...so there's definitely no
hard deadline about to expire on the 6th, you could argue that the two years will be the 2nd anniversary of the Air/Mini launch in November 2020 and there have been one or two little unforeseeable road bumps since then which give Apple a pretty good excuse for being late.
...and at the Mac Studio launch in March, Apple said that "there was just one more product to go" - Mac Pro, which was "for another day". I think Apple usually choose their words very carefully and while you could twist that into "we're going to announce, another day, the Mac Pro is staying with Intel" I think that's stretching credibility. Note, though, that Apple are still selling Intel Mac Minis so saying that a product has "transitioned" doesn't mean "we've stopped selling the Intel version".
Yeah, I think that's more significant than what Apple may or may not have said about a 2 year transition. Pro users who may currently be thinking about specifying Mac Pro for an upcoming product are currently in limbo, with no clue as to what an Apple Silicon Mac Pro may comprise or how long the Intel version will continue for. Maybe the really top-tier customers have been given more info under NDA, but who knows?
What I still don't quite see is how Apple Silicon is going to
work for something like the current Mac Pro. The M1 Ultra already delivers the raw performance, but the MP is also about supporting massive amounts of RAM and PCIe expansion, including discrete GPUs, afterburners etc. There are other uses for PCIe slots - e.g. specialist A/V input output cards - but they generally don't need all those resource-hungry 16-lane MPX slots. By contrast, many of Apple Silicon's strong points come from having everything, including GPUs and the rough equivalent of Afterburner, on-package if not on-die (and those also play a role in it's low power consumption). So while I don't know of any technical reason why Apple
couldn't build a Xeon-killer chip with the necessary RAM and PCIe capacity, would there really be any big advantage, apart from scoring points against the x86 instruction set? Machines like the M1 Max MBP and the M1 Ultra Studio are what Apple Silicon is really good at.
Any new chip
just for the Mac Pro - probably Apple's smallest-volume Mac - would be awfully expensive, too. And even if some sort of "quad" M1 Max chip did have the necessary RAM capacity, boy is it gonna cost, and producing different packages for every reasonable step from 128GB to 1.5TB would be cumbersome...
Really, this is an opportunity for Apple to do something "courageous" with the new Mac Pro, while keeping the reasonably-up-to-date Intel Mac Pro on the books for a while.
I'm going to stick with two wild, mad guesses for an Apple Silicon Mac Pro:
Option 1: Basically a rackmount M1 Ultra Studio (maybe with a couple of extra TB ports - if the M1 Max can support 4 then the Ultra ought to be able to support 8) - that can be paired with a (possibly 3rd party) TB to PCIe enclosure (which you
wouldn't be expected to use for GPUs).
Option 2: M1 Max or Ultra
on a MPX card - with a new "Mac Pro" that is basically just a MPX backplane
but also compatible with existing Intel MPs. MPX would make sense for this as it adds extra power, thunderbolt and video connections to basic PCIe.
...both of those with a side-order of xGrid 2 to support cluster computing. No, that's not the same as more CPU/GPU cores, RAM or proper multi-processor architecture
but the alternative is making that expensive, low volume custom ARM CPU.
