You’re right, the answer for “who does Apple think these computers are for” is, for most folks posing the question, “Not Me”.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Gurman: All-New Mac Pro Still in Testing, But 'M2 Extreme' Chip Likely Canceled
- Thread starter MacRumors
- Start date
- Sort by reaction score
If they throw away the performance advantages of Apple Silicon, like on-package unified RAM shared directly between CPU and GPU and GPUs optimised for MacOS/Metal, then there might not be any point transitioning the Mac Pro to Apple Silicon. Just keep supporting the 2019, which isn't really obsolete yet.
Mac Pro isn't designed as a server, it is a high-end workstation - designed for interactive use by a single "power user", and the Xeon-W isn't a server-class CPU, the 'W' stands for "workstation". Sure, you can call any computer a "server" if it's sharing files or running a web server, but computers designed as servers have quite different design priorities. The only time Apple offered server-class hardware was when they made the XServe. Heyday of the XServe was when the PPC chip was still cool, before Linux had so much industry acceptance, and MS Windows Server was knobbled by huge per-user licensing fees. Today, in server applications where the MacOS UI isn't significant, there's no advantage to MacOS on a server and no real reason for Apple to have skin in the server game.
Amazon and others already have server-class ARM chips that will out-server Apple Silicon, and run Linux which has supported ARM since the 1990s & who's developers have a culture of hardware independence. Apple would be playing catch-up & designing a chip for a tiny niche of customers who want a MacOS server.
Not Apple. Unless unless they do two U turns (a W turn?) on (a) supporting non-Apple GPUs on Apple Silicon and (b) burying the hatchet and having anything to do with NVIDA.
Your 4090 will work very nicely in a bog-standard Xeon-W or AMD Threadripper tower, and if you're doing the sort of work that needs one then it won't run significantly faster or cooler if it is plugged into Apple Silicon or x86.
For some tasks, maybe a cluster of Mx Ultras would be more power-efficient than a single 4090 - for which a 1U rackable Mac Studio or Mx Ultra "compute unit" on a MPX-like card (maybe even plugging into a 2019 Mac Pro) might do the trick.
Reality check here - apart from whatever advantage you may or may not get from the T2 chip running the SSD - the 2019 Mac Pro is just a particularly fancy-looking Xeon W tower "blessed" to run MacOS and a neat scheme for routing extra power and Thunderbolt to "MPX" cards which are otherwise the same hardware that you'd plug into a PC. They're of interest to a small but deep-pocketed niche of users with high-end pro workflows expensively tied to MacOS, a niche that is probably going to shrink. Apple could develop a custom ASi chip for that niche but without high-volume sales from MBPs etc. it would cost a fortune - and if it didn't provide a seamless transition for those complex, legacy workloads, many of the potential customers would take that as a nudge to switch to Windows or Linux.
Machines like the MBP and Studio are a much better use of the strengths of Apple Silicon - I'm not sure it makes sense trying to turn it into something it isn't.
The only reason Mac Pro should exist is expandability, as in RAM, GPU, and storage, and expansion slots. With RAM integrated into Apple Silicon, and for good reason, I am not sure how Mac Pro will accommodate RAM expansion, if any. Without that, one can make a case for Mac Studio, which could accommodate eGPU and external storage.
Why is using a 5 year old computer computer sad? The longer it lasts, the higher the quality. I still use my 2008 iMac. I can't believe this thing has lasted so long and I'm pretty sure it'll still work in 10 years. This is the reason I only buy Macs.
The competition’s workstations using amd or intel CPUs have 16-64 cores right now and soon up to 96. For gpus they easily supports at least 2 high end cards (like 4090 or rtx 6000) with a combined Compute power of more than 160 TF. And that is not even discussing the specialized Raytracing hw or tensor cores for AI. A mac pro that would only provide 2x m2 max will be a joke compared to these. Apple knows this of course and will not release a MP that is that bad. Right? There must have been many different approaches considered internally in case somehow didn’t pan out. Maybe we’ll only get a new Mac studio in spring and maybe a preview of the new MP at wwdc since they has to redo everything? The other option is that apple goes back to intel xeons with quad amd gpus just like now and just says that was always the plan 😂
John Ternus said the Ultra was the last SKU in the family of Apple Silicon. Mark is the one that pulled this speculative 'extreme' edition of thin air. But I think Apple likely might have planted the story to trace who is leaking details. That person right now is probably receiving their pink slip.
If you buy the low-end configuration of the Mac Studio, the CPU is called "Max".
Words stopped making sense to them a long time ago.
For M1. But for M2 and beyond, it’s anyone’s guess if they will switch up the SoCs.
Nobody. That GPU requires an 850 watt power supply just for it.
I presume Apple could have on-package memory up to a certain limit (<256GB) and then additional DIMM-based memory using an external memory controller (and running at slower speeds with higher latency). So that would allow a significant amount of highest-speed memory for the majority of tasks, but the ability to run truly massive amounts of memory for those tasks that need it (with a performance penalty).
The Mac Pro is aimed for macOS users who need the most power they can get. It has never been aimed at the general workstation market because PC vendors have always offered much more scalable and powerful designs.
I don't see Apple going back to Intel. Worst-case, they just put a single Ultra model SoC in it. Offering PCIe slots (even if they do not support GPU/MPX cards) would still give it value over a Mac Studio.
To be fair, John Ternus stated that well after Mark (and others) were reporting that Apple was intending to develop an M1 SoC that was equivalent to four M1 Max. So it sounds to me like Apple was working on a 4xMax/2xUltra SoC and just could not make it economically workable.
Last edited:
Yes. Was that also about the time when there wasn’t a Mac Studio and people complained about the Pro’s lackluster performance and upgradability?
Question I’d have for Gurman: when he said that the 2026 Apple Car is gonna use a chip that is “the sum of 4 of Apple’s highest-end Mac chips”, he (or his source) meant 4 x Ultra or 4 x Extreme?
What if such a chip also ends up in a MacPro? It’s all about offsetting the development cost by using it in as many products as possible. Is a 100k$ car less of a niche than a 10k$-15k$ MacPro?
What if such a chip also ends up in a MacPro? It’s all about offsetting the development cost by using it in as many products as possible. Is a 100k$ car less of a niche than a 10k$-15k$ MacPro?
IF they do it that way complexity would be the same as the M2Max and yields aren't much of an issue as any set of 4 that fails could still be cut down to 1 M2Max + 1 M2Ultra.
The switch to Apple Silicon started with a bang and ends with a whimper?
The main problem is that Mac is great only for 2D workflow, not 3D because Nvidia GPU is essential to get high performance on 3D software. Even M1 Ultra is not even close to RTX 3090 in 3D and many 3D software aren't even interested in Mac so far. Blender? With M1 Max, it performs as good as GTX 1080 which is horrible. FLOP already shows that AS Mac is slower than RTX 30 series which isn't 5nm. RTX 40 series are 5nm based btw.
Yes, Mac is great for video and music stuff but if they wanna expand to 3D field, then they really need to deal with 3D software developers to attract them.
Yes, Mac is great for video and music stuff but if they wanna expand to 3D field, then they really need to deal with 3D software developers to attract them.
I’d like to see how Apple makes an expandable Mac Pro? So far we are in the dark given the other AS macs using soldered RAM and SSD’s.
The M1 Ultra was released 18 months after its A14 generation-starter.
The M2 Ultra will be released roughly 18 months after its A15 generarion-starter.
Sounds like the time for smallest chip to Ultra scaling is ~18 months.
The Extreme is a step further and I would expect it to take more time anyway.
Even Intel would take like 2 years to go from the smaller chips to the biggest Xeons within a generation. Getting the M2 in July 2022 and the M2 Extreme in early 2023 is unrealistic. It would have been possible if we were still talking M1 Extreme (Jade 4C-Die).
The M2 Ultra will be released roughly 18 months after its A15 generarion-starter.
Sounds like the time for smallest chip to Ultra scaling is ~18 months.
The Extreme is a step further and I would expect it to take more time anyway.
Even Intel would take like 2 years to go from the smaller chips to the biggest Xeons within a generation. Getting the M2 in July 2022 and the M2 Extreme in early 2023 is unrealistic. It would have been possible if we were still talking M1 Extreme (Jade 4C-Die).
To be pedantic, it's not "on-die", it's "on-package".
<Raising Hand> Mac Pro user since the dawn of time, PowerMac before that, you can scroll backwards in time on my profile and nearly all messages revolve around Mac Pro.
Single question to anybody in this thread who actually owns or uses a Mac Pro: so... how's this different from always?
If Apple manage to release a new Mac Pro every half a decade, then for them it's an extremely short timeline, which could've lasted 7 years (it was a decade from Cheesegrater 5,1 to Cheesegrater 7,1; only 7 years if counting the trash can).
It is what it is, but it's not exactly shocking. Extremely limited target demographic whose sales generate profits in the realm of what amounts to a rounding error for Apple. Yes, flagship product, halo effect, blah blah blah, but at the end of the day ... focus resources on newer, new Mac Pro design, or... let's add some emoji's, litter more of the All Singing All Dancing crap of the world into macOS, and focus on the iPhone and services. Hmmm, tough choice!
Anyway... should obviate need for wheels, and replace with anti-gravity pads + Mr. Fusion for power needs.
seems like a BS rumor - they’re not going to release a Mac Pro that does not have a configuration faster than the Mac Studio.
The real issue is the yield rate for an "Extreme" SoC. If we presume that an "Extreme" SoC uses four M1 Max with UltraFusion connectors on all four sides, one would need all four dies to be perfect in order to make an Extreme SoC. Each 300mm wafer can hold under 150 M1 Max dies so than means 30 "Extreme" SoCs per wafer assuming 100% yields, which we know is not the case.
A perfect M1 Ultra is a $2200 upgrade over a perfect M1 Max. Some of that is no doubt "Apple Tax" and $400 is the mandatory 32GB to 64GB upgrade, but a significant part of that is probably to account for lower yields and by extension higher product costs. The "Extreme" will be twice as complicated so I can understand why Gurman believes a base Mac Pro with an "Extreme" SoC and 128GB of RAM could be $9999 (or more) compared to the $5799 of a Mac Studio with an M1 Ultra and 128GB of RAM.
At a minimum it appears to have PCIe cards (based on claimed first-hand experience with prototypes) and Apple could allow additional RAM via DIMM slots.
A Mac Pro with "only" a single Ultra class SoC but with PCIe slots and maybe significant memory expansion (via DIMMs) would still be desirable to a number of people who find the Mac Studio too limiting in terms of connectivity.
It’s now called the studio.