MP 7,1 What does Apple's switch to ARM based processors mean for the future of the Mac Pro

[deconstruct60]

Here is something I was just thinking of that I haven't seen mentioned in this entire thread. The entire purpose of the Mac Pro was that it can be customized to the users needs and desires,

The External ones should "just work" in almost all cases. an external USB device doesn't "have to" inject" any drivers to run. That is what the standard product is for. Same with DisplayPort. if implementing the protocol it should work. All the stuff with external edge ports should "just work". ( there are some funky USB drives that try inject drivers on first insertion to enable non USB features. Those are going to die ... and die hard. Loading drivers on the 'fly" just isn't going to work on Apple Silicon Macs across the whole range very well. And over time ( at least when get to stopping the Intel port, if not sooner. ) not at all. )


The current Mac Pro has a couple of internal connectors. USB... see above not a problem. SATA. Again should be a problem for a pure SATA drive not deviating from the protoocls. PCI-e cards. Here may be getting into a bit of a slippery slope as cards sometimes have UEFI boot drivers so can do work before get to a full operating system. Technically, most of those are platform specific. ( either x86/ARM64/Power/etc specific. ) Pretty good chance that that will be a problem not specific to just Macs. ARM based laptops, servers, workstations from other vendors will stumble upon that isue also if just have x86 drivers built into card's firmware.

UEFI 2.5 has a bytecode (EBC) driver extension put in but probably not many folks following that.

UEFI - Wikipedia

en.wikipedia.org

It pretty much was an attempt to go back and cover what OpenFirmware had, but of course with a different implementation. Byte codes are a bit slower, but in the platform neutral boot manager phase, where just running driver through an interpreter because not trying to do whiz bang applications. Just trying to set some parameters and pick a disk image to proceed with.

and that there will be an ecosystem of accessories for it.
...
I don't see that happening and I think the transition to ARM will ensure it doesn't happen for the Intel Mac Pro.

USB , Thunderbolt , DisplayPort, HDMI , etc ... not really going to be a huge problem. ( Thunderbolt is on a slippery slope but if it is still bascially univeral across the whole Mac line up then there will be a 20+ Million per year market for those accessories just from Macs. Even if Windows on ARM 'fails' and only throws another 1-2 M on top that is still going to big enough for a viable ecosystem. )

There is a potential problem in the discrete GPU space though. If Apple is out to remove them from a larger fraction of the Mac line up. So for example, MBP 16" and iMac 21-24" and the entry iMac 27" all loose them. That many also diminish some of the external PCI-e enclosure ( and eGPU) market. if it is just Mac Pro's that don't use MPX modules then the sales could get low enough to where get back to 2009-2017 era where there is little to no movement on "Mac Cards" in large part because market is too small. AMD and Intel fighting over a shrinking pool of discrete GPU opportunities may see one (or both) largely quit chasing Macs. ( Apple could counter that by throwing both some money to stay attentive until most of the transition is over. )

For compute targeted cards though the server Linux market probably does mean will get some either arm64 or EBC. At least more vendors who thought about doing a non x86 firmware. Similar for fast (not fastest but better than 10GbE) networking cards. ARM servers will present a bigger market Mac Pro may be able to bow-wave off of a bit. Add in Card SSDs... same thing.


There is a fair amount of work for the Add-in-card vendors to get done here. Probably better that they have have a two year lead time. Folks won't be able to "just ship the 3-8 year old" card and firmware as an option. But incrementally bigger firmware storage and a EBC option could get an "old, paid for' card going on the new systems when coupled to the macOS drivers that Apple is going to make just about all of the write anyway. ( IOKit is dying long live DriverKit. but that means everyone with a driver has work to do. ). The switch to ARM is going to kill off old , static products just like the 32-bit purge did. Probably bigger though for Mac Pro targeted cards.

If the Osborne effect sharply tanks Mac Pro sales and accessory makers largely don't get a return for doing MP 2019 products that could have a very chilling impact on the prospects for an arm64 based Mac Pro. But that is low sales as much as it is ARM. Thunderbolt on the arm64 other Macs could offset that a bit as there would be other PCI-e slots to fill there. (or PCI-e embedded like digital I/O box such as an video en/decode box for a laptop. )

if Apple and the accessory makers make good enough sales in the 2 year gap then it probably won't be as bad. By two years out reasonable card vendors should have some that x86/arm64 firmware or x86/EBC .
Similarly if the Windows 10 ARM desktop market makes any progress. ( not holding my breath on that one but it would have an impact here of cracking the "x86_64" only firmware mindset on a broader set of products. )

There is zero reason for Apple to drop out of the MPX module making business though. If all the other GPU A-i-C vendors flake, Apple will have to be the primary lead on solutions. I do not think though that Apple is going to try to push both AMD and Intel out of the Mac Pro MPX GPU chip business though. ( if they do a good job on following roadmaps and delivering top tier Metal code. )

P.S. Customizing the CPU package options 2-3 years after buying probably is going down. But the other parts ( RAM , storage ,etc ) are probably still going to be there.

 

[ssgbryan]

You're comparing a 6-W iPhone ship to a 105-W CPU. The latter has 16 cores. Sixteen A13 performance cores would still not draw 105 Watts.

The A13 CPU already matches the best desktop CPU in single-treaded tasks. More performance cores would make a good workstation CPU. Yeah, you need to figure out how to interconnect many cores, but I don't see why Apple could not do it.

[automerge]1593374892[/automerge]

There are HPC servers using ARM CPUs.

I am not the one claiming that an A12 series chip beats Intel Desktop CPUs. The statement was made, and I debunked it. Single threaded performance doesn't really mean that much in 2020.

I get it - True Believers gotta Believe.

Those of us that have been in computing for the past 40 years or so have seen a large number of CPU design failures. Apple won't be any more immune to that possibility than any other CPU design team.

At the end of the day, the True Believers are building a very large house on a very slender reed......

 

[JayKay514]

Interesting posts. I am using my 7,1 for music production and am heavily invested in 100s of plug ins. Moving to ARM, these would all have to be recompiled which is a massive amount of work for the companies involved some of which are only very small set ups. Added to the other issues mentioned above, I don't think this is going to happen in a hurry.

Yes, that's why it'll be at least a 2-year transition as they noted.

I'm also interested in audio performance. I don't know if it's possible to do a proper DAW benchmark comparison right now as the platforms are so different.

I think there's definitely folks with thousands of dollars invested in UAD and ProTools HDX cards that would like to be able to port them into a future A-Pro.

But while we're on the topic, what I would really love is if Apple put the equivalent of a UAD-2 Quad right into the SOC, to accelerate AU plugins. The SHARC chips on the UAD-2s are relatively old now and not even the latest die processes, so if Apple were to add their own equivalent-performance audio DSP cores to the SOC, they could in theory at least equal an Octo card?

For sealed systems like the MBP, iMac and mini, it could be a huge shot of audio performance. If designed properly maybe we could get some of that holy grail stuff like native near-zero latency monitoring with effects, the kind of stuff we spend $$$$ on external processors or cards to do now.

 

[Kpjoslee]

I am not the one claiming that an A12 series chip beats Intel Desktop CPUs. The statement was made, and I debunked it. Single threaded performance doesn't really mean that much in 2020.

I get it - True Believers gotta Believe.

Those of us that have been in computing for the past 40 years or so have seen a large number of CPU design failures. Apple won't be any more immune to that possibility than any other CPU design team.

At the end of the day, the True Believers are building a very large house on a very slender reed......

Do you think entire line of Apple Silicon Apple Macs will be stuck with 2 or 4 big cores like iPhones or iPads?

No one is claiming that A12z itself beats 8-core latest Intel processors, let alone 16-core 3950x. The fact that Apple already has big core that is already good enough to hang with the very best Intel or AMD has to offer shows that Apple Silicon with 8, 12, 16 Cores will have very good potential to offer great performance while consuming less power than best of x86 processors of same number of cores.

Of course, Apple can fail, just like any other company. But it is not like Apple A series is something of a completely unknown quantity, as we already seen its potential. I am honestly excited to see how it will perform when it is fed sufficient amount of power.

 

[AidenShaw]

Do you think entire line of Apple Silicon Apple Macs will be stuck...

Yes.

Think about it - there's no CPU or GPU other than the T3 chip. The T2 hasn't been a very smooth transition.

 

[Kpjoslee]

Yes.

Think about it - there's no CPU or GPU other than the T3 chip. The T2 hasn't been a very smooth transition.

I don't expect Apple's ARM macs will have less cores than Intel models it will be replacing with. We will see.
If Mac Pro is getting updated with Apple Silicon, it will happen within next 2 years. I expect Apple will still use AMD on dGPUs.

 

[goMac]

Do you think entire line of Apple Silicon Apple Macs will be stuck with 2 or 4 big cores like iPhones or iPads?

Yes.

Well, A12Z is already at 8 cores, so...

I think the rumors have low end Mac Apple Silicon at 12 cores.

 

[jccrtv]

I would imagine apple's new processor would eventually apply to their whole mac lineup. Wondering if that means Mac Pro 7,1 will be phased out sooner or later? Perhaps my concerns don't make sense, but I wanted to ask the board their thoughts around the future of the Mac Pro with apple's own ARM processor.
[automerge]1591753616[/automerge]

I dropped $14,000 6 months ago expecting to own it for a long time and take advantage of the upgradeability as time went on. 6 months later the news that the new Silicon and tuned OSX is the future. This sucks for me unless there is a paid upgrade for motherboard and processor. I'll be fine for a few years but then who knows? Ouch.

 

[deconstruct60]

Well, A12Z is already at 8 cores, so...

I think the rumors have low end Mac Apple Silicon at 12 cores.

There are only 4 P ( 'big' ) cores in the A12Z. The other 4 E ( 'small' , energy efficient) are not. Apple going to 12 has a decent chance of just being 8 P cores with yet another round of 4 E ones.

https://www.macrumors.com/2020/04/23/12-core-arm-macs-2021-report/

Shifting those 4 E's to P's will make a difference if can keep them all feed with work. But if the workload is fairly intertwined then those E's will be kicked to just doing background 'scut work' for the OS and deamons. ( that helps. )
But not particularly going to bring it into parity with MP 2019 12 core on a heavyweight workload.

 

[deconstruct60]

I dropped $14,000 6 months ago expecting to own it for a long time and take advantage of the upgradeability as time went on. 6 months later the news that the new Silicon and tuned OSX is the future. This sucks for me unless there is a paid upgrade for motherboard and processor. I'll be fine for a few years but then who knows? Ouch.

About the same thing would have happened if Apple didn't move to ARM. Either an Intel or AMD processor in late 2022 would probably put a big gap on your current system. There would have been no motherboard or processor upgrade from Apple. There are really no new CPUs coming for your $14K outlay ( other than eventually used stuff retired. ) . You could have gotten cheaper , more dated (at time) CPUs in the future, but that socket was a 'dead end' the day it shipped.

If had work that a $14K could bring a return on capital equipment investment then it is still can accomplish that task now as it did months ago. If it was a $14K vanity project. Well, vanity is expensive. Disposable income disposed ... is just disposed. It never was going to appreciate in value just sitting. there.

If Apple is just as late arriving as the last two Mac Pro iterations, then there is at least another two years until the "count down" even starts on your system. That is a few years right there. After that there is probably at least twice as long as that window going forward to get work done on the system .

The PCI-e slots will be just as upgradable as they are now. Newer SSD on add-in-cards will work in the system. If apple drops a Afterburner 2 in two years.... yep probably will also work in system. The monitor is modular . The add-in 3.5/2.5 drive brackets will probably be more affordable. etc.

 

[handheldgames]

Has everyone forgotten PA Semiconductor?

 

[cmaier]

Has everyone forgotten PA Semiconductor?

No. But what about it?

”has everyone forgotten Intrinisty?”

I can play this game all day.

”has everyone forgotten Exponential Technology / MicroUnity / Transmeta / Rise Technology / NexGen...”

 

[deconstruct60]

Yes.

Think about it - there's no CPU or GPU other than the T3 chip.

May not be a T3 depending upon the sequencing here and what priority Apple puts on matching the new boot sequence environment of the new arm64 Macs. The boot security options are a bit different so may need a T3 to match them up.

However, if the new Intel Macs are arriving before the arm64 Macs then probably low desire to match the new systems are they aren't even out yet. For example, new Intel iMac 27" in July-October with T2. That would be different than if there was a new IMac Pro in Feb-March 2021 with a T2. iMac Pro about same time as the first laptop arm64 goes out the door ( Nov-Dec) they a T2 would probably work there also.

[/quote]
The T2 hasn't been a very smooth transition.
[/quote]

it is a smother transition than if they had jumped straight to where going to with the new arm64 boot set up of zero OS instances that aren't signed by Apple.

The incrementally improved from locking out all external boot by default. ( undoing that in part for currently signed macOS. ). They "broke' some stuff but have also had two years to fix it at this point too.

 

[Wilderness-1902]

I do not know why people keep bringing this up. I was able to use my 2010 Mac Pro until fall of 2017 as my primary system. Plus another two years after for encoding videos all day. The ONLY reason I am not using it today is because it cannot produce HEVC content. I have a gaming system I built in 2015 that I still use today. Using a GTX 1080 (non Ti) which was released in 2016 that is still a beast of a computer and can still play the latest games wonderfully. I also use this as another video encoding machine, and it still performs wonderfully in that area too.

I do not know where this 3 years = NO LONGER USABLE timeframe came from. But it is not true. I know professional photographers that still use Power Mac G5. They don't need all the useless stuff (to them) in the new Photoshop versions. They get their work done and get paid for it on an old system. And that system is going on 17 years old now and is STILL usable.

Your system still work, because you can get software updates and security patches. When Apple stops providing OS updates and patches, what will happen? For example,

• You might not use amazon because the root certificates expires.
• The browsers are too old to be supported by web sites
• You might not get software updates either, because developers have dropped the platform
• You might be vulnerable for hackers.

So this kind of computers are not suitable for the primary usage. But you can still run software of old versions.

 

[jeanlain]

I am not the one claiming that an A12 series chip beats Intel Desktop CPUs. The statement was made, and I debunked

Whoever made that statement, that wasn't me.

 

[koyoot]

... and the Vega 11, which may be AMD's best iGPU (AFAIK) averages 19 fps.

Here we have a game that has been ported from DX12, a port that probably cost performance. It was coded for desktop GPUs, not for TBDR mobile GPUs. This mismatch also costs performance. Then it ran under emulation, which costs about 30% CPU performance, and some GPU performance due the Metal validation layer to ensure the absence of visual artifacts.
And even after all that, the A12Z matches or beats the best iGPUs running the game native?
So, either:
The A12Z/Metal combination beats the competition (i.e, mobile GPUs and iGPUs).
The scene they demoed was much less taxing that the benchmark.
The quality settings were lower than medium (definitely, but that can't explain everything).
The WWDC demo did not run on an A12Z, but some other prototype. (We shall know when someone tries to replicate the demo on the dev kit.) EDIT: Federighi confirmed in an interview that the demos were running on the dev kit, hence the A12Z.
Any combination of the above.

Which is exactly the point. It does not appear at all that Rosetta2 is eating up performance, like people expect, or believe it to do.

Which makes the case for those CPU benchmarks appear a bit... worse for Apple ARM design.

 

[cmaier]

Which is exactly the point. It does not appear at all that Rosetta2 is eating up performance, like people expect, or believe it to do.

Which makes the case for those CPU benchmarks appear a bit... worse for Apple ARM design.

It’s a nice result, but it’s only one datapoint, for an unusual kind of app. I’m confident that Rosetta will be fine, but too early to claim vindication based not his small amount of data.

 

[gnomeisland]

There are only 4 P ( 'big' ) cores in the A12Z. The other 4 E ( 'small' , energy efficient) are not. Apple going to 12 has a decent chance of just being 8 P cores with yet another round of 4 E ones.

https://www.macrumors.com/2020/04/23/12-core-arm-macs-2021-report/

Shifting those 4 E's to P's will make a difference if can keep them all feed with work. But if the workload is fairly intertwined then those E's will be kicked to just doing background 'scut work' for the OS and deamons. ( that helps. )
But not particularly going to bring it into parity with MP 2019 12 core on a heavyweight workload.

and why would Apple only put 12 cores (I agree 8P and 4E are likely) in ANY desktop computer? Given current TDP of Apple’s existing silicon, that 12 core part is a 15watt chip. Is the 12 core MP using a 15watt chip?

this is the point us “fanboys“ are trying to make! Even accounting for increased over head for tying all the core together, you’re looking at 40+ P-class cores at the current MP thermal limits. Apple has invested a lot in that design so artificially limit their chips to laptop TDPs? Other companies like Ampere and Fujitsu have shown it is possible. Do you think Apple is LESS capable? In fact each design will increas Apple’s R&D so my guess is the new Pro will have only a couple of processor tiers starting with lots of cores and then jumping to ridiculous numbers of cores—probably with SVE2 extensions that take AVX512 to task.

 

[netkas]

Don't know if this was posted or not, xcode12 for universal apps has boot.h for arm64 mac and it indicates the use of iboot , not efi as boot-up system.

 

[Boil]

I have no doubt that Apple is going to manage to knock it out of the park in regards to a HPC SoC for the iMac Pro / Mac Pro / (Mac Pro Cube?!?); the real question(s) in regards to Apple Silicon is are they going to roll their own dedicated GPUs or will they continue to use third-party solutions (maybe bring Nvidia back in as an option), and will we see a bunch of fancy dedicated bits for faster video editing / rendering & digital audio DSPs and such?

Come on Apple, please give us a new arm64-based Cube that is a modern day version of a SGI O2, a digital content creation powerhouse workstation! Keep the (massive) Mac Pro for serious horsepower workstations, but give us a 'entry level' 3d / video / (etc.) workstation that can be had for a 3k starting price?

 

[ZombiePhysicist]

They actually said they will have completed the transition within two years. They said they have unreleased Intel Macs, but that likely just means the odd laptop will get a CPU refresh. They will, however, support Intel Macs for years to come. Given that the most popular models are the lower end laptops that will benefit most from the ARM transition, however, I can see them being big sellers. Once a large proportion of the user base is on ARM machines, I'm not sure how much effort Apple will put into the legacy (Intel) version of the OS. PPC only made it as far as Leopard before being left to wither.

Yea, and that by implication means for 2 more years they will still be selling intel machines. The clock stops when the last Intel Macs are sold as to how long intel machines get supported. So if the last Intel Macs are still being sold in 2022, and they support for say 4 years after that, you would still be on a current OS in 2026.

 

[leman]

Single threaded performance doesn't really mean that much in 2020.

Single threaded performance matters enough. But what matters even more is efficiency. A12 offers more performance per watt than anything Intel or AMD have on the market. You are looking at the absolute performance, but the point of the story is normalised performance aka. performance per core. Apple claims their architecture scales well. And even if it doesn't — an 8-P-core equivalent of A12 running at slightly faster nominal clock (let's say 3Ghz) it is still going to be a sub 30W design with the total performance that rivals any comparable Intel or AMD CPU. And A12 is not even the most efficient core Apple has out right now.

 

[jccrtv]

About the same thing would have happened if Apple didn't move to ARM. Either an Intel or AMD processor in late 2022 would probably put a big gap on your current system. There would have been no motherboard or processor upgrade from Apple. There are really no new CPUs coming for your $14K outlay ( other than eventually used stuff retired. ) . You could have gotten cheaper , more dated (at time) CPUs in the future, but that socket was a 'dead end' the day it shipped.

If had work that a $14K could bring a return on capital equipment investment then it is still can accomplish that task now as it did months ago. If it was a $14K vanity project. Well, vanity is expensive. Disposable income disposed ... is just disposed. It never was going to appreciate in value just sitting. there.

If Apple is just as late arriving as the last two Mac Pro iterations, then there is at least another two years until the "count down" even starts on your system. That is a few years right there. After that there is probably at least twice as long as that window going forward to get work done on the system .

The PCI-e slots will be just as upgradable as they are now. Newer SSD on add-in-cards will work in the system. If apple drops a Afterburner 2 in two years.... yep probably will also work in system. The monitor is modular . The add-in 3.5/2.5 drive brackets will probably be more affordable. etc.

Yeah that's true. Difference is that if they had stayed with intel (like they did from the time I owned the 2013 MacPro till 2019) I'd get the speed benefits of future OSX improvements and FCPX speed improvements for many years. That now has a shorter end date with the future being APL Silicon machines. I'm a video editor so I went for the best machine we could afford and then filled it with memory. It will return investment while I own it with speed. It just is sad that I went for the mega Hype Dog and Pony show of the New machine for professionals (we were all waiting years for) only to announce a roadmap change 6 months later. I can afford to buy it all over again but I won't. I'll go for almost-as-fast lower end machines from now on. I stand here holding the bag.

 

[Unregistered 4U]

It's also not just about enabling macOS features or letting FCP in particular handle multiple streams of 4K video. For general software, especially cross-platform applications like Adobe CC and Maya, there is no substitute for actual GPU grunt.

That’s the thing, cross-platform applications like Adobe CC and Maya already perform better on non-Apple platforms. So much so that anyone still using macOS is doing so not because they want the best performance, but because they prefer the OS and the Pro Apps (including developers and Xcode). So, it really is just about enabling the macOS features/performance for those users remaining that prefer the OS and Pro Apps.. All Apple HAS to do at the high end is “perform better than the last highest performing Mac”.

CoreML is still massively behind anything Nvidia has to offer. Both in software and hardware.

I say this from personal experience. Trying to port anything slightly advanced in ML to iOS is like pulling teeth.

OH, I’m sure it is. NVIDIA has been successful at promoting and protecting their proprietary CUDA solution. It’s never going to be it easy to do anything with CUDA where NVIDIA hardware isn’t attached.

If you want more advanced things you have to manually implement them yourself in Metal on GPU to get any performance.

This isn’t surprising, though. Just like to get performance using NVIDIA solutions, you use proprietary NVIDIA API’s, to get performance using Apple solutions, you use proprietary Metal.

 

[goMac]

Shifting those 4 E's to P's will make a difference if can keep them all feed with work. But if the workload is fairly intertwined then those E's will be kicked to just doing background 'scut work' for the OS and deamons. ( that helps. )
But not particularly going to bring it into parity with MP 2019 12 core on a heavyweight workload.

The E's are still fairly fast. Ars dug into that somewhere, I'm still looking for the article, but it's like 1.7 ghz vs 2.4 ghz. E cores aren't slow, just clocked lower. They operate just like normal cores and will pick up foreground work no issue. The system just prefers to use them first for lower loads when the rest of the CPU is clear. It doesn't automatically prioritize work to a P core.

If an A series chip is at full load, all cores will run that load. E cores aren't "reserved" for backgrounds tasks or "scut work." That's not how Apple's scheduler works. If the processor is _only_ doing background tasks, then the tasks will be pinned to the E cores. That's the only special feature for the E cores. They're just there to prevent your battery from being drained by a P core when Mail needs to just check for new messages or something.

It's entirely possible that a Mac Pro CPU would not even contain E cores unless Apple was worried about thermals.

Don't know if this was posted or not, xcode12 for universal apps has boot.h for arm64 mac and it indicates the use of iboot , not efi as boot-up system.

Completely possible they ship with iboot on the final ones, EFI wasn't mentioned during the WWDC session. But also worth pointing out that Apple DTKs don't usually ship with the right firmware due to their kit bashed sort of nature. The Intel DTKs didn't come with EFI either.

If the ARM DTKs seem like they've had work put into pre-boot stuff like the boot picker, that might be more of a sign that iBoot is coming to the final Macs. Which would be a bummer.