Claim - New 12 CPU Core M1 SoC for iMac Pro

[theluggage]

Ok, for the people who kept suggesting that Apple would disable 8 good cores on the M1 Ultra so they can make a 12-CPU version, where you at?

Read the title of the thread: “Claim - New 12 CPU Core M1 SoC for iMac Pro”

…so we were discussing how Apple could come up with a 12 core M1 SoC if that claim was true. Two of the quotes you posted actually acknowledged that it could just be a false rumor.

If and when Apple release a purpose-built 12 core die (a whole new die design just for one performance tier between Max and Ultra, after all the effort put into making the Max scalable) then you get to say “I told you so”.

Nothing has happened so far to cast doubt on the M1 roadmap that has been predicted for the last 6 months: M1, M1 Pro, M1 Max, M1 Max x 2 (Ultra), and maybe Max x 4 for a Mac Pro. 8, 10, 20 and 40 cores. Any other number of cores would most likely come from binning.

 

[crazy dave]

Nothing has happened so far to cast doubt on the M1 roadmap that has been predicted for the last 6 months: M1, M1 Pro, M1 Max, M1 Max x 2 (Ultra), and maybe Max x 4 for a Mac Pro. 8, 10, 20 and 40 cores. Any other number of cores would most likely come from binning.

That one in bold is looking increasingly unlikely.

 

[theluggage]

Well then maybe Apple could not get "Jade4C-Die" to work and had to scrap it.

If Apple's claims that the M1 Ultra outperforms the 28 core Mac Pro (i.e. the $13k version!) and its discrete GPUs check out in real life, then maybe it's just not worth making the quad version for such a small target market: if the Ultra is so powerful, the Quad starts looking like a very expensive way of supporting twice the RAM, which still won't be enough for people who actually need the 1.5TB capacity of the Intel Pro to work with huge datasets (and not just because they can't be bothered to close tabs in Chrome ).

The M-suffix Xeons that Apple needs to support the full RAM capacity of the Pro are already hideously expensive (even compared to the non-M Xeons) - and Intel has a far larger customer base than Apple, who would be making a special SoC just for the highest tiers of the lowest-selling Mac. Maybe there's a better way of supporting the few customers who need that sort of RAM - perhaps some sort of intermediate external cache/virtual memory built from 'cheap' DIMMs that are still an order of magnitude faster than SSD?

I'll be honest, I have no idea anymore what the Mac Pro will look like at this point.

Yeah - trying to turn Apple Silicon into some sort of drop-in replacement for a Xeon-W seems like using a screwdriver to crack a nut. Needs a bit of "think different".

Apart from the RAM issue, a 1/2U rackmount version of the M1 Ultra Studio with a matching rackmount TB4->PCIe cage (for things like specialist A/V cards, not GPUs) would pretty much do the trick. However, I think releasing any new M1 Ultra system just a few months after the Studio would really tick off lots of people who bought the Studio Ultra.

 

[crazy dave]

If Apple's claims that the M1 Ultra outperforms the 28 core Mac Pro (i.e. the $13k version!) and its discrete GPUs check out in real life, then maybe it's just not worth making the quad version for such a small target market: if the Ultra is so powerful, the Quad starts looking like a very expensive way of supporting twice the RAM, which still won't be enough for people who actually need the 1.5TB capacity of the Intel Pro to work with huge datasets (and not just because they can't be bothered to close tabs in Chrome ).

The M-suffix Xeons that Apple needs to support the full RAM capacity of the Pro are already hideously expensive (even compared to the non-M Xeons) - and Intel has a far larger customer base than Apple, who would be making a special SoC just for the highest tiers of the lowest-selling Mac. Maybe there's a better way of supporting the few customers who need that sort of RAM - perhaps some sort of intermediate external cache/virtual memory built from 'cheap' DIMMs that are still an order of magnitude faster than SSD?



Yeah - trying to turn Apple Silicon into some sort of drop-in replacement for a Xeon-W seems like using a screwdriver to crack a nut. Needs a bit of "think different".

Apart from the RAM issue, a 1/2U rackmount version of the M1 Ultra Studio with a matching rackmount TB4->PCIe cage (for things like specialist A/V cards, not GPUs) would pretty much do the trick. However, I think releasing any new M1 Ultra system just a few months after the Studio would really tick off lots of people who bought the Studio Ultra.

Yeah I don't see how to do it either ... at least not gracefully ... but Apple has said there will be a new Mac Pro and there will eventually be one with AS. So I dunno, I'm just really interested to see what that means. We have the long awaited xMac, now we have to see the what the future holds for the workstation.

 

[CWallace]

I still see no reason why Apple cannot design a systemboard with two M1 Ultra SoCs electrically connected, even if they are not physically connected - so they would be in separate packages and therefore would not be a "new chip".

These separate packages could also have access to a pool of off-package memory to allow more than 256GB of system memory (the A5X and A6X had off-package memory).

Yes, this could involve a performance hit since the two M1 Ultra could have more latency on intra-SoC communication and it might impact the ability for the two GPU clusters to communicate at maximum efficiency.

 

[theluggage]

We have the long awaited xMac, now we have to see the what the future holds for the workstation.

I don't know that there's any consensus as to what an xMac was. My image was always a sort of 'official' Hackintosh - an expandable mini-tower Core i PC in a nice case sans anachronisms like PS/2 ports, with a choice of prosumer GPUs, without the Xeon/ECC/Workstation GPU price premium of the Mac Pro towers. Relatively cheap to develop and - 2010 onwards when laptops were king - wouldn't have cannibalised MacBook sales too much.

That probably ceased to make sense with the arrival of Apple Silicon.

The M1 Max Studio might meet some people's model of the xMac (minus any sort of internal expansion).

The M1 Ultra Studio, however, is absolutely the return of the Trashcan as Apple's idea of a workstation except, maybe, with custom Apple Silicon replacing whatever Intel and AMD deigned to produce, this time it will be more successful. Also, when the trashcan launched, the classic Mac Pro had been ignored to death for several years. Today, the 2019 Mac Pro is still as viable as it ever was (...for anybody who needs RAM and PCIe and can't countenance switching to Windows or Linux).

If I were Apple, I'd just make the Pro a rackmount version of the studio with matching storage and PCIe housings.

 

[theluggage]

Yes, this could involve a performance hit since the two M1 Ultra could have more latency on intra-SoC communication and it might impact the ability for the two GPU clusters to communicate at maximum efficiency.

Trouble is, M1 relies quite heavily on that very efficiency to maintain its edge over x86 and discrete GPUs. Especially in desktop applications where power efficiency is less important.

 

[crazy dave]

I don't know that there's any consensus as to what an xMac was. My image was always a sort of 'official' Hackintosh - an expandable mini-tower Core i PC in a nice case sans anachronisms like PS/2 ports, with a choice of prosumer GPUs, without the Xeon/ECC/Workstation GPU price premium of the Mac Pro towers. Relatively cheap to develop and - 2010 onwards when laptops were king - wouldn't have cannibalised MacBook sales too much.

That probably ceased to make sense with the arrival of Apple Silicon.

The M1 Max Studio might meet some people's model of the xMac (minus any sort of internal expansion).

The M1 Ultra Studio, however, is absolutely the return of the Trashcan as Apple's idea of a workstation except, maybe, with custom Apple Silicon replacing whatever Intel and AMD deigned to produce, this time it will be more successful. Also, when the trashcan launched, the classic Mac Pro had been ignored to death for several years. Today, the 2019 Mac Pro is still as viable as it ever was (...for anybody who needs RAM and PCIe and can't countenance switching to Windows or Linux).

If I were Apple, I'd just make the Pro a rackmount version of the studio with matching storage and PCIe housings.

I agree to a certain extent, except I think they really are going to keep the Mac pro around. I just don't know how: as you said, a lot changed with AS, so some parts of the workstation ethos won't make sense with it, but they said it will transition. So I'm just going to keep an open mind as to what that might mean. I've seen a lot of suggestions on these forums, some very creative ones, but none so far have seemed quite right. By the time, Apple released the latest Mac Pro in 2019, hell by the time they got down to really designing it, they had to know the future AS roadmap. They named this one the studio not the pro. I think the overall Mac Pro form factor will stay. I'm just not sure how they're going to marry the workstation to Apple silicon. There's another chip, the 6500, to come. It should be interesting what it's made of. If it's even relevant.

 

[CWallace]

Trouble is, M1 relies quite heavily on that very efficiency to maintain its edge over x86 and discrete GPUs. Especially in desktop applications where power efficiency is less important.

As macOS continues to diverge away from x86, it soon won't matter how good Intel and AMD are because macOS and Mac apps won't run on them. I am sure the latest IBM Power CPUs are probably beasts compared to the G5, but good luck booting Lion on one and running Adobe Creative Suite 5.5.

As long as Apple silicon runs macOS and Mac apps great, that is all that matters.

 

[oz_rkie]

As macOS continues to diverge away from x86, it soon won't matter how good Intel and AMD are because macOS and Mac apps won't run on them. I am sure the latest IBM Power CPUs are probably beasts compared to the G5, but good luck booting Lion on one and running Adobe Creative Suite 5.5.

As long as Apple silicon runs macOS and Mac apps great, that is all that matters.

Maybe for first party applications and other niche platform specific apps. Third party app developers will keep developing for platforms that still have enough of a market share.

 

[CWallace]

Maybe for first party applications and other niche platform specific apps. Third party app developers will keep developing for platforms that still have enough of a market share.

Well Mac market share has only improved as more models move from Intel to Apple silicon, so...

 

[UBS28]

Well Mac market share has only improved as more models move from Intel to Apple silicon, so.

Some developers will never develop for Mac and you need Windows for their software.

 

[oz_rkie]

Well Mac market share has only improved as more models move from Intel to Apple silicon, so.

Yeah, I know. I am not anti apple or anything. I daily drive both macs and windows/linux. Just pointing out that it does not really matter these days in terms of architectural differences of the hardware as to whether software is written for it or not. The landscape for software development has changed dramatically compared to before.

With software development these days, its much easier to target multiple platforms, there are very strong abstractions and high level APIs around that will compile to your desired targets anyway. Very few cases where you are going to dip down to a low enough level where architectural differences come into play. Plus, most general computing is done on the web anyway.

The stronger deciding factor for which platforms software is written for these days is financial viability and if you are first party (read apple because microsoft is much more open with their software these days atleast) then you might think about vendor lock in like xcode, finalcut etc.

 

[CWallace]

Some developers will never develop for Mac and you need Windows for their software.

And if I only use macOS, I should care why?

If one wants to run Windows apps, one would have a Windows PC.

(For the record, I do run Windows apps, which is why I also have a Windows PC in addition to my Mac.)

 

[senttoschool]

Read the title of the thread: “Claim - New 12 CPU Core M1 SoC for iMac Pro”

…so we were discussing how Apple could come up with a 12 core M1 SoC if that claim was true. Two of the quotes you posted actually acknowledged that it could just be a false rumor.

If and when Apple release a purpose-built 12 core die (a whole new die design just for one performance tier between Max and Ultra, after all the effort put into making the Max scalable) then you get to say “I told you so”.

Nothing has happened so far to cast doubt on the M1 roadmap that has been predicted for the last 6 months: M1, M1 Pro, M1 Max, M1 Max x 2 (Ultra), and maybe Max x 4 for a Mac Pro. 8, 10, 20 and 40 cores. Any other number of cores would most likely come from binning.

You haven't followed the thread close enough. It starts on page one. Everyone acknowledged that nothing is official and that everything in this thread is basically speculation. It's just a matter of discussing which speculation made sense and which didn't.

2x M1 Max cut down to 12-cores made no sense. Others insisted that it does make sense and Apple could do something like this. I thought there was no way. I was right.

 

[theluggage]

Everyone acknowledged that nothing is official and that everything in this thread is basically speculation. It's just a matter of discussing which speculation made sense and which didn't.

Yes: you basically just repeated what I already said. You even quoted the bit of my old post where I said exactly that.

2x M1 Max cut down to 12-cores made no sense. Others insisted that it does make sense and Apple could do something like this. I thought there was no way. I was right.

What about?
There were always only 3 possibilities:

(a) the rumour was just false
(b) Apple were building a completely new M1 variant with 12-cores per die just for a couple of intermediate options
(c) Apple were going to sell a 12-core binned version of the M1 ultra (...and only Apple know what M1 yields are like)

Everybody (as you admit above) accepted (a) as a possibility, if not the most likely solution. Everybody - not just you - was right. Whether (b) was ever a better argument than (c) remains unproven. If the 12-core M1 rumour was still on the table I'd still vote for (c) over a custom die. Of course, once M2 rolls around, everything gets a new die so all bets are off.

 

[theluggage]

Well Mac market share has only improved as more models move from Intel to Apple silicon, so...

We've had a boom in personal computer demand during the pandemic and Apple seemed to have been less affected by the chip shortage than their competitors (being good at logistics, having your best-selling computers using your own system-on-a-chip and being immune from the GPU price/supply madness help).

Apple Silicon is at its most impressive in low power applications like the MacBooks, which blow comparable PC laptops out of the water (...and give non-comparable laptops a run for their money).

However, we're talking specifically about higher-end Mac Desktops here - which were always the least popular Macs, and which don't benefit quite so much from Apple Silicon's low power advantages. Apple make impressive claims for the Ultra, but I'm betting those are based on Mac-optimised software that makes good use of the media engine, neural engine etc.

And if I only use macOS, I should care why?

If you just use FCPx, Logic, MS Office, maybe Adobe CS and browser-based Apps on mid-range Macs you may be fine.

Otherwise, you're relying on third-party developers to keep supporting and improving the Mac versions of their high-end Pro software, despite most now being available on PC and the PC market offering far more potential customers.

If Apple relies on established MacOS customers sticking with the pro Macs and doesn't succeed in attracting new users from PC land, then you're looking at a pool of users that is continually shrinking as people get picked up by PCs that offer more bangs-per-buck or even just a hardware solution that is better tailored to their needs. That leads to a vicious circle - fewer customers, less software support, more customers jumping ship. Sooner or later, the pool dries up.

Back in the late 80s/early 90s, Macs could do things that PCs - when Windows was still basically a DOS shell, and not even full 32 bit - simply couldn't, which is how they got established in pro audio/video*. Those days are long over.

(*although, interestingly, the first viable looking personal computer-based non-linear video editing system I saw ~ 1990 was running on an ARM-based Acorn workstation, which was briefly ahead of the game in it's ability to play moving video from disc. We're talking about building an edit decision list from grotty-looking, quarter-screen 'proxies' that had been compressed by a lab using specialist hardware, of course, but it was still better than a Mac or PC could do).

 

[CWallace]

If Apple relies on established MacOS customers sticking with the pro Macs and doesn't succeed in attracting new users from PC land, then you're looking at a pool of users that is continually shrinking as people get picked up by PCs that offer more bangs-per-buck or even just a hardware solution that is better tailored to their needs. That leads to a vicious circle - fewer customers, less software support, more customers jumping ship. Sooner or later, the pool dries up.

By that logic, Apple should never have abandoned Intel or nVidia then. That way they can just continue to coast on their coattails.

Back in the late 80s/early 90s, Macs could do things that PCs - when Windows was still basically a DOS shell, and not even full 32 bit - simply couldn't, which is how they got established in pro audio/video*.

Apple invested probably billions in adapting Apple silicon for macOS and making it as "best in class" as they could so that Macs will once again be able to do things that PCs either can't do or can't do nearly as well.

 

[CWallace]

(*although, interestingly, the first viable looking personal computer-based non-linear video editing system I saw ~ 1990 was running on an ARM-based Acorn workstation, which was briefly ahead of the game in it's ability to play moving video from disc. We're talking about building an edit decision list from grotty-looking, quarter-screen 'proxies' that had been compressed by a lab using specialist hardware, of course, but it was still better than a Mac or PC could do).

For me, it was the 68000-based Amiga 2000 and a Video Toaster.

 

[JMacHack]

Isn’t this the same crap we’ve heard a thousand times? “Oh nobody will want to develop for Mac!”

It’s still here, and I see no signs of platform shrinkage.

 

[JMacHack]

Some developers will never develop for Mac and you need Windows for their software.

The only thing I’ve encountered are games and Adobe Framemaker. And Framemaker gives me ass cancer.

 

[arvinsim]

The only thing I’ve encountered are games and Adobe Framemaker. And Framemaker gives me ass cancer.

Heard there are a lot of CAD software that are Windows only.

 

[satcomer]

Well Mac market share has only improved as more models move from Intel to Apple silicon, so...

To bad many shareware are still releasing Intel only software! That tells me most of these programs were made by bad developers! I mean M1 has been out for over a year now and all the Macs products, except the Mac Pro and Intel apps are still being develop! Heck at least they could make it Universal at least!

To me the fly by night developers of DAW plugins acting non-alive now in the Mac or PC world! that tells me once they get their money they move on to better things! meanwhile their plugins are dying on the vine in Mac and /windows machines sense 11 came out too! fly by night seems to be their MO!

 

[arvinsim]

To me the fly by night developers of DAW plugins acting non-alive now in the Mac or PC world! that tells me once they get their money they move on to better things! meanwhile their plugins are dying on the vine in Mac and /windows machines sense 11 came out too! fly by night seems to be their MO!

Unless you are willing to pay for subscription or fund their development, it's unreasonable to expect developers to support their offerings forever.

 

[satcomer]

Unless you are willing to pay for subscription or fund their development, it's unreasonable to expect developers to support their offerings forever.

Yea but I hear/see many complaints about plugins not working sense the first M1 Mac!