If Apple does a 'quad tile/chiplet' SoC with eight ProRes/H.265 de/encode hardware units in it ( the Max has 2 the Ultra 4 ) there is very little chance any PCI-e add in card is going to compete with that. Especially one that is kneecapped with the 7,1 relatively old age PCI-e v3 infrastructure.
First, If Apple adds more codecs to the Afterburner it will only get more expensive. Relatively few folks by the Afterburner card now. If Apple makes it substantively more expensive that is just a pricing death spiral. ( even fewer folks buy so price goes up . Rinse and repeat for next iteration). If anything need to take Afterburner costs down because the same or better is basically free with the M-series ( starting at M2 generation going on. Even basic version is getting some de/encoders. )
Second, no way it is going to compete with number of concurrent decoded streams being displayed. A separate PCIe card has to ship all that uncompresses/decoded video over the PCI-e channel. At some point just run out of bandwidth. Even Afterburner 1.0 is capped. So for 6-12 , 8K multicam edit bays, a modular card isn't going to make a difference because actually introducing a choke point into the system.
What Apple/AMD can do is roll out drivers that actually fully take advantage of the hardware that is there. Metal 3 will probably allow more apps to squeeze more performance out the 6000 series that Apple already sells. Extremely likely that the VideoToolbox is not completely optimized for the VCN 3.0 hardware present in the latest MPX modules either. ( as even Windows drivers were not until Spring time 2022 either. If not on the mainstream OS driver targets than likely not on side niche macOS builds either. )
As far as "ignoring the 7,1" goes for new GPUs. When did they really put any huge substantively effort into the 5,1. From 2010 -> 2012 there was zero GPU updates to the configurations. It was the same old stuff. Apple did sell some 3rd party cards that somewhat popped up on the vacuum between 2010 and 2012. However, by in larger the 5,1 got "hand me downs" card from new Macs that used new GPUs. The MP 2013 spawned hand me downs for the 5,1. The iMac Pro and iMac did the same. The 580X on the MP 2019(7,1) is basically a hand me down also.
Very similar pattern in eGPU support space. Where GPU got used in an embedded fashion in a next generation Mac it got nominal support as Macs got more aligned UEFI boot support. But the coverage didn't really go to stuff that Apple had never selected and was substantively different at the driver ID family and frame buffer level.
The 7,1 was a bit new behavior in that got 6800 and 6900. But was that because a W-6300 (IceLake) stopgap update was a stillbirth and Apple rolled out the MPX modules that were done anyway. Or Apple had some deep love of independent GPU module sales into 'old' Macs regardless of new Macs? Hard to tell. Other last generation Macs embedded GPU had moved to the 6000 generation also. So there may have been no AMD 7000 investment planned at all. And the 7,1 already go the "hand me down".
If Apple had already put aside a budget for doing a Xeon W-6300 system development and killed it then perhaps could take some of the "left over" money and put it into a AMD 7000 update that is paid for mostly on "sunk cost " money. A 7,1 + 7600-7800 modules would be cheaper development path than a much wider update. It is a better 'stop gap' than just sitting on aging hardware. And if the 8,1 is a "one slot wonder" box then probably better to keep a multiple slot box in the line up. ( very similar reason kept the Intel Mini around for two years. The "capacity chop" for some significant submarkets is just too big. Set the 'entry' level higher 16 cores , 48GB , 2 TB , 7700 with fewer folks on fatter margins. Use that to pay down some of the development costs. )
There is a long standing "big lie" that Mac Pro GPU card ecosystem was independently stable , healthy ecosystem It really wasn't. 2006-2015 era was really about very , very few vendors producing cards and them secondary market that preyed on other folks work to hack/plifer boot ROMs and those vendors being under paid to get return on their investment. The market was never large enough were people want to pay for the work to make it grow. Similarly, the dGPUs of the MBP and iMac's was driving most of the inertia that Apple was putting into the embedded portion producing "hand me downs" as the EFI->UEFI shift took place.
There is no UEFI on the "bare metal" M-series for macOS. Apple has good enough GPUs to eliminate them from the rest of the Mac line up ( iMacs and MBPs ). Can quibble about whether has 'high end enough" GPUs going forward , but anything about how Apple cannot cover the lower end -to nominal-affordable-mid-range is basically dead in the water.
The dGPU card sales volume is primarily down to upgrades to the last Intel Mac Pro and eGPU enclosures on Intel Macs. Basically a non growth market.
One version of the big lie was that the Mac Pro was pulling the GPU updates fro the rest of the Mac line up along. Wasn't really true before and definitely not true now in the Apple GPU era.
The other part was that hardware standard slots 'solved' everything in and of themselves. Again nope. No drivers , no pragmatically working solution. Drivers also don't tend to get done for free. If there is not a large enough revenue path back to the dev team then work will probably stop. The lack of a healthy, viable retail GPU card market limits the ecosystem. There is lots of hype that this market is bigger than it really is without Apple priming the pump with embedded dGPU sales.
