of the CPU business so repeating that NVidia doesn't have an x86 solution is a moot point. AMD on the other hand build their company reputation with their x86 CPU. They bought ATI to get into the GPU market while NVidia as always been a GPU business.

 

[MacVidCards]

2 to 1 ratio is an even split to you?

This would be a good example of "spurious logic" I mentioned in another thread.

In a Presidential election, one candidate getting twice as many votes as the other would be a "landslide". Here in our Apple spun world it is "evenly split".

 

[goMac]

But as demonstrated, AMD are getting out of the CPU business so repeating that NVidia doesn't have an x86 solution is a moot point. AMD on the other hand build their company reputation with their x86 CPU. They bought ATI to get into the GPU market while NVidia as always been a GPU business.

AMDs not getting out of the CPU business. What on earth would make you think that? They just inked deals to supply the CPUs for the Xbox One and PS4.

 

[deconstruct60]

I was unaware that OpenCL 1.2 addressed that.

Both aspects, the SPIR extension, and further heterogeneous interoperability with OpenCL 2.0 are new. Technically SPIR is still provisional for 6 months for no one is putting it into production in 2013. It will take sometime to deploy.

I don't think the desktop cpus have the eDRAM option.

This round. I suspect one reason that sizable fraction of Intel's desktop CPUs are going BGA next round (v4 Broadwell ) is that they will be CPU-GPU + eDRAM packages similar to the mobile versions this round. The line between upper level mobile and lower level desktop will blur like it has in the Nvidia GPU line up. Intel needs another round of improvements to get closer to entry level Desktop performance on a broader range of performance envelopes.

The BGA option is targeted at all-in-ones of which the iMac is an instance of.

It seems to be limited to relatively expensive cpus designed for notebooks.

Partially because Intel is getting its feet wet easing back into the DRAM business.

http://www.realworldtech.com/intel-dram/

I don't think anyone has said that was wrong and the key take away is that Intel is fabbing the eDRAM. The more they sell the better margins they make and the busier they keep their fabs running even with the decline in PC sales. ( by taking fab work away from VRAM folks ). If legacy PC sales keep tracking down expect Intel to be all over that like cold on ice.

In about a year,, they should operating pretty smoothly to start to attack on cost competitiveness. It is up the dGPU and VRAM folks if they can counter, but Intel definitely is hunting to take that business.

I didn't forget about the Tegra. I just didn't comment on it, because I don't know how well that research integrates with the Quadro/Tesla lines where they derive their higher margins. It's certainly a long term bet on the direction of technology.

At the top end it is slightly conflicted. The performance at all cost runs counter to performance/watt. Tegra has its own short term drama with Qualcomm/Adreno and Imagination Tech/PowerVR. If they don't counter them what trickles down long term from Quadro/Telas isn't going to matter. So far in 2013 Nvidia is having problems.( Nexus 7 ... lost as latest example).

Apple (PowerVR so far) and Samsung ( seemingly switching to ARM Mali ) and the other major phones on Qalcomm/Adreno means CUDA matters very little in the handeld computer space. OpenCL perhaps eventually. CUDA not. The open ARM SoC market is not largely looking for proprietary single vendor software/hardware stacks.

 

[deconstruct60]

In a Presidential election, one candidate getting twice as many votes as the other would be a "landslide". Here in our Apple spun world it is "evenly split".

There is no "it doesn't matter who is elected" option in a President election.

Far closer analogy to the actual situation is that overriding a veto takes far more than just half the votes.

And frankly these landslide victories of 2-3% percentage points .... well that go covered a long while back in 1981.

http://www.youtube.com/watch?v=xSOp507HJMA

 

[seveej]

I (as Tim) would decree, that there should be a BTO-option of the mac pro utilizing only one (≈150€ -class) desktop (not workstation) GPU.

Also, I'd make patent the connection scheme of the GPU cards and make the standard openly available so that anyone could build cards for the nMP.

Oh, yeah. I'd also have them ship one to my home address, with 27" LCD.

RGDS,

 

[goMac]

I (as Tim) would decree, that there should be a BTO-option of the mac pro utilizing only one (≈150€ -class) desktop (not workstation) GPU.

I think two GPUs comes from the number of Thunderbolt ports. One isn't enough to power six Thunderbolt ports.

What I'd hope to see is being able to mix and match GPUs. Maybe one low end one high end?

Also, I'd make patent the connection scheme of the GPU cards and make the standard openly available so that anyone could build cards for the nMP.

I could see this maybe happening with both the card slot and SSD connector. More likely with the SSD connector...

The GPUs are going to be funny in that the SSDs are mounted on them. Bleh.

What I could see happening eventually if CPUs and GPUs are combined is that instead of one CPU board and two GPU boards, you have two combo CPU/GPU boards, and all the storage moves off onto the third board.

 

[deconstruct60]

Nowhere in this does it says that NVidia is droping out of the dGPU market anytime soon.

Care to pull out the press release where AMD says they are dumping GPUs?

Never said they were dropping out. But frankly they are presuing more than one avenue to the future. Futhermore they are going out of their way to point out that they are not a one-trick-pony to the investors and analysts. Being non-diversified is a problem they have been tagged with, that why it is being addressed and highlighted in their quarterly results.

On the contrary, the fact that Tegra is moving more and more toward Cuda integration means that their SoC solution will work in tandem with their more powerful brothers.

Chuckle. Yeah what share of the andriod market does Nvidia have? Something similar to AMD's of the PC market. CUDA in ARM SoC solutions a dominate factor? You are smoking something.

Oh you mean you missed the whole droping of dedicated mainstream desktop CPU and going to crappy under performing APU.

http://www.tomshardware.com/news/amd-intel-cpu-apu-processors,15741.html

Oh you mean the part of the article where it says

"... It would be a bit premature to assume that AMD is not developing chips that are at least somewhat performance-competitive, but there is a clear shift in thinking that appears to be moving away from Intel being the focus to the building threat from ARM chip makers, including Nvidia. .... "

And between the May 2012 and May 2013 a constant decline in the legacy PC market. Intel clearly sees ARM as a threat. Why shouldn't AMD?

In constrast could to jump to 2013 for the future predictions.... for example

http://www.amd.com/us/press-releases/Pages/amd-unveils-2013june18.aspx

Shifting from competing on GigaFlops and core count wars to Flops-per watt and more cost effective solutions is far from abdication.

What is still being released are the leftover design. They'll be completely out in 2015.

Doubtful given still in the server game. Operton single processor package can be made workstation as well as a single package Xeon.

People don't buy computer or hardware for what they could be in the future. They buy them for what they can do today!

chuckle, that's why there are gazillion comments in this forums archives about how the Mac Pro is better than iMac/Mini/etc because it has expandibility and is "future proof".

Sure there are buy now elements the decision but at one point you were huffing and puffing about what AMD is going to do in the future. Perhaps what will enabled in the future has to do with where AMD is investing their R&D resources into now.

And as of today OpenCL is lagging behind Cuda.

Quite similar context when Jobs penned Apple's notes on Flash. Similarly the markets move off of old legacy PC form factors neutralizes that legacy inertia effect. If align two industry inflection points they tend to generate lots of synergy.

 

[goMac]

Quite similar context when Jobs penned Apple's notes on Flash. Similarly the markets move off of old legacy PC form factors neutralizes that legacy inertia effect. If align two industry inflection points they tend to generate lots of synergy.

There's also the small problem of how Apple was the group that invented OpenCL. And yet people wonder why Apple isn't backing CUDA.

 

[tuxon86]

AMDs not getting out of the CPU business. What on earth would make you think that? They just inked deals to supply the CPUs for the Xbox One and PS4.

Those aren't DESKTOP cpu...

----------

Care to pull out the press release where AMD says they are dumping GPUs?

Never said they were dropping out. But frankly they are presuing more than one avenue to the future. Futhermore they are going out of their way to point out that they are not a one-trick-pony to the investors and analysts. Being non-diversified is a problem they have been tagged with, that why it is being addressed and highlighted in their quarterly results.





Chuckle. Yeah what share of the andriod market does Nvidia have? Something similar to AMD's of the PC market. CUDA in ARM SoC solutions a dominate factor? You are smoking something.








Oh you mean the part of the article where it says

"... It would be a bit premature to assume that AMD is not developing chips that are at least somewhat performance-competitive, but there is a clear shift in thinking that appears to be moving away from Intel being the focus to the building threat from ARM chip makers, including Nvidia. .... "

And between the May 2012 and May 2013 a constant decline in the legacy PC market. Intel clearly sees ARM as a threat. Why shouldn't AMD?

In constrast could to jump to 2013 for the future predictions.... for example

http://www.amd.com/us/press-releases/Pages/amd-unveils-2013june18.aspx

Shifting from competing on GigaFlops and core count wars to Flops-per watt and more cost effective solutions is far from abdication.






Doubtful given still in the server game. Operton single processor package can be made workstation as well as a single package Xeon.






chuckle, that's why there are gazillion comments in this forums archives about how the Mac Pro is better than iMac/Mini/etc because it has expandibility and is "future proof".

Sure there are buy now elements the decision but at one point you were huffing and puffing about what AMD is going to do in the future. Perhaps what will enabled in the future has to do with where AMD is investing their R&D resources into now.




Quite similar context when Jobs penned Apple's notes on Flash. Similarly the markets move off of old legacy PC form factors neutralizes that legacy inertia effect. If align two industry inflection points they tend to generate lots of synergy.

What!.?!

I said AMD is getting out of the desktop CPU not the GPU... You really are trying too hard now.

As for the rest of your diatribe... Who care really. It looks as if your trying to convince yourself that you really know what you're talking about...

 

[MacVidCards]

There is no "it doesn't matter who is elected" option in a President election.

Far closer analogy to the actual situation is that overriding a veto takes far more than just half the votes.

And frankly these landslide victories of 2-3% percentage points .... well that go covered a long while back in 1981.

http://www.youtube.com/watch?v=xSOp507HJMA

Wow, do you even read this stuff you are posting or do you just Copy & Paste it right out of the emails from Cupertino?

I am including a screen grab of the Poll as it exists, that you call "evenly split" along with this completely 100% nonsense response.

You apologists don't help your case by posting gibberish.

"Evenly split"...I just fell out of my chair laughing.

 

[deconstruct60]

There's also the small problem of how Apple was the group that invented OpenCL. And yet people wonder why Apple isn't backing CUDA.

They aren't as openly hostile to it as with Flash. I doubt Apple kneecapped Nvidia on the Mac Pro for that purpose. Flash was basically an outcast. Nivida is in the rest of the Mac line up , in part because Nvidia reversed course on performance/Watt and price and also support OpenCL (at least some places ). It could be a problem, but not a major hurdle on OS X.

Long term, yes. The OS X system libraries are only going to use OpenCL. Apple is telling no developers to put effort into CUDA. CUDA is likely never going to show up in iOS whether developers want it or not. Apple is going to have zero interest in solutions that can't possibly cross those two platforms. They are not pressing iOS hard to expose OpenCL to general developers yet.

That appears to be part of Nvidia's excuse.

Imagination Tech/PowerVR .... check (2 years ago. I wonder who a major user their GPU is? Hmmmmm ) http://withimagination.imgtec.com/index.php/news/powervr-sgx-cores-get-opencl-conformance

Adreno ... check 320 has OpenCL support https://developer.qualcomm.com/discover/chipsets-and-modems/adreno-gpu

Mali .... check http://malideveloper.arm.com/develop-for-mali/sdks/mali-opencl-sdk/

Nvidia .... apparently it doesn't matter on any of them.

"... "Today's mobile apps do not take advantage of OCL (OpenCL), CUDA or Advanced OGL (OpenGL), nor are these APIs exposed in any OS. Tegra 4's GPU is very powerful and dedicates its resources toward improving real end user experiences. ... "
http://www.brightsideofnews.com/new...cs-disappoint2c-nvidia-in-defensive-mode.aspx

well geee, if it is not there uniformly it is pretty obvious the OS aren't going to expose it. Wonder if someone cares.... [ same article]

"... To put this answer in perspective, Nvidia - a company almost always known for innovation in the desktop and mobile computing space - does not consider that API's such as OpenCL and its own CUDA are important for ultra-efficient computing. This attitude already resulted in a substantial design win turn sour, as the company was thrown out of BMW Group, a year and a few quarters after it triumphantly pushed Intel out of BMW's structure.
... "

Vivantae .... check OpenCL and Android's Renderscript http://www.vivantecorp.com/index.php/en/technology/gpgpu.html