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

[nbwallace]

Well, we are in a different situation than before. ARM is rising especially for the server. Ampere made 80 cores ARM chip and others too. But obviously, macOS itself isn't good for the server system so maybe they need improve both CPU and software?
[automerge]1592112668[/automerge]

I think we’re all aware that OS X is a very nice wrapper around a Linux kernel. What difference does it make if a server has some easy to use interfaces. These days there are plenty of them that run on any brand of Linux.

Having all that overhead on a server doesn’t make much sense.

What we are talking about here is not instruction sets, compilation or anything else, those are just abstractions. Today and in the future efficiency is what matters, if you can produce more flops per watt you win. Media streaming and blockchain are using significant resources and contributing significantly to climate change. Apple’s current ARM based chips are very efficient, and like most modular designs can be effectively scaled. They can and will build fast, wide memory access, it’s been done before, it’s not magic, it’s engineering.

There are no show stoppers here, all that’s going on is either easy recompiles (whether it’s JIT or pre=compile is comically irrelevant) there is no debate here at all. You build an abstraction layer somewhere and to the users, and likely the developers, things look very similar if not the same. The cost of the abstraction layer in terms of compute will be smaller than the gains from a smaller TDP.
[automerge]1594044429[/automerge]

This is not a like for like comparison and therefore irrelevant.


I think an Apple ARM CPU has potential. The question is: To what degree. I've read a lot of pro Apple ARM posts on this site and if the performance claims are to be believed then Intel needs to liquidate or find some other pursuit.


Then let's me be perfectly clear: Intel has floundered as of late. That out of the way does their floundering mean an Apple ARM implementation will be significantly better? Or have longevity?

I've been through this before with the PPC. According to the Mac advocates PPC was the best thing to happen to the processor market. It was new, exciting, and it was RISC. But then something happened...it ended up being a dead end which led to Apple's adoption of x86.

Will ARM follow suit? Only time will tell. But the transition from PPC to x86 had one significant benefit which I see as a sep back for x64 to ARM...the ability to natively run x64 software. Which meant the ability to natively run Windows and its huge software library. Will Mac purists (i.e. those only using macOS specific software) care? Unlikely. But I think the move by Apple to use x86 had a huge benefit in being able to natively run Windows and its associate software. Thus it was a low risk because if a user couldn't find a native macOS application they could always fall back to Windows with very little penalty.

[automerge]1594044980[/automerge]

This is not a like for like comparison and therefore irrelevant.

This is the only thing that matters. All the rest is just engineering and a bit of code that seems to be available already. You’re concerns are not even real.

There is really very little difference between some number of A12Xs and whatever you seem to think a ”desktop” processor is, you want to address 1.5 TB of ram, fine widen the bus and add some cooling. It just doesn’t matter. These are really just semantics. If someone took the developer kit mini and stuck it in a tower, I am not sure you’d notice a difference.

“But I run Maya!” It’s just code and faster/wider busses. Untwist your undergarments and figure out a way to think more abstractly. That’s how hard problems get solved. Eventually, and probably sooner than you think you’ll be able to buy a laptop which smokes many “desktops” except for how much ram it can physically hold.

 

[AidenShaw]

Well, if we're being pedantic, Mac is not an acronym, it's an abbreviation.

I threw you a bone

 

[09872738]

I think we’re all aware that OS X is a very nice wrapper around a Linux kernel.

Its not, its a unix (Mach) kernel. Just sayin‘

 

[Stephen.R]

I think we’re all aware that OS X is a very nice wrapper around a Linux kernel.

Ahem. No, it is not.

It's a "very nice wrapper" around XNU, which is a hybrid kernel with heritage and code coming from both Mach and the FreeBSD kernel.
[automerge]1594048825[/automerge]

What difference does it make if a server has some easy to use interfaces. These days there are plenty of them that run on any brand of Linux

Well for one thing if that UI is dependent on a beefy GPU and can't be run without it, that means your server needs a beefy GPU to run a UI nobody is looking at.

Edit: removed word salad.

 

[ZombiePhysicist]

Ahem. No, it is not.

It's a "very nice wrapper" around XNU, which is a hybrid kernel with heritage and code coming using some from both Mach and the FreeBSD kernel.

really!? When the hell did this happen!? What version of OS X?!

 

[leman]

really!? When the hell did this happen!? What version of OS X?!

The very first one. OS X never used the Linux kernel.

 

[Stephen.R]

really!? When the hell did this happen!? What version of OS X?!

Sarcasm doesn't play well in text. You were being sarcastic right?

 

[leman]

Sarcasm doesn't play well in text. You were being sarcastic right?

Oh, they were sarcastic? The joke's on me then. My humor detector automatically turns off when I am on MacRumors. Otherwise I would assume that every second post is sarcastic.

 

[Stephen.R]

Oh, they were sarcastic? The joke's on me then. My humor detector automatically turns off when I am on MacRumors. Otherwise I would assume that every second post is sarcastic.

 

[ZombiePhysicist]

The very first one. OS X never used the Linux kernel.

Not true. The first version of OS X was definitely a mach kernel.
[automerge]1594053266[/automerge]

Sarcasm doesn't play well in text. You were being sarcastic right?

Sorry, not being sarcastic. Something is clearly going over my head. My understanding is that apple has been using a Mach kernel variant from the NeXTstep days to pretty much currently. If they changed at some point, I sadly missed it.

 

[leman]

Not true. The first version of OS X was definitely a mach kernel.
[automerge]1594053266[/automerge]


Sorry, not being sarcastic. Something is clearly going over my head. My understanding is that apple has been using a Mach kernel variant from the NeXTstep days to pretty much currently. If they changed at some point, I sadly missed it.

I'll just leave this link here: https://en.wikipedia.org/wiki/XNU

NeXTSTEP never used Mach kernel directly. It was always a hybrid.

 

[Stephen.R]

Sorry, not being sarcastic. Something is clearly going over my head. My understanding is that apple has been using a Mach kernel variant from the NeXTstep days to pretty much currently. If they changed at some point, I sadly missed it.

... and how does that conflict with what I said:

XNU, which is a hybrid kernel with heritage and code coming from both Mach and the FreeBSD kernel

 

[ZombiePhysicist]

... and how does that conflict with what I said:

I'm not challenging you bro. Chill. I was asking.

Anyway, it's interesting. Still seems that it's basically still a mach core:

With macOS, the designers have attempted to streamline some tasks and thus BSD functions were built into the core with Mach. The result is a heavily modified (hybrid) OSFMK 7.3 kernel, Apple licensed OSFMK 7.3, which is a microkernel,[4] from the OSF. OSFMK 7.3 includes applicable code from the University of Utah Mach 4 kernel and from the many Mach 3.0 variants forked from the original Carnegie Mellon University Mach 3.0 microkernel.

Which explains while everything is still mach_o stuff all over the place. But wasnt aware that the micro/macro mashup warranted a new moniker for the kernel, XNU, nor that it was a mash up of Utah mach 4 kernel (but I did know it was part of the Carnegie mach 3 kernel). Interesting. Thanks for the lesson!

 

[macguru9999]

I am not quite sure who, if anyone, would buy a 7,1 mac pro at this point. It would be interesting to look at sales figures. There may be a scenario that applies to the big end of town, but surely anyone who can wait would do so ?? unless they are unaware of the switch.

 

[ZombiePhysicist]

I'll just leave this link here: https://en.wikipedia.org/wiki/XNU

NeXTSTEP never used Mach kernel directly. It was always a hybrid.

Um, yea, it was always a mach kernel when NeXT started up.

http://vtda.org/docs/computing/NeXT/NeXT%200.9-1.0%20Release%20Description.pdf

After Apple acquired NeXT, the Mach component was upgraded to OSFMK7.3 from OSF,[2] the BSD components were upgraded with code from the FreeBSD project, and the Driver Kit was replaced with a C++ API for writing

During the next days it was always mach plus a BSD 4.3 layer. This XNU nomenclature came later.
[automerge]1594058017[/automerge]

I am not quite sure who, if anyone, would buy a 7,1 mac pro at this point. It would be interesting to look at sales figures. There may be a scenario that applies to the big end of town, but surely anyone who can wait would do so ?? unless they are unaware of the switch.

Odds are the wait is about 2 years for something comparable in power, and likely worse for any intel based virtualization.

 

[Blair Paulsen]

The crux of this issue is time. How long until Apple can put out a legit ARM workstation? How long until apps other than in house programs like FinalCut and Logic are optimized for ARM architecture? Etc.

If you knew, for certain, that a stable and well supported high performance ARM shredder would be on the market two years from today - then you could run the numbers on whether getting a 7,1 today is a good investment for your particular business model. The reality is that in July of 2020, there are way too many variables to accurately forecast when that day will come.

FWIW, it's not like a 7,1 becomes a paperweight overnight. Look how long many people kept getting work done on pimped 5,1 cheese graters. If the delta between how much you can crank out on a MBP vs a 7,1 MP is significant enough to pay for the tower over a couple of years I wouldn't sweat what's next. When the performance delta between the 7,1 Intel MP and an 8,1 ARM MP reaches a certain level then upgrade.

 

[iindigo]

Which explains while everything is still mach_o stuff all over the place. But wasnt aware that the micro/macro mashup warranted a new moniker for the kernel, XNU, nor that it was a mash up of Utah mach 4 kernel (but I did know it was part of the Carnegie mach 3 kernel). Interesting. Thanks for the lesson!

Interestingly, they're pushing XNU back in the microkernel direction lately with all the stuff that's been moving to userspace. The plan is to eliminate kernel extensions entirely, perhaps as early as macOS 11.1 or 11.2 based on the current rate of deprecations.

I am not quite sure who, if anyone, would buy a 7,1 mac pro at this point. It would be interesting to look at sales figures. There may be a scenario that applies to the big end of town, but surely anyone who can wait would do so ?? unless they are unaware of the switch.

Depending on what the pending iMac refresh ends up looking like, I might pull the trigger on a 12/16 core W5700X 7,1. It's a lot more than I'd like to spend, but if the iMac refresh fails me it's the only real Mac that fits the specifications I'm looking for.

Do agree that Mac Pro sales numbers right now can't be huge.

 

[Pressure]

The crux of this issue is time. How long until Apple can put out a legit ARM workstation?

The timeframe given were 2 years from start to finish and that the transition will start at the end of this year.

The Mac Pro may be the last in line like the last time around, where it took 7 months from the release of the Intel based MacBook Pro to the Mac Pro.

 

[Ethosik]

I am not quite sure who, if anyone, would buy a 7,1 mac pro at this point. It would be interesting to look at sales figures. There may be a scenario that applies to the big end of town, but surely anyone who can wait would do so ?? unless they are unaware of the switch.

I am looking at buying one. I need the 8 cores performance and I need a more quiet machine than my i9 2019 iMac provides. This thing gets way too hot and loud way too fast. The expandability over the iMac Pro is also preferred as I can upgrade the RAM, GPU, Storage and more on the Mac Pro over the next few years.
[automerge]1594071937[/automerge]

The crux of this issue is time. How long until Apple can put out a legit ARM workstation? How long until apps other than in house programs like FinalCut and Logic are optimized for ARM architecture? Etc.

If you knew, for certain, that a stable and well supported high performance ARM shredder would be on the market two years from today - then you could run the numbers on whether getting a 7,1 today is a good investment for your particular business model. The reality is that in July of 2020, there are way too many variables to accurately forecast when that day will come.

FWIW, it's not like a 7,1 becomes a paperweight overnight. Look how long many people kept getting work done on pimped 5,1 cheese graters. If the delta between how much you can crank out on a MBP vs a 7,1 MP is significant enough to pay for the tower over a couple of years I wouldn't sweat what's next. When the performance delta between the 7,1 Intel MP and an 8,1 ARM MP reaches a certain level then upgrade.

Yes agreed. People use more than Final Cut Pro X and Logic Pro X. Other than Adobe and Microsoft, we might be looking at a slow transition of Intel software being changed to ARM. We might also lose some popular software because developers won't want to do this.

 

[macsound1]

While there is of course much on the horizon, everything currently runs on Intel macs and if you need a Mac Pro, you're going to buy a Mac Pro.

At my old job, I was in the boat of purchasing all new hardware before the PPC-Intel transition. Dozens of video ingestion and simple audio editing iMacs, a handful of PowerMacs for video editing and audio recording.

It was 2 years before they released the Intel MacPro and iMacs, we replaced about half the first year and the other half the following year, right on schedule with their normal replacement cycle.

A couple PPC iMacs became printing stations or on a cart for quick audio recording around the buildings, but eventually they all went by the wayside, just as they would have if there was no chip transition.

 

[Ethosik]

While there is of course much on the horizon, everything currently runs on Intel macs and if you need a Mac Pro, you're going to buy a Mac Pro.

At my old job, I was in the boat of purchasing all new hardware before the PPC-Intel transition. Dozens of video ingestion and simple audio editing iMacs, a handful of PowerMacs for video editing and audio recording.

It was 2 years before they released the Intel MacPro and iMacs, we replaced about half the first year and the other half the following year, right on schedule with their normal replacement cycle.

A couple PPC iMacs became printing stations or on a cart for quick audio recording around the buildings, but eventually they all went by the wayside, just as they would have if there was no chip transition.

We also went from PPC to Intel which was going towards an industry standard. With a benefit of using Bootcamp. Now things are changing away from a current industry standard. I will be very shocked if all the apps get converted over. I think we will end up losing a few very good apps due to this transition.

 

[ssgbryan]

We also went from PPC to Intel which was going towards an industry standard. With a benefit of using Bootcamp. Now things are changing away from a current industry standard. I will be very shocked if all the apps get converted over. I think we will end up losing a few very good apps due to this transition.

You can count on it - just look at the last transition. The software that did transition took about about 4 years 1st version was Intel native, and that was about it - it wasn't until version 2.0 (Intel) that software houses hit their stride.

A good chunk of that was Apple's fault (P.T. Barnum was pushing 64-bit Carbon, since the Cocoa libraries weren't ready, and then did an about face 12 months later 64-bit Carbon is dead - you can trash everything you converted over the last year) - which was why Adobe & Microsoft were so late to the conversion.)

Then look at how long 10.6.8 hung on because of Rosetta. I didn't upgrade until 10.10 came out, and I was far from the only one.

 

[macguru9999]

I am looking at buying one. I need the 8 cores performance and I need a more quiet machine than my i9 2019 iMac provides. This thing gets way too hot and loud way too fast. The expandability over the iMac Pro is also preferred as I can upgrade the RAM, GPU, Storage and more on the Mac Pro over the next few years.

I suppose there are tasks where the 12 core 5,1 is too slow, the i9 imac too hot and noisy etc.... they may sell more 16inch macbooks in the interim although they must be pretty hot as well when worked hard.

 

[Ethosik]

I suppose there are tasks where the 12 core 5,1 is too slow, the i9 imac too hot and noisy etc.... they may sell more 16inch macbooks in the interim although they must be pretty hot as well when worked hard.

5,1 doesn't export HEVC content and doesn't have the most up to date GPUs without a lot of modifications. Which is why I demoted my 2010 Mac Pro for just rendering non-HEVC footage.

 

[Unregistered 4U]

I think we will end up losing a few very good apps due to this transition

Any application developer that’s gone through the hoops to get their app up and running for Catalina is already almost there. I’d figure the list of very good apps is mostly apps that aren’t currently running on Catalina.