MP All Models Is the Mac Pro still relevant in 2025

[Terraaustralis]

Is the Mac Pro relevant in 2025?

I think not.

As theluggage and Apple Knowledge Navigator say, the M4 Studio answers the demand originally targeted by the MacPro.

From my perspective, the ARM solid state M series changes everything. The power of Apple silicon and the ‘configure at time of purchase’ approach hugely simplifies choice.

Today, a consumer simply asks ‘Do I want a desktop, laptop or mini/studio with a dedicated monitor’.

According to work requirements, any of the above three options can be purchased configured at minimal base entry level up to monster multicores. The previously complex line up of CPU/GPU graded products with Entry and Pro in different Apple product classes are gone.

Choosing a Mac now is much easier and the product will better match individual requirements.

The question of affordability and value for money is however, an entirely different discussion.

 

[Flint Ironstag]

It's SUPER relevant to me as long as peak single precision TFLOPS maxes out at what - sub 20 TFLOPS maybe on Mx chips?? 🤣

 

[theluggage]

Then again, an XBox can outperform anything Apple makes at the things it's good at, so is it a powerful computer in its own right?

Yes, of course it is. Running video games is computing, and architecture-wise its much the same as a "proper" PC - certainly one that was assembled with gaming as a priority. The only thing stopping it being a useful general purpose computer is MS's artificial & arbitrary restrictions on what software it can run, much like the iPad. In the past, it has been possible to run full-blown Linux on the XBox.... Or maybe, by your argument, Windows PCs are just re-skinned XBoxes?

If Macs and iPads cost as little as an XBox, maybe it wouldn't be such a problem for them to be sealed & disposable.

You seem to have your own personal definition of "computer" that requires internal expansion. That wasn't even true in the 1980s...

I think you're crafting your definition of a tradesman's van metaphor to specifically exclude Apple's hardware from it.

By Tradesman's van I mean every slotbox - from i5 single slot, all the way up to dual processor 9 slot sytems.

I'm 'crafting my definition'??? Here's your original analogy:

A Mac Pro is Apple's plain white tradesman's van. The Volkswagen Crafter / Transporter, or Ford Transit.

You specifically referenced two iconic mass-market, small vans (and the source of the "white van man" meme). Now you want "white van" to cover everything from a hatchback car to an articulated lorry? Sorry, no. If you move the goalposts that much then it becomes worthless as an analogy.

The 2019 Mac Pro was priced comparably to Lenovo's P-series & HP's Z-Series.

A basic Xeon workstation with 3-4 PCIe slots could be had for $3000 or so in 2019. A Core-i or AMD tower could be had for around $1000. Those are your (original definition) "white vans".

The 2019 Mac Pro started at $6000 for a base configuration that would barely out-perform an i7. That price could be justified by the use of the (then) brand new Xeon-W3000 processir with greatly increased PCIe bandwidth and RAM capacity - meaning the MP had 8 PCIe slots, could support multiple PCIe x16 GPUs and take up to 1.5TB of RAM. The 24 core cpu options used the M-suffix Xeon-W chips to get support for >1TB of RAM (which added thousands to the cost of the chip c.f. the 1TB max versions). If you hunted down a Xeon-W3000 system with comparable PCIe and RAM capacity then it probably would be "competitive" - but still way out of the original "white van" pricerange - and virtually useless unless you're going to fill it with thousands of bucks worth of RAM and expansion cards.

Funny you mentioning everyone going to laptops, partner's game studio is dumping the Core i9 / 4070 laptops they switched everyone to, and switching everyone back to minitowers.

...sure, serious PC gaming being one of the last remaining consumer niches for PCIe tower PCs - a shrinking pond that's still worth targeting because the users have deep pockets & spend $silly continually upgrading their GPUs to get a few more FPS (so I guess developers have to keep up). I suspect that an increasing number of their customers are using laptops and handheld gaming systems, though.

Literally every component maker; motherboards, the PCI standard, graphics cards, storage vendors, RAM standards... the entire tech industry (apart from Apple) are putting all their money into making better white vans.

Ok, so given that many component makers are self-evidently putting significant money into developing new CPUs, GPUs etc. for laptops & phones (e.g. Qualcomm's Snapdragon X in MS's Copilot+ laptops), Samsung et. al. are continuing to turn out new non-Apple phones with better processors, graphics and displays, Intel and the USB-IF are continually expanding Thunderbolt and USB4 (external PCIe-class expansion for laptops, SFF etc.) while NVIDIA et. al. are pouring money into AI products for everything from cars to big-iron supercomputers... that would mean that your already elastic "white van" analogy now covers all of these things...

Nah.

 

[mattspace]

Yes, of course it is. Running video games is computing, and architecture-wise its much the same as a "proper" PC - certainly one that was assembled with gaming as a priority. The only thing stopping it being a useful general purpose computer is MS's artificial & arbitrary restrictions on what software it can run, much like the iPad.

But I wouldn't use use a games console as the architectural basis of a Mac Pro.

You seem to have your own personal definition of "computer" that requires internal expansion. That wasn't even true in the 1980s...

No, my definition is specific to what should be called a Mac Pro.

You specifically referenced two iconic mass-market, small vans (and the source of the "white van man" meme). Now you want "white van" to cover everything from a hatchback car to an articulated lorry? Sorry, no. If you move the goalposts that much then it becomes worthless as an analogy.

You are misrepresenting that my definition was about the paradigmatic nature of the vans, not their size.

Size of van is irrelevant, A VW Crafter, Mercedes Sprinter, Even a Citroen 2CV Van, Ford Escort Van or Original Austin Mini Van, they are all the exact same paradigm - a shell, with minimal prebuilding, where everything is a post-purchase reconfigurable add-in.

A Mac Pro is Apple's plain white tradesman's van. The Volkswagen Crafter / Transporter, or Ford Transit. It's a blank canvas with a stripped-out interior providing the bare minimum scaffold for the owner to make the business plant and equipment that suits them (documented drill-safe points in the van's body framework for fitout, etc). It's never been the fastest machine in the range for more than a couple of months after release, it was always the post-purchase reconfiguration machine (granted the 2013 let you change the ram).

That is the sum total of the metaphor I made, all this size, colour and price malarkey is torture you have brought to the metaphor, presumably in an attempt to discredit it.

A basic Xeon workstation with 3-4 PCIe slots could be had for $3000 or so in 2019. A Core-i or AMD tower could be had for around $1000. Those are your (original definition) "white vans".

These are all white vans, by structure.

The 2019 Mac Pro started at $6000 for a base configuration that would barely out-perform an i7.

The 2019 Mac Pro was effectively the same price as the equivalent configuration Lenovo P-Series, or HP-Z.

The mac Pro had ECC RAM. If you need ECC RAM the i7's performance is irrelevant.

That price could be ... expansion cards.

All irrelevant.

...sure, serious PC gaming being one of the last remaining consumer niches for PCIe tower PCs - a shrinking pond that's still worth targeting because the users have deep pockets & spend $silly continually upgrading their GPUs to get a few more FPS (so I guess developers have to keep up). I suspect that an increasing number of their customers are using laptops and handheld gaming systems, though.

"Everyone is gaming on handheld systems" is a hype cycle, like blockchain, cryptocurrency, NFTs, and AI. This studio is pivoting to PC and Console, because Mobile is a low-return, low-prestige casino cesspool for everyone who is not Fortnite and COD.

The systems they use to develop the games don't need to be the devices people play games on. Apple Arcade titles aren't actually made on Apple computers. They're made on Windows PCs, overwhelmingly. And as this studio is finding, laptops for gamedev (the entire chain of gamedev, especially the Art department) have a venn diagram overlap of expense, low performance, poor portability, and no real advantage over a desktop system when everyone working remotely is VPNing in to the company servers anyway, so there's no point in a computer which moves.

Ok, so given that many component makers are self-evidently putting significant money into developing new CPUs, GPUs etc. for laptops & phones (e.g. Qualcomm's Snapdragon X in MS's Copilot+ laptops), Samsung et. al. are continuing to turn out new non-Apple phones with better processors, graphics and displays, Intel and the USB-IF are continually expanding Thunderbolt and USB4 (external PCIe-class expansion for laptops, SFF etc.) while NVIDIA et. al. are pouring money into AI products for everything from cars to big-iron supercomputers... that would mean that your already elastic "white van" analogy now covers all of these things...

Nope, that's all happening as well, but when you get a new Thunderbolt in your laptop, it's the half-chewed table scraps of the version of PCI that was made for Desktop several years prior. Your entire expansion bus, and everything connected to it, is sitting on one quarter of what one device gets in a white van computer.

Anyway, I'm kinda done with this because your whole shifting of goalposts by trying to introduce pricing, and colour options to the original tradesman van metaphor, when it was clearly explained that it was about the structural paradigm of the devices seems frankly disingenuous to me.

My case:

• A "tradesman van" slotbox is a paradigm of computer in which individual components, processor, memory, storage, graphics etc are user replaceable and upgradable on standard connections. That structural architecture is the only criteria that is relevant to the metaphor and categorisation.
• The purpose of these machines is explicitly to express this structural paradigm, because it is inherently useful and utilitarian in of itself. Fast, slow, cheap, expensive; these are all just variants of the paradigm.
• Every Mac Pro, save for the 2013 and 2023 has explicitly followed this paradigm. As has every Powermac (and Performas) in the 603/604 onwards era (prior machines had soldered processors which could count them out). Whether Apple marketed them for this characteristic or not is irrelevant. Whether a person thinks customers bought them for this characteristic or not, is likewise irrelevant. The capability was there, regardless.

If you are reliant on the Apple ecosystem, you are a coerced participant in the market. We know how coerced the Apple customer base is, how strongly Apple believes its customer base is coerced because they literally said as much in their internal emails (eg iMessage for Android), as revealed in multiple trials.

Apple apologists argued that Butterfly keyboards were popular and beloved, because Macbooks kept selling despite the keyboards being unfit for sale. In a single-supplier, competitively distorted market customers buying the only product choice offered is not a self-fulfilling confirmation of its popularity.

Arguments that Apple's current paradigm is what people "like" or want, because it sells, when it's the only option Apple is selling are similarly facile.

I would bet the Ultra Studios don't sell any better than the 2019 Mac Pros. We've seen once, directly what happened when Apple offered a compact "Pro" machine at the same time as the slotbox with the same specs and same price - The G4 cube was a sales and technical disaster, because no one valued the small size when they had to give up the expansion capabilities.

Is the Mac Pro still relevant in 2025? Not the one that Apple is selling. They don't even bother to test NVME storage, which is the only non-thunderbolt use case for PCI cards on an AS Mac.

 

[Joe The Dragon]

The 2019 Mac Pro was effectively the same price as the equivalent configuration Lenovo P-Series, or HP-Z.

The mac Pro had ECC RAM. If you need ECC RAM the i7's performance is irrelevant.

The mac pro had an high start point Other system had starting points and configs that started at much lower pricing with more choice.
ALSO did not have storage at apples pricing that is like 3-4X what you can get on your own.

 

[M4th]

It's SUPER relevant to me as long as peak single precision TFLOPS maxes out at what - sub 20 TFLOPS maybe on Mx chips?? 🤣

Maybe Apple won't ever deliver (and for good reason as many have expounded on), but ASUS seems to want to take up the higher end desktop mantle?

 

[m1maverick]

It's SUPER relevant to me as long as peak single precision TFLOPS maxes out at what - sub 20 TFLOPS maybe on Mx chips?? 🤣

Unless I am misunderstanding what you're saying it seems to me the M4 Studio Max would be more relevant.

 

[mattspace]

The mac pro had an high start point Other system had starting points and configs that started at much lower pricing with more choice.
ALSO did not have storage at apples pricing that is like 3-4X what you can get on your own.

Yes, Lenovo and HP both make lower end (and higher end) P and Z series products, respectively.

But the P and Z series that were of comparable config to the Mac Pro, were comparable in price. Here in Australia the local arms of Lenovo and HP didn't offer configure to order options for P and Z machines for a long time - all they offered were fixed good, better, best configs, which really tracked strongly to Mac Pro pricing.

No one's ever forced you to use Apple's storage - I think if we look at folks here, for a lot of us the built in storage is really just a giant bootrom, and we run the systems from NVME drives.

 

[Terraaustralis]

’Is the Mac Pro still relevant in 2025? Not the one that Apple is selling. They don't even bother to test NVME storage, which is the only non-thunderbolt use case for PCI cards on an AS Mac.‘

I am curious, what are your grounds/evidence for saying Apple does not test NVME storage? Please explain.

 

[mattspace]

I am curious, what are your grounds/evidence for saying Apple does not test NVME storage? Please explain.

Go read the very long thread in this forum about NVME issues under Sonoma. The TLDR version, from memory, is Apple replaced the external storage drivers in macOS with new versions derived iOS, which generally works fine over thunderbolt, but when used for internal PCI storage resulted in storage failing to register on boot, disappearing during use, and requiring a full power-off and power-cord-pull to recover. It seems to have stabilised in Sequioa, but there was more than a year of the primary use case for the 2023 being unfit for purpose.

 

[iPadified]

Oh, and I've checked on Wikipedia and a handheld cellular phone cost $4,000 in 1984 - IBM PC money, even then. So not sure what your point is there. Anyway, a modern smartphone is a computer - the actual phone bit is rapidly sliding down the priority list.

Point is: for every new CPU/GPU generation, fewer and fewer need the high ends and hence Mac Pro like computers loose relevance. Desktops as a segment looses relevance. Add to that cloud based compute solutions will diminish local compute capacity.

Yes cellular phones anno 1984 was not computers and now they can theoretically run MacOS.

 

[theluggage]

The 2019 Mac Pro was effectively the same price as the equivalent configuration Lenovo P-Series, or HP-Z.

"Equivalent configuration" is doing a lot of work in that claim.

The 2019 Mac Pro is a specific model with a specific logic board, a specific case, a specific power supply and a choice of about 4 Xeon W3000-series plug-compatible processors.

P-Series/Z-Series are Lenovo/HPs catch all name for "pro" computers and currently include everything from "laptop workstations" through small-form-factor desktops to proper rackmount systems (i.e. not just a desktop tower turned on its side). There are dozens of completely different base models to choose from, before you get to the lost of cpu/gpu/ram options that are available for that model. To find a P/Z series with an "equivalent configuration" you'd have had to look at the top of the desktop range and find a base model with Xeon W3000 processors, 8 PCIe slots with an unusually high proportion of 8 and 16-lane slots and sockets for at least 1.5TB of RAM. Compared to a ~$3000 basic Xeon tower "P/Z series" that would mean a different motherboard, more expensive base CPU, bigger power supply and probably a different case/cooling system.

Yet, until you started using all of that extra PCIe and RAM expansion, you wouldn't see any performance boost and that $3000 system would still have a couple of free PCie slots, hard drive bays, M.2 slots and space for maybe 256GB or 512GB of RAM.

That's part of why your "white van" metaphor doesn't work for the 2019 Mac Pro (apart from the nonsense of ignoring price and size) - it was not a blank canvas, Apple have already steered you down a particular path by choosing a single Xeon-W3000 as the processor, ruling out both cheaper Xeon/Core/AMD options and more powerful Scalable Xeon multi CPU options.

Which is really Apple's problem with the 2019-style Mac Pro - they're competing with the PC market that offers something far closer to your "blank canvas" where you can choose from a vast number of permutations of case, motherboard, processor, GPU, storage... even if you can't build your own PC there are many companies that will build them to order, and while compatibility problems do exist, every PC component manufacturer has to offer Windows support. (Even with Hackintosh - or Macs that did have PCIe slots - you had to carefully pick-and-choose MacOS-compatible components).

Whatever Apple do, unless they radically change their business model, Apple would be offering a one-size-fits-all "shell" with a limited range of CPU and GPU options.

The mac Pro had ECC RAM. If you need ECC RAM the i7's performance is irrelevant.

The $5000 iMac Pro had ECC RAM and comparable performance & better GPU c.f. the $6000 base 2019 MP even before Apple bumped the base iMP to 10 core.

Most of those $2000-$3000 Xeon towers also had ECC RAM (since that was a major reason for paying a premium for Xeon) so that only really eliminates the much cheaper core-i options.

"Everyone is gaming on handheld systems"

Straw man - I never said that. They are however growing increasingly popular & reaching the "good enough for most" point where they can deliver near-photorealistic 3D games (maybe at a slightly lower frame rate and without some gameplay-irrelevant bells and whistles you'd get with a water cooled GPU the size of Manhatten).

I would bet the Ultra Studios don't sell any better than the 2019 Mac Pros. We've seen once, directly what happened when Apple offered a compact "Pro" machine at the same time as the slotbox with the same specs and same price - The G4 cube was a sales and technical disaster, because no one valued the small size when they had to give up the expansion capabilities.

Apart from the absurdity of digging back to 2001 for an example (I mean, its not like the PC market has changed at all since then) Apple have been successfully selling compact, non-expandable Mac Minis alongside "slotboxes" since 2005.

The G4 cube was a flop because it was overpriced and had widely publicised problem with developing cracks (kiss of death for a machine sold partly on its looks). Those are the two principle reasons you'll see cited in any account of its failure. I don't recall it ever being pitched as a "pro" machine - part of the pricing problem was that it was pitched more as a home machine in competition with the (cheaper) iMac.

The G4 Mac Mini was basically the same concept, better implemented, at a more reasonable price and sold successfully alongside the G5 tower. The Intel Mini sold alongside the "classic" Cheesegrater. It's nadir was probably 2014-2018 when Apple weren't making slotboxes and desperately trying to push the trashcan and iMacs as "pro" solutions - and dumped the higher-end Mini models that had been available pre-2014.

(It's also proven the perfect machine for developers who just need a Mac for MacOS/OSX builds and user testing),

Meanwhile, the Ultra Studio is Apple's most powerful current system and costs accordingly - but Apple also now have a whole range of headless desktops including the M4 Pro Mini and the M4 Max Studio which (whether or not they meet your personal "Mac Pro" criteria) are all suitable for serious creative/development work.

The systems they use to develop the games don't need to be the devices people play games on. Apple Arcade titles aren't actually made on Apple computers. They're made on Windows PCs, overwhelmingly.

So how does that support the need for a Mac Pro? The Mac is not going to suddenly change into an AAA gaming platform, and if studios do want to develop for Mac they can do it on PCs. Even if there was a suitable Mac tower, studios would still kit themselves out with cheap Windows PCs built from commodity components - and maybe a couple of Minis, Macbooks and iDevices for testing. The main case for more gaming on Apple Silicon is the speed-vs-power of Apple Silicon GPUs in MacBooks and iPads as delivery platforms. A "Mac Pro" with discrete AMD/NVIDIA GPUs has nothing unique going for it as either a delivery or development platform for games.

 

[Regulus67]

A "Mac Pro" with discrete AMD/NVIDIA GPUs has nothing unique going for it as either a delivery or development platform for games.

I don't know if it wise of me to comment here or not 🤣
But I do feel macOS is/was a very important part of why many prefer mac. And yes, even for a Mac Pro.

Do Pros need a Mac Pro?
Probably not, for all kind of reasons.

Is the Mac Pro still relevant for high-end users?
I would think the users that had the 4,1 and 5,1 for years will find the 7,1 useful.
And for everyone who use bootcamp. That's why I am commenting here.
So those users don't go "meh, I'll skip that"

I certainly feel Apple let us down by not releasing drivers at least, for newer generation AMD cards.
At least I am able to boot into Windows for gaming with my new Radeon RX 7900 XTX card. Without having to build a PC as well.

 

[keksikuningas]

I wonder if the addition of CUDA (NVIDIA's proprietary API for interacting with their GPUs) to Apple's MLX framework (Machine learning library for Apple Silicon) might signify something for Mac Pro. A big chance it's just to allow developers to work with CUDA on their MacBooks and then do the real stuff on the AI servers with NVIDIA cards, but NVIDIA GPU support would be fun to have.

I was wondering for a while if GPU support is possible for M2 Ultra on Asahi Linux. There are ARM drivers after all for both NVIDIA and AMD.

 

[Terraaustralis]

Go read the very long thread in this forum about NVME issues under Sonoma. The TLDR version, from memory, is Apple replaced the external storage drivers in macOS with new versions derived iOS, which generally works fine over thunderbolt, but when used for internal PCI storage resulted in storage failing to register on boot, disappearing during use, and requiring a full power-off and power-cord-pull to recover. It seems to have stabilised in Sequioa, but there was more than a year of the primary use case for the 2023 being unfit for purpose.

Thank You mattspace. All understood.

 

[theluggage]

But I do feel macOS is/was a very important part of why many prefer mac. And yes, even for a Mac Pro.

The question is, do enough people prefer MacOS to create a significant market to justify the substantial investment in developing a whole new class of Apple Silicon Die for a new Mac Pro (or continuing to actively support x86 forever, despite 95% of the Mac market having successfully moved to Apple Silicon).

I would think the users that had the 4,1 and 5,1 for years will find the 7,1 useful.

Yes, but that's hardly a growth market. When x86 offers more bangs-per-buck in a market where bangs-per-buck are key there is going to be a stream of people making the move to PC or Linux - or simply decide that a MacBook is now so powerful that they don't need a tower. What is going to attract new users to Mac to replace them?

This is the problem: Apple Silicon as its stands just isn't the best tool to build a multiple-discrete-GPU tower like the 2019 MP running high-end AMD and NVIDIA GPUs which is always going to be playing catchup - in terms of power, cost, software choice and hardware configurability - with Intel/AMD based towers. Currently only the Mx Ultra provides enough PCIe lanes to make a viable "big box'o'slots" and that doesn't really have enough for the sort of multiple GPU beast that you could build with Intel/AMD CPUs.

However, Apple Silicon is the best tool to build low-power ultra-portable laptops, tiny. quiet SFF systems or even 'pro' tablets that bat way above their weight compared with x86 mobile/integrated GPU systems.

So, if you're Apple, do you put your money into tablets/laptops/SFF where you actually have a sporting chance of growing market share, or into a tower system which is only really going to sell to a shrinking pool of locked-in MacOS users?

And for everyone who use bootcamp. That's why I am commenting here.

Even on Intel, dual-booting Windows was an increasingly niche use as virtualization tools improved - and was really a bit of "low hanging fruit" that Apple could offer easily because Intel Macs were a gnats whisker away from being PC clones. Bear in mind that "bootcamp" was just a set of easy-to-use helper tools for partitioning your drive, tweaking the Windows installer and installing a bunch of Intel and 3rd Party Windows drivers, plus a bog-standard BIOS Emulation module for EFI that had been left off early Intel Macs. You could do it yourself (same procedure as installing Windows on some non-standard PC hardware) if you were a massochist.

The #1 reason there is no bootcamp equivalent for Apple Silicon is that there is only one viable "bare metal" alternative operating system to install alongside Mac OS - Asahi Linux (which I assume comes with its own installer tools). There's nothing (apart from rumoured MS/Qualcomm agreements) stopping Microsoft from producing "Windows 11 Mac Edition" apart from the hard work of producing stable hardware drivers for Apple Silicon & the fact that, in many cases, virtualising it under a MacOS host is a far more practical solution.

A big chance it's just to allow developers to work with CUDA on their MacBooks and then do the real stuff on the AI servers with NVIDIA cards, but NVIDIA GPU support would be fun to have.

First question is whether CUDA support (which AFAIK runs best on NVIDIA hardware anyway) would actually take advantage of Apple Silicon's somewhat different GPU/Neural Engine/Unified RAM architecture and provide adequate performance. Anyway, the value of developers to Apple is that they develop Apple Silicon optimised code that provides a good client experience on Macs and iDevices.

However, for development/training why not just use the Mac as a front end to the AI server (whether its in the cloud or a private machine room)? The performance bottleneck isn't going to be the code/images/video etc. running over your broadband link, its the bandwidth between the AI server and its data (which will most likely also be on the interweb or in the machine room rack).

I was wondering for a while if GPU support is possible for M2 Ultra on Asahi Linux. There are ARM drivers after all for both NVIDIA and AMD.

I can't imagine it is impossible but what hardware is it going to run on without being knobbled by PCIe bandwidth and/or sky-high costs?

The only ASi chips with enough spare PCIe lanes to support a single, full width (x16) GPU are the Ultra variants which cost an arm and a leg for just 24 spare PCIe4 lanes. You're paying Xeon prices to get i7-level PCIe bandwidth.

 

[Regulus67]

The question is, do enough people prefer MacOS to create a significant market to justify the substantial investment in developing a whole new class of Apple Silicon Die for a new Mac Pro (or continuing to actively support x86 forever, despite 95% of the Mac market having successfully moved to Apple Silicon).

I suppose not. Looking at the Pro lineups the past decade. Where they offered only one machine, with no follow ups.
But the Mac Pro 7,1 had GPU upgrades. Four AMD generation MPX cards exist, with full support.
GCN 4.0, GCN 5.1, RDNA 1.0 and RDNA 2.0 generations.

I ran the DaVinci Resolve benchmark from team2films on my Mac Pro (24-core).
With 2 Radeon Pro W6800X Duo cards, it has the same score as ONE Radeon RX 7900 XTX card!

Pro users would probably have loved the extended life their systems would get, if Apple supported RDNA3.0 cards.

test A 296 seconds - Ryzen 9 5900X 64GB RAM Radeon RX 7900 XTX - 19.1 Studio (@vwgtidude)
test A 299 seconds - M1 Ultra (20 CPU 48 GPU) - 19.1 Studio (@cpaq)

edit:
Apple likes to massage their comparisons.
The Mac Studio M3 Ultra is compared to a "reference" Mac Pro 7,1 16-core, Radeon Pro W5700X (RDNA 1.0)
That is not a proper comparison

 

[maflynn]

The question is, do enough people prefer MacOS to create a significant market to justify the substantial investment in developing a whole new class of Apple Silicon Die for a new Mac Pro (or continuing to actively support x86 forever, despite 95% of the Mac market having successfully moved to Apple Silicon).

What can a Mac Pro offer that a Studio cannot? back in the dark times, the Mac Pros had expandability, in ram, storage, and expansion cards.

I'm no expert but it seems many of the users of the cheese grater Mac Pro (and earlier models) relied heavily on the expansion slots, but now it seems (if youtube, and various online articles are to be believed) that many sectors have moved to thunderbolt. I'm sure there's still some industries that rely on expansion cards, but I think they're the minority (just my uneducated opinion).

My point is, if expansion slots are not needed by and large, what could a re-engineered Mac Pro offer that the Studio cannot? A discrete GPU, replaceable ram? Maybe but as you mentioned is there significant market to justify that move, and if they go down that path would that also be open to the other Mac Models - if so, then that brings us back to the original question. What would the Mac Pro offer that is not available in the studio?

 

[iPadified]

What can a Mac Pro offer that a Studio cannot? back in the dark times, the Mac Pros had expandability, in ram, storage, and expansion cards.

I'm no expert but it seems many of the users of the cheese grater Mac Pro (and earlier models) relied heavily on the expansion slots, but now it seems (if youtube, and various online articles are to be believed) that many sectors have moved to thunderbolt. I'm sure there's still some industries that rely on expansion cards, but I think they're the minority (just my uneducated opinion).

My point is, if expansion slots are not needed by and large, what could a re-engineered Mac Pro offer that the Studio cannot? A discrete GPU, replaceable ram? Maybe but as you mentioned is there significant market to justify that move, and if they go down that path would that also be open to the other Mac Models - if so, then that brings us back to the original question. What would the Mac Pro offer that is not available in the studio?

You a summing it quite nicely, fewer and fewer need internal expansion cards. With the Mx release, Apple is not interested in internal GPU expansions from third parties making Mac Pro less appealing for Apple.

 

[theluggage]

I'm no expert but it seems many of the users of the cheese grater Mac Pro (and earlier models) relied heavily on the expansion slots, but now it seems (if youtube, and various online articles are to be believed) that many sectors have moved to thunderbolt. I'm sure there's still some industries that rely on expansion cards, but I think they're the minority (just my uneducated opinion).

I think we're basically agreeing with each other that the Mac Pro tower is a transitional product for a shrinking market that needs non-GPU PCIe cards- but the math on expansion cards is that TB is still a long way behind internal PCIe on total bandwidth and bandwidth-per-lane. An Xeon or Threadripper tower workstation can provide 64 lanes of PCIe 5 (and PCIe 6 is coming Real Soon Now) - a Thunderbolt port can provide the equivalent of 4 lanes of PCIe and the new shiny Thunderbolt 5 has just upgraded from PCIe3 to PCIe4.

GPUs and some specialist I/O cards can use up to 16 lanes of PCIe each. A single high-end M.2 SSD can use 4 lanes. Most will work with fewer lanes, but at reduced bandwidth. Devices can share lanes via a switch - at reduced bandwidth.

So a tower workstation system still offers a lot more expanson bandwidth for interface cards and storage than you can get with 4-6 thunderbolt ports, and many individual PCIe cards can use more bandwidth than a single TB port would provide. You can get TB to PCIe enclosures with 16x PCIe slots but they only actually run at x4 bandwidth.

Then there's the appeal of having all of your cards & storage in a single unit (and storage pretty much == PCIe now if you want lots of high-speed storage) vs. multiple trailing TB and power cables.

My point is, if expansion slots are not needed by and large, what could a re-engineered Mac Pro offer that the Studio cannot? A discrete GPU, replaceable ram?

A discrete GPU would need a 16 lane PCIe slot. 2019 Mac Pro users were plugging in multiple discrete GPUs.

There's a bit of chicken-and-egg here: Apple not supporting discrete GPUs has ruled out one major use case for multiple 16x PCIe slots. U turning on that would also U-turn one of the key features of Apple Silicon which is having the CPU, GPU, SSD controller and neural engine sharing the same unified RAM. I guess you could have external unified RAM but there are speed/power advantages to LPDDR RAM which has to be soldered (press-fit LPCAMM modules don't seem to have caught on). The glass half empty is CPU RAM topping out at 512GB - the glass half full is GPU RAM going up to 512GB, which could be good for some AI applications.

I don't think non-upgradeable RAM is really a fundamental problem - people paying for stratospheric amounts of RAM will be doing the research before buying. The problem is where non-upgradeability meets Apple's silly RAM prices.

In all these cases, you're ot just talking about a re-engineered Mac Pro you're talking about a new class of Apple Silicon SoC with more PCIe lanes, support for external RAM etc.

My suggestion for a new Mac Pro would be the equivalent of a Mac Studio in a 1U 19" rack mount format that could fit into a cheap(er) lightweight rack rather than a full blown machine room rack. I.e. the same sort of size as most rackmount A/V units - and a range of matching TB storage, hubs etc. so everything could go into a rack with the cabling hidden at the back.

Or maybe a Eurocard-sized Mac & peripherals so you could have a bunch of them in a 3U rack unit...

 

[PowerMike G5]

What can a Mac Pro offer that a Studio cannot? back in the dark times, the Mac Pros had expandability, in ram, storage, and expansion cards.

I'm no expert but it seems many of the users of the cheese grater Mac Pro (and earlier models) relied heavily on the expansion slots, but now it seems (if youtube, and various online articles are to be believed) that many sectors have moved to thunderbolt. I'm sure there's still some industries that rely on expansion cards, but I think they're the minority (just my uneducated opinion).

My point is, if expansion slots are not needed by and large, what could a re-engineered Mac Pro offer that the Studio cannot? A discrete GPU, replaceable ram? Maybe but as you mentioned is there significant market to justify that move, and if they go down that path would that also be open to the other Mac Models - if so, then that brings us back to the original question. What would the Mac Pro offer that is not available in the studio?

I agree. As someone who has owned mostly all the tower Mac Pros, it was always because of the expansion capabilities.

Now that I've been on the Mac Studio for some time, it's the first desktop Mac that hasn't made me miss the tower Mac Pro. Thunderbolt is getting fast enough for expansion peripherals and I have been able to use my PCIE cards in external thunderbolt enclosures just fine.

The main thing is the M-series soocs have allowed for the kind of power that was expected in the Pros to fit in these small all-in-one machines and at price points that are a good deal lower than what was expected. Using a $4K Mac Studio that's quite a bit faster than the $12K I put into my 2019 Mac Pro has really warped expectations of the Pro tower.

A $2K base M4 Max Mac Studio has faster computing power than the 2019 28-core Mac Pro, which was some insane $7K upgrade on the base towers price of $6K. It really makes the business use case really narrow now for the Mac Pro.

I'd love to see the Mac Pro return to glory, but I don't know if Apple really cares that much, based on history. They'd need some serious differentiation from the studio to widen and bring back appeal. Like if they created a machine that had a few PCI slots, but then also had a way to use their ultra fusion connect as additional slots where you can upgrade with custom Apple GPUs, RAM, and media engines that further enhance the base sooc in the machine. I suppose that'd be the way Apple can accommodate pro users while keeping everything on their hardware.

 

[avro707]

I won’t go Mac Studio because native windows does not work.

Yet people say, oh, virtual machine windows. But then say elsewhere “build a PC desktop”. But why do that, wasn’t the virtual machine so great? Oh, your use case isn’t necessary anyway.

I need real GPUs in Windows, virtual machines and the silicon macs cannot do that. If Apple would get off its stubborn high horse and put some 7900 Radeon Pro drivers in macos (or 7800) or even the regular consumer versions of either that would keep me on macOS for longer and maybe buying at least some other Apple products.

But without that, I am not a future Apple customer. Next screen will not be Apple for instance. Next desktop will be a Lenovo workstation.

A $2K base M4 Max Mac Studio has faster computing power than the 2019 28-core Mac Pro

Why compare a 2025 machine to a 2019 one?

I am sure a 2030 model Mac Mini base spec will trash the M4 Max studio at less cost as well, but of course that comparison is irrelevant - technology moves on and what was the top version, thst performance becomes available to everyone eventually.

And would a cheaper PC also be just as fast as a Studio?

 

[m1maverick]

I'm no expert but it seems many of the users of the cheese grater Mac Pro (and earlier models) relied heavily on the expansion slots, but now it seems (if youtube, and various online articles are to be believed) that many sectors have moved to thunderbolt. I'm sure there's still some industries that rely on expansion cards, but I think they're the minority (just my uneducated opinion).

The question is did they do so willingly or because they had no other option?

 

[maflynn]

The question is did they do so willingly or because they had no other option?

Does it matter? either way the end result is that PCI cards are not much of a need as they once were. I'm sure there's still some professions, where its adventitious to use them.

The bottom line as I see it, is that the $6,600 plus MacPro has little to no advantages over the Studio which base model costs about a 1/3 of the Mac Pro

 

[m1maverick]

Does it matter? either way the end result is that PCI cards are not much of a need as they once were. I'm sure there's still some professions, where its adventitious to use them.

I think so. If the options are no longer available does that mean end users didn't want the discontinued options?

The bottom line as I see it, is that the $6,600 plus MacPro has little to no advantages over the Studio which base model costs about a 1/3 of the Mac Pro

I suspect it's because the Mac Pro is essentially a Mac Studio. Apple has effectively neutered the Mac Pro into a Mac Studio.