Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
My personal opinion is that Apple cobbled the Mac Pro together in order meet their transition to Apple Silicon deadline.
 
  • Like
Reactions: Harry Haller
Plot twist: Studio and Pro are retired, get replaced with a new mini-tower.
I simply would not mind.

power-mac-g4-hero1526412492837.JPG
 
That really wasn't true for the Intel Mac Pro's for the iteration cadence that Apple was on for over a decade. Intel switched sockets at the end of every tick-tock iteration during the hey-day of Intel Macs. The 'new CPU' also brought a new socket about every two years (around 36 months 2-2.5 years) ... so no you couldn't drop a 'new' CPU because the socket changed.

Not only the socket but the PCI-e and other I/O elements were changing over time also.

A socket is more so that you can by older stuff ( either at 'used' prices or at 'mature' now more discounted prices). It is not to keep up with the leading edge over an extended period of time. If the objective is to buy up 'scrapped' server room cast offs CPUs at cheaper prices , then yeah.... it is a match.


A highly proprietary daughter cards aren't going to necessarily do it either ( The pins on the daughter cards are off to I/O which is moving also). The Mac Pro 2009-2012 got no new daughter cards.

Doing a relatively ginormous SoC is expensive to create ( R&D costs , very very low userbase to apply the amortization rate , etc. ) . Folks expecting the Mac Pro's to iterate at anything like yearly rates a likely pretty misguided. The Mac Pro is missing something that is bigger than an Ultra. Not necessarily a 'quad' but something incrementally more than just a laptop optimized die pretending it is a 'chiplet' (when it is not).

The volume that made the Intel Xeon workstation class CPUs used in the the older Mac Pros work was entirely based upon non Mac sales. Completely!!!!! Decoupled from those the new SoC rate is not going to increase. What have is a completely different sized ecosystem. There is no viable buying loosey, new CPU only off the shelf market.
Apple controls all aspects of the hardware now so the only reason why the I/O elements would change now is if Apple decides to do that. They could have made the SoC upgradable so you don't have to buy a new MacPro body just to get the latest chip. They could have also made the MacPro dual SoC with an ultrafast bridge; and while two separate M#s wouldn't be as fast as a single SoC with 2 M#s, it would be a lot faster than a single M#.

The point of the Mac Pro is to have the fastest and most upgradable desktop Mac. Not being able to upgrade the SoC runs contrary to that concept.
 
I hope Apple has made progress on previously rumoured "Extreme" chip, and release a Mac Pro with a M3/M4 Extreme chip: with 4 Max chips fused together. That Extreme chip will probably, finally justify the tower form factor and be the "real" Mac Pro.

The M-series concept is great for everything up to the Studio but just isn't what Apple need for a 2019 Mac Pro-like workstation. The 2023 Pro serves a particular niche who are OK with the CPU/GPU/RAM capabilities of the Ultra but need non-GPU PCIe expansion for specialist interface cards and/or tons of internal storage. Its got plenty of CPU power, plenty of PCIe lanes if you're not adding GPUs.

The rumoured "Extreme" chip was "just" 4 M? Max dies "joined together". It would still be based around the idea of integrated GPUs sharing on-package LPDDR RAM with the CPU. Twice the RAM of the Ultra and more of the same GPU cores might be pretty powerful, but isn't going to satisfy people who want a 2024 evolution of the higher-end 2019 Mac Pros, with terabytes of RAM and multiple GPUs (...and a lot of the target market really want NVIDIA, not AMD anyhow).

A "real" Mac Pro would need a chip designed to run multiple AMD/NVIDIA PCIe GPUs and regular DDR5 RAM. No reason Apple couldn't do that - but it probably wouldn't sell enough Mac RealPros to justify designing a whole new CPU die and supporting AMD/NVIDIA drivers.

Of course, we're assuming that the M3 Ultra is going to be 2 M3 Max dies "joined together" - perhaps Apple will do something different and have an Ultra die fused to a GPU die, NVIDA Grace-Hopper style.
 
Why is no one talking about how Macrumors is insisting that the M4 chips are launching mid year. Yes the space between the M2 Pro and M3 Pro is 10-months, but are we really going to be seeing two chip generations in the same year?

And if Vpro actually rolls out with M2, would that be THREE chip generations in the same year?

I have to think Vpro will "surprise" with M3... but nothing suggests that so far.
 
The ASMacPro is a slap in the face to Apples Pro users, (not both cheeks, just one). At least if they could use eGPU that would open a lot of avenues for users without have a huge impact on Apples own software?

Honest question - what areas could macOS running on Apple Silicon SoCs benefit from being able to access the latest NVIDIA and AMD video cards?

It won't be Windows games, because almost all of them do not run on Windows ARM (and even if they did, they would only run in a VM on Apple Silicon Macs, which imposes its own performance restrictions).

For AI / Machine Learning, does NVIDIA's technologies run under macOS? And if they do, are they as efficient/effective as Windows or Linux on x86 compared to macOS/Linux on Apple Silicon?

For video editing, based on real-world benchmarks and user reports, it appear that unless you are working in 8K RAW RED or Canon, Apple Silicon seems to be more than fast enough, especially thanks to the hardware decoding/encoding it offers for codecs like ProRes, h.265/AVEC and h.264.
 
Hmm, I wonder when I will upgrade my first gen Mac Studio, which has the M1 Ultra and 64gb ram. The ultra having double the memory bandwidth is something that doesn’t seem to get mentioned very often, over the Max variants.

I think I’ll be good with my machine for a while, but it’s kind of sad to see the single core benchmarks start to get trounced by these newer generations. An older M1 ultra in multicore benchmarks still seems competitive with the newer generations, from what I understand?
 
My personal opinion is that Apple cobbled the Mac Pro together in order meet their transition to Apple Silicon deadline.

They probably did, and the reason they did that was in discussions with their 2019 Mac Pro customers, enough said "I just need PCIe slots for non-GPU PCIe cards and am willing to pay a premium over the 2019 model" that reusing most of the 2019 model made it cost-effective to do so.

And this is likely why the Apple Silicon Mac Pro will see extended refreshes just like the Intel model did. Clearly a significant number of Intel Mac Pro users did not need faster CPUs or PCIe slots on a regular basis, so Apple did not need to update the motherboard and CPUs every time Intel released a new generation. They did need better GPUs, so Apple did release new cards as AMD did.
 
  • Like
Reactions: lkrupp
A "real" Mac Pro would need a chip designed to run multiple AMD/NVIDIA PCIe GPUs and regular DDR5 RAM. No reason Apple couldn't do that - but it probably wouldn't sell enough Mac RealPros to justify designing a whole new CPU die and supporting AMD/NVIDIA drivers.

They could, but it would have worse memory bandwidth, for a start. Their SoC / RAM on package approach affords them unusually good latency and bandwidth.

Low sales numbers aside, I think it's also the software complexity this would add. They would love to move macOS further and further towards an iOS world where the system architecture is unified and simple. "Sometimes, the RAM isn't on the package", "sometimes, graphics boot from an entirely different chip, with third-party drivers", etc. add complexity that Apple doesn't really care for.

For AI / Machine Learning, does NVIDIA's technologies run under macOS? And if they do, are they as efficient/effective as Windows or Linux on x86 compared to macOS/Linux on Apple Silicon?

Isn't that a chicken and egg thing?
 
Last edited:
The Mac Pro is literally the studio with PCIe…why would that have to be explicitly mentioned to know it’s going to be updated?
 
Hmm, I wonder when I will upgrade my first gen Mac Studio, which has the M1 Ultra and 64gb ram. The ultra having double the memory bandwidth is something that doesn’t seem to get mentioned very often, over the Max variants.

I think I’ll be good with my machine for a while, but it’s kind of sad to see the single core benchmarks start to get trounced by these newer generations. An older M1 ultra in multicore benchmarks still seems competitive with the newer generations, from what I understand?

Based on synthetic and real-world benchmarks, it seems that even with the lower bandwidth of the M3 generation over the M1/M2, it doesn't seem to have much of an impact.

I purchased an M2 Mac Studio MAX and while I am sure the M3 model will be amazing, the M2 is already overkill for what I need so I should comfortably get many, many years out of it. And by the time I am ready to move to M8 or whatever, it will be mind-blowing, I am sure. :)
 
Honest question - what areas could macOS running on Apple Silicon SoCs benefit from being able to access the latest NVIDIA and AMD video cards?

It won't be Windows games, because almost all of them do not run on Windows ARM (and even if they did, they would only run in a VM on Apple Silicon Macs, which imposes its own performance restrictions).

For AI / Machine Learning, does NVIDIA's technologies run under macOS? And if they do, are they as efficient/effective as Windows or Linux on x86 compared to macOS/Linux on Apple Silicon?

For video editing, based on real-world benchmarks and user reports, it appear that unless you are working in 8K RAW RED or Canon, Apple Silicon seems to be more than fast enough, especially thanks to the hardware decoding/encoding it offers for codecs like ProRes, h.265/AVEC and h.264.
Choice. That's what we like everywhere we go.
The majority of Macs I see, (yeah maybe anecodtal), are running Windows. I work mainly in data centres, banks and design houses.
I'm not sure about the Nvidia machine learning thing as we don't have the drivers to compare.
Fitting a Mac Pro with multiple 4090s or 7900s, (yes I've used an extreme example), I'm sure would trounce any Silicon model?
 
While all this seems completely logical and I never understand the need for any Silicon Mac Pro that didn’t include support for 3rd party GPU PCI card. The very limited [some what outdated] Apple sales numbers present a paradoxical view [that I still don’t understand]. Specifically the Mac Pro units out sold the Mac Studio and the Mac Mini units combined!?! This was supposed to have been based on sales September 2022 quarter (M1 generation).

View attachment 2333031

I imagine that the old Intel model was fairly popular for bulk purchases to businesses where IT could later upgrade the innards at a fraction of the cost. Sadly we lost that ability with the M2 model. Either businesses will hold onto them for dear life for years to come or Apple will push out a MacOS update that is ARM only and make them obsolete over night.

In either case the cheaper Mac Studio would make more sense as an all-in-one replacement.
 
I still think the use case for the MacPro should be the ability to run multiple SOCs in a cluster mode. This would be super powerful for machine learning or heavy developer virtualization scenarios.
 
I do think that there is still something possible. I'm sure Apple has something else in its stash... With many cores, and / or several processors and current implementation of ram, pci, ssd's etc., they could certainly get the boost they need. Maybe liquid-cooled...

The only question remains; Who needs something like that. 20k$+...
 
Hmm, I wonder when I will upgrade my first gen Mac Studio, which has the M1 Ultra and 64gb ram. The ultra having double the memory bandwidth is something that doesn’t seem to get mentioned very often, over the Max variants.

I think I’ll be good with my machine for a while, but it’s kind of sad to see the single core benchmarks start to get trounced by these newer generations. An older M1 ultra in multicore benchmarks still seems competitive with the newer generations, from what I understand?
So what do benchmarks have to do with anything other than bragging rights?
 
  • Love
Reactions: Chuckeee
I imagine that the old Intel model was fairly popular for bulk purchases to businesses where IT could later upgrade the innards at a fraction of the cost. Sadly we lost that ability with the M2 model. Either businesses will hold onto them for dear life for years to come or Apple will push out a MacOS update that is ARM only and make them obsolete over night.

In either case the cheaper Mac Studio would make more sense as an all-in-one replacement.
"MacOS update that is ARM only and make them obsolete over night."

It took three years for them to drop support for the PowerPC architecture after they went Intel. It’s already been two years now since Apple Silicon debuted so that could happen any time now.

Mac OS X Snow Leopard (10.6), released in August 2009, was the first version of Mac OS X (later macOS) to require a Mac with an Intel processor, ending operating system support for Power PC Macs three years after the transition was complete.
 
For a company that makes the very best notebooks in the world, it’s funny to see how desktop computers are a laborious affair with Apple Inc.

When you sell around 7-8 laptops for every 2-3 desktops, is it really funny to think Apple would favor spending time and money on the former?


Choice. That's what we like everywhere we go.

But we never had choice with Mac video over the past decade as AMD was the sole supplier of dedicated GPUs for Macs. And they were mostly the sole-supplier before that, as well, except for a period when NVIDIA was the sole-supplier.


The majority of Macs I see, (yeah maybe anecodtal), are running Windows. I work mainly in data centres, banks and design houses.

And I expect every one of those Macs have an Intel CPU in them and either an Intel integrated GPU or an Intel integrated GPU and AMD dedicated GPU.

So again, zero choice, just like with Apple Silicon Macs.


Fitting a Mac Pro with multiple 4090s or 7900s, (yes I've used an extreme example), I'm sure would trounce any Silicon model?

Yes, but of those scenarios that benefit from multiple 4090s and 7900s, how many would perform better with that level of GPU power if run on macOS (Intel or Apple Silicon) compared to Windows or Linux?
 
Don't forget that Apple kept selling the old trashcan Mac Pro for almost 6 years without upgrading it.

I expect that they will upgrade the current Mac Pro occasionally or maybe never, but will continue to sell it. They can drop the price over time, or not, depending on how locked-in the businesses are that need it.
 
  • Like
Reactions: Chuckeee
It took three years for them to drop support for the PowerPC architecture after they went Intel. It’s already been two years now since Apple Silicon debuted so that could happen any time now.
This is a valid concern. But I believe Apple's newly introduced Game Porting Toolkit relies on Rosetta 2, and the Toolkit has been featured heavily in Apple's messaging to developers for how to port games to both Mac and iPhone. So I think Rosetta 2 will not be going away in 2024.
 
  • Like
Reactions: H2SO4
Don't forget that Apple kept selling the old trashcan Mac Pro for almost 6 years without upgrading it.

I expect that they will upgrade the current Mac Pro occasionally or maybe never, but will continue to sell it. They can drop the price over time, or not, depending on how locked-in the businesses are that need it.
I rather think - whatever apple comes up with - we will read here again: "please apple, take my money", "I already ordered today, "I can't wait anymore...". The mac pro has always been a prestige object for apple. That's why there will definitely be a new model.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.