I think it would be a nice move on Apple's part if the next Mac Pro could use a similar form factor to the 6,1, enabling Mac Pro owners to use their MPX modules. Things like the W6800X Duo aren't easily matched by Apple silicon yet,
The point of the additional MPX connector is primarily not the "horsepower" of the W6800X Duo , but three other functionalities.
1. Apple 'hates' wires. So Power delivery. In some sense this is correlated with the high power consumption of the Duo cards, but it is more the additional wires and non-neatness that is in play here.
2. Provisioning up to four Display Port video data channels back to the on-system Thunderbolt controllers (one top/front TB controllers and one back TB controller )
3. Provisioning two x4 for the on MPX TB controllers if present.
Apple's M-series (Apple Silicon) blows at least two of those away. The TB controllers and GPU are on the same die. So Apple does not need a complex system to route between display output and TB controllers. Cheaper , more effective, and lower power consumption.
Once the TB and GPU are colocated on the same die it becomes more of a Rube Goldberg contraption to route DisplayPort signals all the back to the primary SoC just so they can go back out again to an outer edge. The signals probably would have to be re-timed on the way back and and well as back out the other direction. The vast majority of other GPU card implementations just deal with the display out ports on the edge of the card. Not sure what all that extra complexity is 'buying'. It is not going to be a "GPU less" SoC in the next Mac Pro. It is not going to be a 'weak' iGPU SoC in the next Mac Pro. ( Pretty decent chance it will start off as an 'entry' GPU with something equivalent to an Ultra (duo Max), 64 Apple GPU cores ).
Apple mentioned in their Studio introduction that the W5700 is the most popular MPX GPU module. There are folks who bought Pro Vega II Duos and W6800X Duos, but that wasn't the bulk of the MPX module buyers. The Duo's were not the top primary functionality of MPX. Contributing factor, but not the primary driver.
not to mention other dedicated modules for audio, video etc. I'm assuming they won't,
There are
zero MPX modules for audio , video , etc. The standard PCI-e cards for audio , video , etc cannot provision for or to TB controllers. They don't have additional MPX connectors.
The only non GPU MPX module is the "storage" MPX module
Apple Store Search Results
www.apple.com
Here roughly 2/3 of the MPX connector there is 'dead'. And it is only a small fraction of the power connection supply to run 4 disks and a RAID controllers. ( 75W off the main PCI-e bus and perhaps another < 75W off the MPX power ( essentially could have been done with a 6-pin 75W molex power cable). It is almost never mentioned in these forums so I'd be shocked if it ever sold in high numbers.
There is an extremely high chance that would get higher polling in favor of bringing 2-4 drive sleds back than for "saving" the Promise MPX storage module. Or Apple provisioning space for a "J2i" bracket being optional in a Mac Pro upgrade.
Promise Pegasus J2i two-drive internal storage enclosure delivers 8TB of SATA storage for your Mac Pro. This module can be installed in your Mac Pro and connects to the main logic board with included cables. You also have space to add a second 3.5-inch SATA drive to your system.
www.apple.com
or a "J3i" bracket with one more SATA connector provisioned.
For most two (or more) slots for 4x M.2 SSD carrier card provisioning would work just fine for most. Perhaps some U.2 PCI-e bays if Apple doesn't want to deal with 3.5" drives internally anymore.
but it would certainly help the upgrade path. If not, I guess there are going be some cheap MPX modules knocking around in the future...
Not going to help any upgrade path if there are no 3rd party GPU drivers. The 'hardware' issues are the most problematical issue here. Apple's dogma of "our GPUs are good enough" and "unified memory is insanely great" is a much bigger issue.
The 2019 "MPX" connector was/is a time stopgap filler for a couple of trends that were coming in 2019 but had not arrived.
1. colocation of GPU and TB controllers. Apple Silicon has 'solved' that problem.
2. PCI-e v4 (and v5) going more mainstream. Bleeding off one x4 PCI-e v3 worth of bandwidth from x16 PCI-e v3 is more "painful" than bleeding it off of. x16 PCI-e v4 (or v5). A standard GPU with a USB-C USB 4 socket (with TB option 'active' ) on it may not be so rare in 2023-24 after DisplayPort 2.0 starts to roll out.
Even with just a DisplayPort 2.0 output port ... probably will be able to drive Apple's TB display docking stations from 2022 going forward in some kind of mode.
3. multiple-chip-module GPUs roll out. So the "two GPU worth of horsepower" in one SoC package gets more mainstream. [ e.g. at M3-M4 stage (TSMC N3 N3P ) stage Apple could shrink the Ultra onto one die. Then two "Ultra single die" could be combined to make a "quad power" Extreme. At that point Apple would probably be covering the W6800X duo. They probably wouldn't cover a. 7800 "Duo" , but would be well into the 'good enough' zone. ]
4. Waiting for 2022+ era when SSD's at higher capacities have killed off more of the user demand for 3.5 HDDs for workloads that need a 8-14TB of capacity storage. And where single (or mirrored) archival 20TB drives are enough to make most folks happy for long term internal storage. (i.e. J2i can solve internal archive problem).
P.S. If Apple is shooting for a "Half sized" Mac Pro and are willing to open the door for 3rd party GPGPU compute card drivers , then it could be useful for a decent number of folks would be a MPX/PCI-e expansion box. Use a PCI-e v4 (or v5) x16 connection between boxes and provision out one or two MPX bays ( 4 standard slots ) for folks that had a relatively large collection of cards. W6800X as pure compute cards would probably find some utility for a decent fraction of upgraders who already had them.
Without some substantive software driver commitment there isn't a huge traction there.
