programmer

It is a common refrain that a platform vendor ought to become a "top rated game vendor" in order to turn their platform into a top games platform.  Nintendo, Sega, 3DO, Sony, Microsoft (XBox), and many more have all tried that.  It does not guarantee success, and it is a wildly expensive undertaking which distracts the platform vendor from its core competency -- being a platform vendor.  The best companies for delivering games to a platform are the game companies whose primary focus is delivery great games in order to turn a profit.  Those companies can be motivated to port to a platform by making that platform easy to port to, and by supplying money to subsidize the porting efforts.  Buying game studios to do such ports creates a very strange business situation which is hard for a public company (like Apple) to manage -- they can't drop the support for existing profitable platforms without destroying the business they bought, but if they don't then they're undermining their platform.  This is not a path I would want to see Apple take.

Apple is a very good platform company, if they chose to, they could leverage that to make their platforms much lower barrier-to-entry for developers than they currently do.  They could even enable the existing catalog on PC to run on their platforms without changes to the games.  Suddenly being able to run most of the PC game catalog on the Mac would immediately make the Mac more interesting to average gamers (you won't capture the high end gamers this way because there will always be a performance penalty for doing this), but that's where Apple could start making porting easier and subsidized (for key titles).  Apple needs to focus on removing barriers in the way of gamers and game developers.  Rather than trying to push the game market piecemeal onto their platform, get out of the way and let it move there on its own.

Sometimes the best way forward is to put something into the market, and increment on it.  I've seen design teams lose their way because they got too detached from market realities.  Just grinding away on a new thing in a lab for years and years doesn't ensure you're going to make forward progress.  Often it can be just the opposite, and what you need is for an exec to step in and put a crisp deliverable in front of you.

Serqetry said:

Most people don't care much about expandability, and Apple sure doesn't.  No one needs the equivalent of an M2 Ultra Mac Studio inside a cheesegrater case.  Apple needs to make something very different from current ASi machines if they want anyone to buy it, and that means way more RAM, much more powerful GPU, etc... whether you order it that way or add it later isn't as important.

And most people don’t buy a Mac Pro.  The sales of that line are tiny compared to the high volume products.  If you don’t need the big chassis with its slots and drivebays, buy a Studio.  The number of users that require more than the 192GB memory an Ultra will be capable of providing is minuscule, and even those are likely fine (or even better off) with a flash or RAM based VMM solution.  Terabyte level RAM sizes are almost exclusively used in cloud servers running large numbers of workloads across many virtual machines, as opposed to a single workload that needs that much.  

Workloads that *are* that large can usually be divided into multiple sub tasks running as a distributed process across many hosts (ie “grid computing”)… and it is usually a good idea to do that for various reasons… but how many of those want to run on macOS instead of a server farm using Linux?  

But if you did have a Mac Pro workstation with a PCI-E backplane as it’s communication fabric, then you could pack a lot of M2 Ultra cards into one box and have a “cluster-in-a-box” kind of arrangement where every card brings up to 192GB of RAM, a bunch of CPUs/GPUs/NPUs/MediaProcessors, maintaining a more even balance between the components.  And thanks to ASi’s excellent power/heat characteristics Apple could do that way more effectively than Intel, and could likely compete pretty well against nVidia and its big multi-GPU boxes.

That doesn’t require Apple to design yet another chip, which despite being built out of common components is really challenging.  Building chips which are higher performance out of the same components is very hard because you quickly start running into many bottlenecks.  The M1 Ultra didn’t scale as well as the Max for that reason, and the harder you push to scale the more bottlenecks you hit.  So rather than a huge investment at the chip level, having the Mac Pro do a system level scaling is likely way more cost effective for Apple.

Serqetry said:

That is what I've always thought.  If Apple doesn't design a new ARM chip series for the Mac Pro that exceeds the limitations of the SoC M-series, there's really no reason at all to even make a new Mac Pro.  The dumb rumors about an M2 Ultra Mac Pro make no sense.  I think that is more likely to be misinterpreted info related to an upcoming M2 Mac Studio refresh.

There might be a new Mac Pro coming too, but it has to do a lot better than M2 Ultra... and it probably isn't coming as soon as people believe.

Why?  The Mac Pro is about a tower chassis with expansion capabilities.  There’s no rule that says it has to have higher performance than every other model.  The Studio has good thermals and that’s the primary limitation when it comes to performance.  And while it’s nice to think Apple could crank out a SoC just for the Mac Pro, the economics of doing that don’t make much sense:  they just don’t move enough volume to warrant it.  When using Intel they had the luxury of a buffet of workstation and server class chips, plus AMD and/or nVidia to provide a selection of GPUs.  This is no longer true with ASi, so they need to figure out how to scale high enough without over investing.  Tricky problem.  

r_mari said:

Apple's Mac Pro could run GPUs on PCIe slots.
But someone - and it won't be Apple - has to write drivers to allow the GPUs to work with MacOS.
Who is going to do this work???  AMD?  nVidia?

In regard to RAM.  Apple could add a memory controller to run external RAM like a huge cache.
Obviously the bandwidth is going to be slower than using Apple's on-chip RAM.  It can never be used as working RAM
since it will be too slow.
But at least it is expandable --- perhaps to 2 TB of external RAM.

I’ve been saying this is a possibility for a while.  And they have the software in the OS to do this already — the virtual memory system.  Paging to/from a RAM disk has been done before, and over PCI-E it is quite fast.  The instant on requirement would mean needing to also write the data to flash, but this too isn’t hard to implement.  The M2 Max scales to 96GB RAM, so the Ultra would be at least 192, which is a very large working set.

Another possibility, mentioned in the thread about the compute module, is that the unit of expandability could be the SoC.  Let the user expand cpu, GPU, RAM by adding M-series processors on pci-e cards, and provide grid compute software.

Will they do these things?  Time will tell.  With a SoC as their basic building block though, they need to approach the Mac Pro differently than in the past, or accept that they will be limited by what their SoC can do.  The Mac Pro is primarily about a big case with slots… it doesn’t necessarily have to out perform the Studio.