Now, if we look at a current Mac Pro it could be outfitted with many different GPUS. The original radeon Vega II each have about the rendering perf of single m1 Max. The single of the shelf (now 600 $) 6900xt
was $999 MSRP when released. I think driving off into the swamp when trying to position Apple's primary target for the Mac Pro is a largely a container for two year old video cards
Apple has been very consistent across the last decade ( MP 2013 , iMac Pro , MP 2019) positioing the primary GPU offerings in this general area of their product line up as being "Pro GPU" offerings. Far more aligned with AMD's Pro card line up than chasing the pricing swings of the consumer gaming GPU card market.
Apple's GPU driver stack has been far closer to "Pro driver" objectives ( focus on stable and longer term) than on chasing the latest shiny game and dozens specific tweaks to goose individual games faster (even if that causes some instabilities).
[ e.g., Nvidia bragging that more than one (close to 2 ) driver release per month is a 'good thing'
https://www.tomshardware.com/news/nvidia-subtly-digs-amd-intel-low-volume-driver
in a production shop were configuration stability is important that is 'plus' ? ]
To go from a W5500X to a W6600X is $300.
is about 35% faster than a m1 ultra. And a system specced out with dual 6800 duos renders at similar perf as 8 m1 Max chips.
I list this a "evidence" that Apple just won't release a new Mac Pro that is worse than the current, almost 4 year old system.
"evidence"? The M1 Mini has a lower RAM capacity cap than the Intel Mini ( in part why it is still for sale). The iMac supports less external monitors. The M1 Studio is less RAM than the iMac Pro .
Apple isn't operating under the some rigid rule that every single sub feature category has to go forward for them to deploy a Apple Silicon solution.
Sure they could make something doesn't beat the top of the line version but if we don't get better perf per $ then what is the purpose? Just for specialty audio work? I mean, for video, does anyone really need anything better that the current Ultra? Or even m1 max?
For as much as Apple charges for RAM and VRAM on the MP 2019 options. If they can do 'double duty' on the system ram and chuck the relatively very expensive VRAM then the pref per $ will go up.
Much of this thread is likely lost on myopically chasing after the Nvidia 4090 (which isn't coming to Macs) and > $25K Mac Pro market.
The conclusion might then be: if there is a Apple Silicon Mac Pro, it will be better than a middle of the road 2019 MP at least. That means rendering capabilities at least better than dual 6900xt but with the benefits of silent running and large VRAM.
Probably not a dual 6900xt. Perhaps a dual 6800. But yes, adding more value in the $6-15K range of Mac Pros than in the $25-50K range. The respective units sold is currently likely not the same size ( higher range , lower in number). Apple shifting to selling more systems (grow the lower range more than shrinking the higher one) then why would they not go down that path? A lower volume of SoC packages isn't going to help them get scale on production of those SoCs..
For that to be true, a M2 "Extreme" without extras will just not cut it. It might touch similar scores but it would't be better. So, apple adds RT HW in the m2 pro series, that could save the day and most people would have a good enough system. Maybe it will not beat a dual 6800 duo or be close to PC/nVidia perf but at least useful.
hardware RT would only perhaps 'save' it on heavy RT workloads. Compute? Nope. It is isn't a silver bullet if trying to drag them into the commodity GPU market pricing dynamics.
A MI210 (or future MI310 ) compute focused accelerator would probably have more synergy between AMD and Apple's objectives. There is no $600 mainstream card for either Apple or AMD customers to price anchor on. So if perhaps $500-1K lower than the standard Linux card ( Apple wrangles some discount because going to help relative increase sales 10-20% AMD goes from 60K/yr to 70k/yr . accelerator sales go up 16% for AMD and Apple gets a slice of the action. The cards generate enough revenue to pay for the extra work for both sides. ).
However, not being able to upgrade the GPUs would make the system a lot less interesting as an investment.
Apple has to get a return on investment for the extra work for enabling something that doesn't exist now. $600 GPU cards that people don't directly buy from Apple generates $0.0 worth of money to pay for that extra work.
Would you guys buy an AS MP 2023 with a really fast 48 core (32+16) CPU but with a GPU in the realm of dual 6900xt at max? List price 12000$ with 192 GB RAM?
A current MP 2019 with 24-28 cores , 192 GB RAM , and single W6900 is $20-21K . So a Mac Pro that is $8K cheaper probably would find more buyers.
Even if 'dialed back' that W6900 (it is priced for the max cryptocurrency craze era ) to a single W6800 the price is $18K ( still have a $6K gap) .
Apple's M-series system RAM prices work better when compared again equivalent capacity VRAM prices. If fold the MPX module cost back down into the base system price more effectively then can deliver relatively higher bang for the buck. For some workloads that will be helpful and for others it won't. For the areas that Apple still does have substantive traction it will probably work. (may get some grumbles from the audio crowd that have close to zero GPU workload. )
No accelerators to be had?
There are other kinds of accelerators than just GPU ones. Or display GPU ones.
There are already 50 PCI-e cards that work with macOS on M-series now. The likelihood that the next M-series Mac Pro has access to zero accelerations is about zero.
Apple probably should try to get some more, but it is a grow a larger market target thing; not whether they are doing it all.
Last edited: