You know what's most interesting?
I had no idea that this happened, but apparently, back in 2020, someone found some code (drivers?) referencing Navi 31 (RDNA 3) in a Big Sur beta, alongside the code/drivers for the RX6000 series (Navi 21).
SOURCE 1
SOURCE 2
SOURCE 3
I've been scouring the net to see if there has been any update on that, but sadly there hasn't. I even found some hackintosh guys that tried getting a 7900xt to work on Monterey, but even with spoofing it didn't work -- Granted, he COULD technically use it, but without HA, which means it didn't work.
So.... It feels like perhaps apple
intended to support RDNA 3, and
then they would start to transition to AS, but something must have happened to accelerate that push? I find it quite odd....
It is not clear there is an intention here. A couple of reality checks.
1. Apple isn't the primary source for low level AMD driver definitions. AMD plays a large role in getting AMD GPU drivers running on macOS. Fragments from AMD's other platform development paths can 'leak' into the macOS driver headers.
2. 'source 2's title ends in "..expected late 2021". There was no Navi31 launch in late 2021; on any platform (let alone macOS). The article then goes on to mention Navi21 by that wasn't released until August 2021
Apple today began offering new high-end graphics upgrade options for both the tower and rack versions of the Mac Pro desktop computer. This comes on...
www.macrumors.com
( AMD did announce the 6800/6900 series in 2020 along with the arrival of Big Sur (Fall 2020), but no finished Mac drivers showed up for almost a year. Oct 2020
www.anandtech.com
)
The likelihood that Navi31 would follow months after the delayed Navi21 arrived is pretty slim. Even the AMD roadmap chart in that article pragmatically places Navi31 in 2022 ; not 2021.
What likely happened is that references to Navi21 is what Apple might have been working on. And the Navi31 probably is just entangled in a header file. ( so the variable is defined but never used. Doesn't really hurt to have a 'stub' so that no one else defines a variable that will be pragmatically reserved alter. ). The 'header file' (.h) doesn't really define substantive code; just what some global variable and function definitions look like. The 'do the work' code is in a (.c/.cc/etc ) file is far more likely different between macOS and other platforms than the header file.
In 2020, Apple had not even launched a Navi21 yet let alone a Navi31. The only thing that fit a "Expected by late 2021" was Navi21 not Navi31. There may have been a "Plan B" or "Plan C" intention in putting Navi31 that wouldn't ship for a couple of years in there
3. Back in 2020 AMDs drivers were a bit of entangled hot mess. There have been mentions of 'Van Gough' in the macOS AMD drivers. ( eventually appeared in the Steam Deck console and never used by Apple).
When RDNA3 ( Navi31 ) did launch in 2022, it has a separate driver bundle than the RDNA2 one on mainstream Windows driver path.
Bloated drivers can't be good for performance
www.tomshardware.com
And only relatively recently unified them.
Reunited, with tons of performance uplifts and new features
www.tomshardware.com
If that driver stack was had high decoupling, better cohesion, and functional decomposition that wouldn't have been necessary to deal with the architectural changes (and the associated new bugs that surfaced ) that RNDA3 brought.
If Navi31 had been a modest evolutionary change in the driver model that would have been more tractable for Apple and AMD to do on a tight budget. The big potholes that AMD hit with even the Windows driver suggest that it would not be cheap or an easy path.
However, I know that's a pipedream, especially since there is no Radeon Pro W7900X MPX module (unless if this is what they are planning on releasing within the 8,1?)
This is all a bit circular. The Intel 8,1 exists so therefore the Navi31 exist. There is no good sigh the Intel 8,1 exists. If going by "hints from random R&D strings in the code" the next Intel Mac Pro would have been "Ice Lake" ( W-3300 ). That is a relatively proven dead end. For single user workstations that was largely a bust. ( runs hotter and no single thread uplift). Navi31 runs hotter (consumes more power) also (not particularly the direction Apple is going).
Another interesting read from back in 2020:
... In fact, we have some new Intel based Macs in the pipeline that we're really excited about.
SOURCE
^the above can be found through many other sources as well. But there you have it folks -- straight from the horse's mouth.
This too is very likely a pipe dream. He said Intel Macs. Apple has probably shipped over several 100K iMac 5K 2020 units after Cook made that statement. Those Macs are just as plural as Cook's "Macs".
If Intel had not screwed up the W-3300. ( if it has performed more like the W-2400 does, but shipped in early 2021 ), then maybe would have gotten another Intel Mac Pro. (and perhaps another intel iMac Pro). But it did not. It appears the Apple effort for a "M1 generation" Mac Pro SoC also was a misfire , but Intel's screw up was effectively worse. ( won't be hard looking at them screwing up in 2020 and not make a sound inference that it would not be until 2023, or later, until they got back on track. Which is simply just too little , too late.)
In 2023 ( i.e.,after the "about two years" deadline), the odds are very slim that anything that Cook was referring too in that "Intel Macs" statement would highly conflict with what had just said about the product line transition.
Oh yeah, and then shortly after Tim Cook announces his resignation 😈 and we get another CEO that is similar to Steve Jobs.... Or at least someone who can actually innovate, and not let all the cash reserves go to waste.
The 'Saint Steve' revisionist history is kind of delusional. Jobs green-lit
all the major Apple Silicon foundational work. ( Intel pitched using x86 for the iPad and Jobs pragmatically laughed in their face and doubled down on Apple designed stuff). Do you really think Jobs would have sat around happy and content as Intel fab abilities drove off into the swamp? Really good chance Jobs would have just spent more money to leave Intel in 2019 instead of waiting until 2020. Jobs would be more likely to the more risky move to dump Intel faster than Cook did. And when new Macs came out the old version "Steve'd" rapidly. The 2012 MPB 13" lingering for 2-3 years afterwards probably wouln't have happened with Jobs in charge.
[ The developer transition kit in 2020 was an A12Z which is really an A12X die with a 'spare' GPU core turned on. The A12X die dates from 2018. Apple could have made that same kit a year earlier if the software was ready. The M1 on TSMC N7P or N6 would have not run as well as N5 but it would have been viable ( especially on the laptops). ]
With Cook, it is more likely to get a 4-6 year transition off of macOS on Intel, because he generally doesn't kill 'old' Mac products as quickly as Jobs generally did. Jobs would have told end users to 'suck it up and move to the future'.
There was a large gulf between 2010 and 2013 Mac Pro because Jobs didn't start anything in 2009-2010 to fill anything in 2012. In 2012 the Mac Pro was withdrawn from sale in EU because it was just that out of compliance with a 4 year previous passed regulation. That law passed on Jobs' watch; not Cook's.
Jobs didn't think the x86 and Intel products were "insanely great". It was purely a matter of convenience; cheaper costs through shared R&D while that platform 'happened to work better than others". Once Apple had their own extreme broad base to distribute shared costs over and Intel wasn't delivering on schedule, Jobs would have likely dumped them like an old dish rag.
P.S. Same with Nvidia. The BS that they threw at Apple... Jobs would have cut them off even quicker. Jobs didn't tolerate BS well.