MP 14,8 The future of Mac Pro

[Xenobius]

Four conditions for the Mac Pro to survive and be a workstation class machine:

1. A heterogeneous shared memory architecture that allows the use of both super fast main unified memory and expansion with DDR5/6 modules (something somewhat similar to what Intel uses in its latest Xeon processors)
2. A large number of real PCIe (in version 5.0 to start) lines for expansion cards instead of the current fake ones
3. PCIe accelerators – expansion cards for Metal/Neural computing
4. Double the number of CPU/GPU cores

 

[avro707]

5. Ability to use whichever non integrated GPUs the buyer wants to use. Be it Nvidia or AMD. Even four of them.

6. Work with Microsoft to get windows back working natively for those who need that.

7. Kill off the Mac Pro and Mac Studio completely, build inexpensive regular consumer products that people throw away each year and tell everyone to buy PC workstations.

5 and 6 might make me look at Apple.

 

[rpmurray]

I don't think the Mac Pro really has a future and that the 2023 will be the last. Realistically they should have called it the Mac Studio Pro.

 

[Mr. Dee]

I see two to three more iterations before Apple cancels it similar to Xserve. The rev is gonna be a redesign maybe in a much smaller chassis and different design for design sake. Apple will carry that on for two more revs before retiring the category. There are factors such as the market this once targeted has mostly moved to Windows and Linux x86-x64 hardware. The word is those industries don't feel like Apple is truly committed especially after the 2013 Mac Pro and 2017 iMac Pro. These businesses like upgrading components at their own leisure.

Creative professionals with existing Apple Silicon such as M1/M2 Pro/Max/Ultra Macs devices are meeting their needs.

 

[elfamosisimoJON]

its obvious that the new Mac Pro was just released to complete the transition, i would expect a proper redesign with all of those things (maybe except the memory modules) in 2025, maybe with an M4 Extreme, skipping the M3 entirely.

 

[Mr. Dee]

its obvious that the new Mac Pro was just released to complete the transition, i would expect a proper redesign with all of those things (maybe except the memory modules) in 2025, maybe with an M4 Extreme, skipping the M3 entirely.

It’s gonna get M3 Ultra, but I suspect the redesign is likely 2025 also. I honestly don’t see what was preventing them from releasing the Mac Pro with M1 Ultra last year either.

 

[Nermal]

I don't think the Mac Pro really has a future and that the 2023 will be the last.

I've privately wondered whether it was a case of "let's make it as bad as we think we can get away with, and then kill it off in a couple of years". Time will tell, and I've been wrong before.

 

[flowrider]

It’s gonna get M3 Ultra, but I suspect the redesign is likely 2025 also. I honestly don’t see what was preventing them from releasing the Mac Pro with M1 Ultra last year either.

Glad they didn't🤗 That would have meant less time for 7,1 NcMP support👍🏻 My 7,1 is going nowhere❗️

Lou

 

[ChrisA]

Four conditions for the Mac Pro to survive and be a workstation class machine:

1. A heterogeneous shared memory architecture that allows the use of both super fast main unified memory and expansion with DDR5/6 modules (something somewhat similar to what Intel uses in its latest Xeon processors)
2. A large number of real PCIe (in version 5.0 to start) lines for expansion cards instead of the current fake ones
3. PCIe accelerators – expansion cards for Metal/Neural computing
4. Double the number of CPU/GPU cores

1) Has anyone complained that they can't get their work done for of memory?
2) Which PCI cards do the actual users have that don't work in the MP's PCI slots?
3) Yes, these are good, But we do this with the network and use remote computers or the even more remote cloud. No one really wants to train networks on their normal desktop computer because it ties up the machine for days and weeks on end.
4) Yes, more is better but at some point, you offload your work to the cloud or the server room.

I think the reason you own a Mac is for interactive work. It seems a waste to run them headless with hours to weeks long tasks. Video rendering is kind of an exception because it is a function of video editing which is interactive.

My main point is that the PCI slots are for legacy applications like fiber optic networks or Blackmagic video and some other special cases and these cards already exist. If your media servers push the data out over fiber, each workstation MUST have fiber. That is why the PCI cards.

My opinion? I think Apple will abandon the Mac Pro after the need for these legacy PCI cards goes away.

PCI was a technology for PCs to allow reasonably fast expansion to 3rd party add-on cards but Aple's GPUs have an order of magnitude faster connection and PCI is just too slow. It can not compete with unified memory

 

[Xenobius]

1) Has anyone complained that they can't get their work done for of memory?
2) Which PCI cards do the actual users have that don't work in the MP's PCI slots?
3) Yes, these are good, But we do this with the network and use remote computers or the even more remote cloud. No one really wants to train networks on their normal desktop computer because it ties up the machine for days and weeks on end.
4) Yes, more is better but at some point, you offload your work to the cloud or the server room.

1) Yes, there are applications that need a lot of memory. Apple praised the Mac Pro's ability to install 1.5TB of RAM to load large sound libraries. Others: virtualisation, scientific computing, machine learning, etc.
2) Primarily GPU-based computing accelerators, but also other high-bandwidth-intensive add-in cards. Currently, if you add one high-speed RAID card to a Mac Pro 2023, there is no more bandwidth for any other card.
3) Apparently those who use multi-GPU configurations for machine learning disagree with you. The same is true for 3D rendering. Of course you can use external services, but it is also convenient to have a quick view of the 3D scene in realtime and to perform at least test renders. For these applications, Macs are currently non-existent - Macs are even dozens of times slower than PC workstations. It didn't have to be that way at all.
4) Workstation is your personal computer for work - the faster it is, the faster you will get the job done. A properly configured Workstastion PC can be many times more powerful than a Mac. This is the sad reality, unfortunately. Apple's processors are awesome for laptops and mini desktops, but when it comes to Workstations, they completely fail. I realise it's not a profitable market for Apple, but I think they should treat the Workstation market in a similar way to how car manufacturers treat F1. Or space technology development.

Graphics cards or compute cards (without video outputs), have internal bandwidths up to many times greater than the GPU in Apple Silicon and are connected to the system via PCIe5. Understand that it is not a question of whether it can be done, but whether Apple will find it profitable for them. It is very possible that Apple has reallocated all possible engineering resources to work on Vision Pro, as this product is now the most important. By doing so, they have sacrificed the Workstation market (hopefully temporarily).

 

[iPadified]

I realise it's not a profitable market for Apple, but I think they should treat the Workstation market in a similar way to how car manufacturers treat F1. Or space technology development.

Only if the technical solution found in workstations can be applied to the iPhone and that is hard to see. No drive for efficiency in WS, no optimisation of a balanced SoC to solve everyday task for normal usage, low consideration about costs, the focus is on scaling upwards and not downwards and hence everthing lower than a desktop chip is not considered when designing solutions. To me the Mac Pro is a really poor development platform to find new tech solutions for the iPhone and ergo not interesting.

 

[Xenobius]

Only if the technical solution found in workstations can be applied to the iPhone and that is hard to see. No drive for efficiency in WS, no optimisation of a balanced SoC to solve everyday task for normal usage, low consideration about costs, the focus is on scaling upwards and not downwards and hence everthing lower than a desktop chip is not considered when designing solutions. To me the Mac Pro is a really poor development platform to find new tech solutions for the iPhone and ergo not interesting.

Fair point, but note that the Mac Pro is the development platform (software, 3D, AI...) for Apple's consumer devices.
That may not be the most important thing, but it's also about the company's image. Apple has shown its most powerful machine, that is far, far behind Intel/AMD/Nvidia performance in the Workstation market. It simply looks terrible.
I hope that Mac Pro 2023 is an accident at work, that it's not the case that Apple released this crap just to prove that no one needs the Mac Pro.

 

[Anaxarxes]

Four conditions for the Mac Pro to survive and be a workstation class machine:

1. A heterogeneous shared memory architecture that allows the use of both super fast main unified memory and expansion with DDR5/6 modules (something somewhat similar to what Intel uses in its latest Xeon processors)
2. A large number of real PCIe (in version 5.0 to start) lines for expansion cards instead of the current fake ones
3. PCIe accelerators – expansion cards for Metal/Neural computing
4. Double the number of CPU/GPU cores

Workstation class machine is a super general term.

What is a Workstation? A CAD/CAM operator, a GIS machine, a 3D Video creator, an Audio Station, Broadcasting?

I think mac Pro is super specialised on video creation/broadcasting and photo editing. But not so much a good contender on 3D (at least not yet with 2nd gen AS)

#1 i think is a great idea until the main unified memory can be expanded to multiple terrabytes which is costly and difficult to hande on wafers alogn with SoC.

#2. They are real PCI slots. Verison 5.0 is just gettting support on PCIE cards today so no big miss there and the PCIE bandwith is not the limiter for more performance apart from SSD arrays. YOu mean external GPU support? I think that is not going to be supported, period. Maybe if M3 Extreme can do wonders if realized.

#3 Yep, why not? Having 7 afterburners embedded is good. Now why not option to add more?

#4 3rd gen AS

 

[mectojic]

We're not getting a Mac Pro redesign soon. A reminder that the Power Mac G5 / classic Mac Pro lasted 9 years.

The current Mac Pro is an extremely well-engineered product for such a niche market, and R&D wouldn't justify a new design when the current one does all it needs to, while having plenty of overhead for potential future M3/M4 Pros that 'might' be more upgradeable.

 

[Spaceboi Scaphandre]

Four conditions for the Mac Pro to survive and be a workstation class machine:

1. A heterogeneous shared memory architecture that allows the use of both super fast main unified memory and expansion with DDR5/6 modules (something somewhat similar to what Intel uses in its latest Xeon processors)
2. A large number of real PCIe (in version 5.0 to start) lines for expansion cards instead of the current fake ones
3. PCIe accelerators – expansion cards for Metal/Neural computing
4. Double the number of CPU/GPU cores

Outside of increasing the CPU/GPU cores, none of anything else you said needs to happen is gonna happen, especially the accelerator cards. Apple's made it clear they have no interest in supporting the high end workstation market anymore outside of video editors and music professionals since it's not profitable compared to the much cheaper Mac Studio and 14/16 inch Macbook Pros. The Mac Studio nowadays is more the Mac Pro and the Mac Pro only exists out of obligation for those who need slots for BlackMagic cards.

When the M3 Mac Pro arrives it's most likely getting a full redesign to reflect this. Half the size shrunk and the expansion even more limited, but the price potentially lower. Or they just cut the Mac Pro entirely and just rebrand the Mac Studio as the Mac Pro. Either way, you're not getting another machine like the 2019 Mac Pro or the original cheese grater designs before the trash can.

But in the end, the market who needs the Mac Pro has shrunk in the last 4 years, which is why the rumored M2 Extreme got cancelled since it wasn't worth it in their eyes to waste all that R&D making a superpowered chip hardly anyone is gonna buy. Eventually the Mac Pro as you know it is gonna go the way of XServe when Apple's data center market shrunk too.

 

[Spock]

I honestly think that Apple should have kept the Mac Pro on Intel and maybe some kind of updated afterburner card that added Apple Silicon compatibility.

 

[Spaceboi Scaphandre]

I honestly think that Apple should have kept the Mac Pro on Intel and maybe some kind of updated afterburner card that added Apple Silicon compatibility.

There's no point in the afterburner card anymore. The thing was made into retroware the moment the M1 Pro came out with it's Media Engine. The Media Engine in Apple Silicon does everything the afterburner card did but better and faster in a low wattage package, and the M2 Ultra is equivalent to 7 of these cards.

 

[brofkand]

None of these workstation use cases run macOS so the Mac Pro is essentially irrelevant. It's really telling that one of your bullet points is to run Windows natively. It's really just a machine for YouTube influencers to have on the desk behind them. It's their halo product.

 

[ThunderSkunk]

I don't have a reason to buy a mac pro that isn't intel based, but after they burned the Intel bridge (& the Nvidia bridge before that) there's no chance. To satisfy both companies it'd take forging some kind of weird hybrid Intel + Apple Silicon homonculus in the fires of deepest hades, and there's just no way it's worth the effort to Apple. Look at their revenue streams. iPhone, services, etc... then there's the lowly mac trudging along, and the pro a tiny sliver of that. Meanwhile Apple's been beating a "post-pc era" drum since 2010, pretty upfront about the mac in general being overkill for most people/most of the time. The company Apple has turned into would be wholly justified just spending as little money as possible selling the least amount of hardware possible to the uncritical masses for as much as possible and raking in shareholder gambling $$$. The mac pro, and probably the mac entirely, fit into that scheme less and less with each year, and each new product category that they'd rather sell.

For me, Apple is now a good phone/watch/AR/VR headset company at this point. But for work, I guess it's going to be a Surface + Nvidia/Intel workstation.

 

[Rob__Mac]

I am very happy with my 2019 Mac Pro. The 2023 doesn't add any performance gains that I'd see, and my AMD GPU's are faster for 3D rendering than the M2 Max. However, Apple hasn't added support for the newest AMD GPU's, so it looks like I might be at the end of the line with my current setup in terms of upgradeability.

If they make it so I can have vastly more GPU cores than the M2 Max supports, I would have no issue in continuing my 20+ years working on a Mac.

The way things are looking at the moment, I'll be speccing up a PC in 2025…

 

[TechnoMonk]

Future of Mac Pro.

1. Keep increasing RAM in each iteration.
M3: 256
M4: 512? 384?
M5: 768?
M6: 1024?
2. In crease CPU cores, add RT and tensor cores in GPU.
3. External RAM isn’t coming.
4. No external GPU support. Probably 0.001% chance of AMD GPU. NVidia:0%.

 

[Mr. Dee]

Glad they didn't🤗 That would have meant less time for 7,1 NcMP support👍🏻 My 7,1 is going nowhere❗️

Lou

Taking into account they are dropping Intel Macs from 2016 on year at a time (they are now at specific 2017 Macs), next year 2018 Macs will be dropped, and most 2019 Macs will be dropped in 2025. Apple was still selling Intel Macs in 2020 and the Intel Mac Pro was launched at the end of 2019. That means Apple is likely to keep support up 2028. macOS support might drop off around 2026, but security updates will be maintained until then.

 

[DrEGPU]

I don't think the Mac Pro really has a future and that the 2023 will be the last. Realistically they should have called it the Mac Studio Pro.

More like the Mac Studio Pro(Res)

 

[leman]

Four conditions for the Mac Pro to survive and be a workstation class machine:

1. A heterogeneous shared memory architecture that allows the use of both super fast main unified memory and expansion with DDR5/6 modules (something somewhat similar to what Intel uses in its latest Xeon processors)
2. A large number of real PCIe (in version 5.0 to start) lines for expansion cards instead of the current fake ones
3. PCIe accelerators – expansion cards for Metal/Neural computing
4. Double the number of CPU/GPU cores

I agree with 4. and partially agree with 2. (I don’t think it really needs too many PCIe lanes, but it would be nice to have at least a few more). I doubt that 1. is that important in the large scale of things (although nice to have) and 3. just doesn’t make much sense for me. The industry is moving towards unified memory compute as bandwidth starts to be one the limiting factor. Why would you want Apple to pursue outdated trends?

What do I want to see? A truly modular system with pluggable SoC boards and high bandwidth bridges. RAM expansion could be done via PCIe. Will Apple go for it? Kind of doubtful given how much investment it would require for a very dubious return, besides the coolness factor of course.

 

[iPadified]

Fair point, but note that the Mac Pro is the development platform (software, 3D, AI...) for Apple's consumer devices.
That may not be the most important thing, but it's also about the company's image. Apple has shown its most powerful machine, that is far, far behind Intel/AMD/Nvidia performance in the Workstation market. It simply looks terrible.
I hope that Mac Pro 2023 is an accident at work, that it's not the case that Apple released this crap just to prove that no one needs the Mac Pro.

I agree. It is much better chance that Apple provides a sufficiently high end Mac Pro for development for iDevices than a a generic competitor to Intel/AMD/Nvidia WS. If Apple decides that 3D rendering/AI is central for iDevice development and that MacOS is necessary, they will solve the lack of top performance. Well, at a rather high price.

"Accident at work", I agree : it is more boring than the Mac Studio external case design which is basically an extruded Mac mini with some new holes in it. Something must have gone wrong. This would not be boring:

A truly modular system with pluggable SoC boards and high bandwidth bridges. RAM expansion could be done via PCIe. Will Apple go for it? Kind of doubtful given how much investment it would require for a very dubious return, besides the coolness factor of course.