I believe it's reasonable to expect macOS on x86_64 will be very likely available publicly until the dawn of next decade. On the hardware side..x86_64 Mac will end with 2019 Mac Pro Refresh. After that no more Intel Mac.
What I meant at the end of my previous post was that when Apple re-thinks its strategy in 10 years time, I won't rule out the possibility of resurrection of Mac Pro on x86_64. By then if AMD has been proven up to the job, 2030 Mac Pro to be on ThreadRipper Pro.
After 10 years on a Apple CPU , they aren't going back. Will be lucky in ten years time if x86 has finally dropped the 80-90's vintage baggage from the 32 part of the instruction set. If x86 is still carrying around 20+ year old , failed excursions into instruction set modifications .... Apple isn't going back. Just won't happen. If x86_64 can't cut loose bad design decisions, then that platform will probably get thrown into the 'fail' bucket right away .
The best probable outcome that x86_64 could do now is somewhat managed to stop the bleed in. 4-5 years. Its extremely unlikely to ever recover back to what it was in the 2000-2010 era. If it never recovers that kind of dominance Apple is never coming back ( barring utter collapse of the iPhone business and Apple having huge cash flow problems. ).
You are trying to spin that the macOS on Intel is some lifeline back to the "safe harbor" of x86_64 land. It is very probably not. In about 6 years, it is probably going to be mostly abandoned code base. Decent chance that Apple "internally outsources" it off to some. "lower wages country of the next decade" so that folks just keep it running after they make major final tweaks after all the non T2 Macs are retired off the end of the Vintage/Obsolete list. (and no, the "edu" iMac 2017 probably isn't going to extend the non-T2 termination deadline much at all. )
The dual edge sword with "free" upgrades is that when the money runs out then the work probably is going to stop. Once Apple stop selling Intel systems there is no "new" money coming in to cover Intel development. Yeah, technically Apple is suppose to "set aside" some money from each Mac sale to cover its lifecycle worth of upgrades , but there is usually some 'slop' in that schedule makes the money run out quicker ( e.g., inflation, unexpected security bugs that a costly , etc. )
Same thing with the Intel Hackintosh crowd. ... as long as it is a relatively small camp and the major Intel money is flowing in then it is a "cost of doing business" annoyance. When there is no Intel money coming in it is likely going to creep up into being an annoyance. Doing Intel macOS upgrades for folks who never paid a drop for them. You think Apple is going to pour tons of money into that project? Probably not.
If Apple isn't happy with Arm's direction in 10 years they are more likely just to "fork" off of Arm and go their own way (especially if still have healthy iPhone/iPad/Mac/AR-VR businesses resting on their homegrown cores. )
Apple could be fabricating a part of their SoC portfolio at Intel in 10 years ... but going back to Intel SoCs ... that basically toast and wandering off into delusional land.
x86_64 dominates PC market and will be for the foreseeable future.
Apple didn't drop x86 from the Mac because it was dominate on Windows. Apple primarily dropped x86_64 because Apple already had a viable SoC creation business based on iPhones and iPads. Windows doesn't matter. That is one reason why the M-series doesn't support natively booting into Windows because it is absolutely
not a primary driver here in the slightest.
As for x86_64 dominance into the foreseeable future .... They are big cracks there.
Samsung’s chip announcement is light on details and lacks performance claims.
arstechnica.com
Modify that SoC with four X2 cores , crank up the memory bandwidth , and scale up the GPU a bit with a bigger cache the dominance on laptops would be smaller. If Qualcomm also delivers then would have two and dominance shrinking even more.
Windows 11 dumping 32-bit boot and kernel is an inflection point that can allow x86_64 to be significantly displaced if it keeps holding onto to obsolete baggage. Microsoft is not trying to solely cling to x86_64. They to a evolving slowly but there is a substantive inflection point now.
We may not see macOS x86_64 public releases someday in this decade, but internally I believe Apple will continue to have builds for x86_64 hardwares.
A. "Plan B" build of just core OS in case wanted to bolt? Pretty likely that would be RISC-V rather than x86_64 as Apple ramps down x86_64 work. More likely than that would be a some other Arm implementer ( e.g., someone more focused on server/workstation SoCs so Apple wouldn't have to bother with doing that. ) . Arm9 gets more overlap with what Apple's implemented. If someone keeps following somewhat closely behind doing what Apple doesn't particularly want to do then that is a more productive option.
The notion that the Arm ecosystem is going to implode over the next 10-15 is on very , very thin ice.
To develop workstation/server class Apple processors just for Mac Pro can't justify the cost IMO.
For a mid-high range workstation class, they don't have to. A top tier "iMac Pro" could probably use a "M1 Max Duo" (or whatever naming gyration that Apple comes up with) if they don't radically thin out the iMac chassis. ( the current iMac Pro chassis would work just fine up the 300W range. A full blown out Quad might not fit. but a large subset could. ) Where the Mac Pro starts at $6K and goes up, there is a unit volume problem. Bring that back to $3K and there wouldn't be as much of a problem. ( not that they'd slide that far back but there is a fuzzy zone in the middle).
The bigger problem is Apple trying to drag the larger iMac and the "Mac Cube" into the thinnest of thin land with the iMac 24" and a chopped down Mini.
For a server chip they could pursue the "forked" Arm macOS strategy above and just use someone else. Mac Pro 2019 has a "T2". They would just need something similar ( a Apple security and magic sauce. chiplet/tile) that could be placed on same package as a the die from someone else. And open the door to 3rd party GPUs. (or slap some 'glue' on an Apple GPU and tile/chiplet it up like the mac securing module. ). That probably won't work for more customized apps which lean on AMX/NPU/ProRes decode very hard, but some kind of corner case where just need virtualize , sub 64 core macOS instances on some consolidated "big iron"...... if needed that 10 years down the road it could work. (huge compile/build farms ). Actual server chips Apple doesn't need. macOS isn't licensed well that way and macOS has limited (for big iron server context) core count limits.
Going down to. 2-4 slots with a modified M1 Max baseline is quite doable. Swap out the Thunderbolt controllers on two tiles/chiplets for PCI-e v4 lanes and have enough to at at least two x16 slots. Doing that tweak to the die isn't going to cost a ton of money. Need 'mirror layout' dies anyway to do a dense pack 4 tile/chiplet configuration anyway.
four PCI-e slots is what the old MP 2009-2012 had and it wasn't the end of the world. Even more so if none of those are occupied by a GPU.
What Apple can't do is get into a "core count" pissing contest with AMD/Intel/Ampere/etc in the server business.
Apple has zero unit volume there now anyway. M-series isn't going to be vehicle to go into markets the Mac has ignored for over a decade.
Let's hope there will be Apple cars or other secret projects that will re-use Mac Pro processors. When such re-use is not happening, on the long run, it's questionable to me if Apple will continue the effort.
Let's hope not. Car would need something far more specific. Far more fixed function logic for camera and Lidar data coalescing and far, far, far bigger NPU than ever going to get on a Mac Pro SoC. For the former there is probably more overlap with the VR/AR headset SoCs than the M-series. For the latter, Apple is probably behind the curve competitive wise (and bigger NPU is going to squeeze out GPU. M1 Max is kind of a GPU with other stuff sprinkled around it. That is not what need for a car. A car is not a video special effects production house. dealing with reality not faking reality. The Mac Pro is exactly backwards in focus. ]
Personally I also don't think Apple will abandon the Mac Pro market nor they will limit Mac Pro to the even more niche 2013 Trashcan market. That gives the possibility of the return of Mac Pro on x86_64 around 2030.
About as possible as a meteor striking a bullseye on the middle of the Apple campus circle. About as likely as PowerPC coming back.
The release of 2019 Mac Pro Refresh with W3300 Ice Lake Xeon is imminent.
The longer this takes the less likely that seems. If get to June-August time frame the Mac Pro refresh very well could be far more like the "Mac Pro 2012": just slap some new components in the relatively old chassis as a super duper cheap upgrade. 2H 2022 maybe if Apple bribes AMD enough money they could slap some RDNA3 GPU cards in there , wrestle some extra steep discount out of Intel to remove the "> 1TB RAM tax" on the "M" modules Apple is using (and shift the CPU packages down to lower system price points ). and call it a "model change". Same motherboard. This time swapping GPUs instead of minor bump to CPUs and slapping a new year on the end of the product name.
One of the primary problems with the Mac Pro 2019 is pricing. For what can get competitively in mid 2022 it has even lower "value" than it had in late 2019 ( still priced too high but Intel still had a deeper hold on market and AMD was still viewed in a "they might fall on their face again" way. ) . In 2017 Apple re-priced the Mac Pro 2013 so they could limp along for another 2 years. Refresh could boil down to "repricing".
The Mac Pro 2019 chassis would limp along better into the future with a W-3300 ( providing for user changes in I/O to make later , upgrades to GPU after come down from the crypto-mining craze levels, etc. ). But does Apple really care? Over last decade they have twice pushed the Mac Pro past the nominal "expiration date" (2012 and 2017).
If they put in the work in during 2020-2021 to get it done maybe arriving short term. If this was some kind of "planned for mid-2021-to-mid-2022" project then it more likely got trashed and Apple is just going to write any of that work off and go the super cheap route.
techcrunch.com
