I've said it before and I'll say it again.. The biggest problem Apple has is graphics drivers .
Nvidia even came out with Mac drivers for the 1080's.
Apple needs to go with Nvidia.
Apple has been going with Nvidia , AMD , Intel , and ,up until relatively recently, Imagination Tech***. The Mac graphics stack has multiple layers to it. The lowest level has been done by the GPUs vendors. Apple has controlled the upper levels of the OpenGL stack (and up through their libraries ). Even Metal has some split. Where subcontractors sat might be varied over vendor, but it is really a vendor assignment.
What Nvidia is doing is releasing drivers asynchronously from macOS distribution. That is a dual edge sword because the APIs to do that aren't pragmatically there and that's why there is an explicit dance and incantations that must be done on each OS upgrades.
AMD, Intel, and Imagination Tech aren't running rogue over macOS releases and Apple's schedule because when you do it doesn't help win new design bake-offs. Nvidia has users hooked on CUDA like crack so it appears they are willing to blow up those design wins to keep dropping crack rocks. When the Mac Pro 5,1 is put into obsolete status in a year or so (and OS updates end) that approach may have blow back.
*** Since technically it is Apple's shader implementation asking ImagTech to do drivers for something they didn't do is task misalignment. Typically, low level implementers do what they have deeper ties with.
[doublepost=1492294491][/doublepost]
With all this talk about what the MP7,1 will be, why can't it just be two things.
Same fundamental reason why there were not two 2008 , 2009 , 2010, or 2012 models. The 2009-2012 models used a daughter card to differentiate between one and two socket systems. That meant that sales volume for the core model was shared across those two.
If split there is little in what Apple talked about for that to be viable for Apple. Apple cut off the 2 processor because it is extremely likely that was the much smaller "half" between the two submarkets. Folks yelping about how they and 100 their buddies all have 2P systems ..... who likely have more detailed comprehensive data about how many of each were sold... you or Apple? There is not rational reason at all for Apple to cut off the larger half. As much as many folks moan and groan about costs the lower half was likely larger.
It isn't like the 2008-2009 models were monster volume sellers either.
There is notion that once split both will "halves" will grow faster. There is alot to what Apple talked about their marketing data in that round table session that says otherwise. Trend-lines with folks moving "down market" to iMacs is one. Sore spot being
single mega GPU (and implicitly workloads going GPU) is another.
There are folks whose workloads are heading toward 20-40 x86 CPUs and/or 4 GPUs but much of that is heading toward computer farms rather than sitting on top of individual desks in close proximity to users.
A small form factor like the MP6,1 but with a single GPU, and a 6 to 10 core CPU selection, plus slots for 2-4 SSD sticks. The second MP7,1, could be a MP5,1 respin with at least 2 TB3 ports, but no more than 4. The big MP7,1 could house 2 full size non-proprietary GPUs, and a wider range of CPU choices, possibly even dual sockets. Thunderbolt support could be provided via a modest mobile GPU on the mainboard to provide the required Thunderbolt DisplayPort signal, and support boot screens if no PCIe GPUs are present.
IMHO, far more likely is a compromise between. Room for two Apple customer GPU, but the Thermal Zone the "Compute GPU" is in would have a standard PCI-e slot ( perhaps shared with that custom slot) and optional card edge out. One CPU. Same baseline design shared by both variants. No mobile GPU. The "video" GPU would be present in each version, but would be a decent to very good desktop option (at least at major product updates).
No proprietarily SLI/Crossfire. No 2-3 way Compute/Inference/Training GPUs.
This seems like the best way to serve the widest possible "pro" user group possible, without creating an overly complicated one size fits all solution,
Apple isn't trying to cover the widest possible group. They said so explicitly in the round table discussion. They are trying to cover the most people with a highly limited set of products. It is a balancing act. They are probably just looking to cover "enough" Pro subgroups to have a viable product. Their objective is not to become top 3 in the comprehensive workstation category. So cover almost all users at higher costs isn't the strategy. They are pockets of pros they know in advance they are going to miss.
Two radically different motherboards with two radically different cases probably does not meet the highly limited set of products criteria at all. A Mac Pro with a default video out mechanism that is completely out of alignment with the rest of the Mac pro systems (which have order of magnitude bigger in deployments) doesn't lower overall ecosystem complexity either. It don't think Apple wants any Mac product that is radically different than the others. That isn't going to help build a cohesive ecosystem long term.
What Apple needs to let go of is funneling all video in every circumstance (even optional add-on. non standard configuration) out through TB also. That is a major component of the 'box' they are put themselves in that Apple can't seem to assign enough good engineering resources to keep up with the market every year. They need an 'out' for when they screw up. I think Apple knows there isn't enough Mac Pro volume and sales to keep a large team permanently assigned to the Mac Pro. They could hope to get to a sequence where there were just occasional gaps, but I doubt they want to commit to doing alot more than they already are. The long term growth projections probably are just not there to support that. ( there is no data showing the workstation market is growing at 2002-2005 era rates . It isn't quickly imploding, but it isn't particularly growing either. )
specifically mentioned high-end cinema, VR content creation, and scientific work as the Pros who need the power a module Mac Pro can provide, verses the iMac Pro they have been working on for other Pros who need more then a MacBook Pro