Apple could design the Ultra as a true "workstation class" SoC like the Intel Xeon or AMD Epyc so that it would scale more linearly in a dual-SoC configuration than the current Ultra, which is two "consumer" SoCs lashed together.
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.
Apple's 2025 M4 Mac Pro: What to Expect
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple could design the Ultra as a true "workstation class" SoC like the Intel Xeon or AMD Epyc so that it would scale more linearly in a dual-SoC configuration than the current Ultra, which is two "consumer" SoCs lashed together.
Spot on. The MacPro will continue to do well for those who need PCIe slots and a robust power supply in a variety of scientific, industrial, defense, professional music creation/production, etc disciplines. Especially in its rack mount configuration.
Not a good comparison. That $52k price for a 2019 Mac Pro maxed out everything including things simply not available on the 2023 - including 1.5TB RAM costing $25k (2023 tops out at 192GB) and $9600 worth of GPUs. I'm not mentioning Afterburner cards since the M2 Ultra's media engines make those obsolete. The M2 Ultra probably does thrash the Xeon W 3275M processor on CPU power, but that "M" suffix doubled the price of the processor (ISTR the processor alone was about $7k) purely for the sake of supporting more than 1TB of RAM.
The $7000 2023 Mac Pro is "just" a $4000 Studio Ultra in a big, grotesquely over-engineered box with some (empty) PCIe slots - that can't take GPUs - driven by the surplus SSD controller on the second M2 Ultra die. If you need fast internal PCIe SSDs, specialist networking or AV production cards then those slots offer far more bandwidth than an external Thunderbolt enclosure hung off a Studio.
If you don't use non-GPU PCIe cards, you don't need the 2023 Mac Pro - its partly expensive because its niche (apart from the $400 wheels, which are just Apple having a laugh).
That’s a terrible comparison
A maxed out Intel Mac Pro was a significantly more powerful, useful and upgradable computer than any other Mac available at the time
Apple silicon Mac Pro is just a studio with pci in a big box for more money
Last edited:
Mac Pro should:
- Support GPU cards
- More than 2 SSD slots, at least 4 but ideally even more
So far, the Apple Silicon concept has hinged on having the GPU and specialist "engines" on chip sharing unified RAM. It does best in Apple-Silicon-optimised applications that benefit more from the GPU having direct, high-bandwidth access to tens of gigabytes of fast RAM than a faster GPU or more non-Unified RAM. U-turning on GPUs would throw away the Unified RAM advantage - and would end up with much the same performance as PCs using the same AMD GPUs (and many of the people asking for GPUs really want NVIDIA which probably ain't gonna happen for political reasons).
If you want a big box'o'slots filled with AMD/NVIDIA GPUs for standard CUDA/OpenCL/OpenGL/whatever workflows then frankly the M-series isn't the tool for the job c.f. a Xeon or Threadripper box - and I don't think the powerful personal workstation is a growth area that would justify Apple developing a more conventional processor. It's worth looking at what NVIDIA is doing for high-end AI applications - which has taken a somewhat Apple Silicon-like direction - and there are rumors that Apple are looking at a high end AI-focussed chip.
As for SSDs - if you have as many PCIe slots as the 2023 Mac Pro, and no GPUs to fill them with, you can add PCIe-to-M.2. SSD cards until the cows come home.
There are a LOT of industrial applications.
Special I/O cards for controlling manufacturing equipment or reading data from same.
When you look at things like scanning electron microscopes, you see big equipment. When you open them up, there are several computers inside, often with multiple components that control different parts of the tool.
Very rarely do you see any of this controlled by USB except the keyboard and mouse. When you need 3nm accuracy, that USB-C doesn't look nearly as sexy as PCIE connections. When ultra precise control of time, again PCIE wins. If you need to control more than 255 parts, PCIE to the rescue. This may seem like a niche to you, but I guarantee you the money is good.
Show up at a vendor with $9 million dollars to ask for something 'pretty good' (not the latest greatest), and, because you are a loyal customer of a major international company, they will deliver it in 'only' 4 years - and that if you put it under contract for 3 years at additional 7 figure costs annually, and you walk away feeling like you just made a steal compared to industry standard.
People who need THAT need PCIE.
I have tools at work with racks of 18 Mac Pros in them. And we weren't the only ones that bought those.
GPU cards are always possible, and the hardware probably supports them, but I doubt we'll get OS support any time soon.
As for SSD slots... If you're buying an $8000 computer, springing $30 for:
OWC Accelsior 1M2 NVMe M.2 SSD to PCIe 4.0 Adapter... at MacSales.com
SACL1M OWC Accelsior 1M2 NVMe M.2 SSD to PCIe 4.0 Adapter Card. Add your own NVMe M.2 SSD to create a speed boosting, capacity expanding solution.
... seems pretty trivial.
I agree that Apple could have put some M.2 slots in there, but it is super easy to mitigate.
Last edited:
It probably won't come down more than this, don't get your hopes up.It’s incredible how much Apple was able to lower the price of the M series Mac Pro from what a maxed out version of the Intel Mac Pro used to cost:
A Maxed Out Mac Pro Will Cost Over $52,000
With Mac Pro orders now live, pricing for all of the build to order options for the machine are available, giving us an idea of what it's going...www.macrumors.com
Hopefully the price can continue to come down with the second generation M series Mac Pro with all of the efficiencies that Apple will continue to improve upon in their manufacturing capabilities
If anything, they should develop a better chip than Ultra and jack the price a little more.
But it's crazy to see how a maxed out Mac Pro 4X less expensive than ever before makes absolutely no sense anyways. The Mac Studio is $3K less if you max it out, this makes much more sense. Oh and you can even save an additionnal $1K by buying a Satechi hub and sticking an 8 TB SSD in it for half of Apple's price if you really need it.
Last edited:
Yes to all of the above. And with the MacPro rack-mount version.
After over a decade with Mac Minis I have zero desire to have a computer tower anymore, at all.
The last supported amd gpu’s still outperform apple silicon in metal
So it’s not just about cuda/opencl/whatever
What I would like to see in the next Mac Pro:
- Monolithic M4 Ultra - Up to 480GB LPDDR5X (inline ECC)
- Two-die M4 Extreme - Up to 960GB LPDDR5X (inline ECC)
- Four NAND blades (end-user configurable w/third-party options)
- Thunderbolt 5
- ASi Compute cards (target-able compute/render nodes)
- M4 Extreme Mac Pro Cube
I don't see that the M series mac pros will ever support 3rd party GPUs
It’s weird that Apple struggles so much to deliver such an obvious product. It’s like they just refuse to give people what they actually want and always has to do something weird or different that compromises the whole thing
In the case of the recent one, it’s not properly expandable. People were saying that it didn’t provide enough RAM for some advanced research and stuff like that
In the case of the recent one, it’s not properly expandable. People were saying that it didn’t provide enough RAM for some advanced research and stuff like that
Indeed, which is why a number of people were asking Apple to just release a new Intel-based Mac Pro. The OS already runs on Intel, the apps run on Intel, but Apple's bent on trying to shoehorn the "consumer" M chips in there instead.
Yeah but the problem with that is Apple hasn’t been able to deliver the necessary GPU power or high amount of RAM some people need
Apple silicon is great for most applications, but they could be permanently damaging their relationship with this market the same way they did with Final Cut Pro X, and that’s not easy to fix
What to expect? Complaints like always.
HobeSoundDarryl
macrumors G5
Ummm, there is ANOTHER option and that is BOTH. No U-Turn necessary. Leave Silicon Graphics as is and also support optional graphics cards in slots. Those who believe there's something magical about Apple graphics don't lose a thing... including the ability to keep using Silicon graphics exactly as they would now. Those who believe they need some third party graphics could add those too.
Intel MBpros used to have both options: intel AND AMD graphics and a user could even select which they wanted to use for some task. This could be the same.
Slot-based computers sometimes will fill those slots with multiple graphics cards, just like they may fill them with multiple audio cards, multiple memory cards, multiple storage cards, etc. A card is just a card. If it can be slotted, don't rule it out to protect some kind of "ours and only ours" turf. Else, customers of pro towers SHOULD opt for PCs who are not deciding such things for their customers.
The change that Apple would need to roll with is giving graphics cards makers a way to have Silicon drivers for their cards and/or build some such support for that themselves. But that's no great leap- just software.
Last edited:
Has IRIX been ported to ARM64 yet? That would certainly be interesting.
A question for the “in the know” users on this thread—for applications demanding $X0,000 in specialty hardware, is it still standard practice to run all of that on a standard off-the-shelf operating system? I know a depressing number of scientific instruments just glue a bunch of third-party controllers together with some amateurish Windows app, thus creating insultingly high prices for janky user experiences, but surely the more software-savvy fields such as scientific computing research would be customizing Arch Linux before they’d even bought their third CUDA card?
…actually, if this isn’t the case, I’m not sure I want to know
…actually, if this isn’t the case, I’m not sure I want to know
There are investment and architectural tradeoffs that Apple is making.
Intel, AMD, and Nvidia can spend a lot of resources on high-performance computing, because they all have huge markets to sell in to. Apple's market size is so small, it must be quite an internal challenge to argue for burning a lot of money with limited RoI.
Had Apple gotten the M1 or M2 Quadra (4x M* Max) working, maybe most of this wouldn't have been as much of a problem.