The problem is that they kind of forced it into being a niche machine. I've always wanted one, but they priced the regular consumer right out of it, and because they let the Mac Pro stagnate for so long they lost many of the professional customers to Windows or Linux. It all started with the Trash Can Mac Pro. They killed it themselves with their own dumb decisions.
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.
What's Happening With the Mac Pro?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Leaving the price the same is one thing when there are yearly releases. But when you let a product stagnate that long, lower the damn price, update it, or discontinue it.
Man, I've said this before too. That would help justify the expensive up front cost too. If I could buy the chassis/mobo once and then swap/upgrade the internals as time went on would be awesome. They could be updating the chassis or parts of the chassis somewhat regularly too, but having some kind of standardized upgradable platform that is supported for x amount of years would be amazing.
Only the PRO market segment value this and the loyal PRO-sumner - I posted same outlook here a few posts back
Last edited:
infiniband as been mentioned a lot in this topic. Doing quick look up online to see if Apple used it or not, I came across this PDF on clustering macs using TB5, it might be of interest to someone.
[PDF] Towards Large-scale Training on Apple Silicon
[PDF] Towards Large-scale Training on Apple Silicon
Apple might walk out with a Ai Mac Pro
Then suddenly Apple is in teh Ai market where it can be stronger. It's still not a chip company right. It has just produced a PRO Ai machine product.
Not sure if I came across this link on this topic but had it open in a tab - https://www.epfl.ch/labs/esl/do-we-really-need-big-data-centers-for-ai/
So considering they have pushed on-device processing for privacy reasons even before LLM mania hit the world, why not tap further into this market. The more dev they have for it on their platform the more Apple Ai can branch and grow and relying on Google ends at some point down the line.
Apple Ai Silicon for their Ai product range.
You buy you base Mac Pro blade system, and load it up with as many Apple Ai Blade modules and you're Go!
Then suddenly Apple is in teh Ai market where it can be stronger. It's still not a chip company right. It has just produced a PRO Ai machine product.
Not sure if I came across this link on this topic but had it open in a tab - https://www.epfl.ch/labs/esl/do-we-really-need-big-data-centers-for-ai/
So considering they have pushed on-device processing for privacy reasons even before LLM mania hit the world, why not tap further into this market. The more dev they have for it on their platform the more Apple Ai can branch and grow and relying on Google ends at some point down the line.
Apple Ai Silicon for their Ai product range.
You buy you base Mac Pro blade system, and load it up with as many Apple Ai Blade modules and you're Go!
Cost to manufacture a bespoke item for limited market.
Instead of building an entirely separate high cost device for a small number of people, just link up smaller mainstream boxes that the customer can incrementally add to without opening the box?
I can see the appeal in theory, but how well that would work in practice is another matter.
I used to assemble my own PCs so the result was, in theory, completely modular and upgradeable. Reality was, that only made sense if you got something wrong with the build and had to upgrade immediately. Otherwise, after a year or two, everything (RAM, CPU, GPU, HD interfaces, USB etc.) had all moved on, you needed a whole new motherboard and all that was re-usable was the case and (maybe) PSU - vs. keeping the old PC in one piece as a server, backup machine or hand-me-down and starting from scratch. Intel & AMD didn't help by continually changing processor socket - but that was only partly forced obsolescence as technical specs advanced as well.
...and you have to bear in mind that if you're upgrading your last-year's model CPU to the new shiny, so is every man and his dog (including anybody who is replacing a failed part) so the resale market for your surplus old-model CPU as a spare/upgrade part is going to be nonexistent.
In the case of the Mac Pro, Intel->Apple Silicon would obviously be a whole new "innards" - even the power supply would be over-specced and inefficient for an AS machine - versus finding someone with a use for a fully working 2019 Mac Pro (still a pretty capable machine!)
For the Apple Silicon Mac Pro, if a M5 Ultra does appear one of the big plusses is likely to be PCIe V5 support - so all the PCIe switch stuff will have to be replaced (OK it might still work as PCIe4 but on a system who's raison d'etre is that TB doesn't provide enough PCIe bandwidth for some, that's not great), so again it's going to be a whole new motherboard. Again, versus finding a use/buyer for a fully working M2 Ultra Mac Pro which is still a pretty powerful rig.
For something like a Studio, the logic board is so small that a socketed processor wouldn't make sense (and would force Apple to 'freeze' the physical design of the chip package) but being able to swap out (say) a M1 Max logic board for a M4 Max board does sound good. However, that's most of the works of the machine so it's not gonna be cheap, and its likely to be easier to sell a fully working M1 Studio than a bare logic board (again, because nobody will be 'upgrading' to those).
...even from the greenwashing point of view you're sending a lot of old motherboards/cpus/gpus etc. (made of 90% nasty plastics and chemicals) to landfill and only "saving" the case and PSU (in Apples case, lots of recyclable aluminium and full of copper coils which stand a much better chance of getting recycled).
I think that's too extreme the other direction honestly. I get what you are saying. But I am typing this from a PC I built my self back in 2020. The case is the easy one in the PC world. I could still build a new computer in my case I bought back in 2012. I tend to build in each case I buy at least twice. The power supply has also mostly been the same compatibility too. So for those parts, they are the most likely to be possible to re-use and it still does take energy and materials to make them even if it is less than that of a motherboard.
On the other parts, The market for re-sale of older parts would be small at the start, but would likely expand as this went on. Lets say they made a Mac Pro like this since M1, I would have a tough time selling the old M1 chip when I upgrade to the M4 variant because anyone who bought one would have an M1 or newer already. But, What if I'm upgrading an M3 to an M5. Maybe someone on an older M1 system still would love to have a cheaper M3 to upgrade their M1 they are still using. I agree a lot of people will be jumping on the new new, but not everyone. People upgrade at different rates and with different goals/financial situations that can also change from original purchase to later upgrades.
For the motherboards. I'm currently using a socket AM4 with the AMD 5950X on my PC I built. That AM4 platform was around for a number of years, and it is possible I could have bought a 1800X on AM4 when they launched and then upgraded to my current 5950X later without upgrading the motherboard at all. There are always times that require change, but The AM5 platform now has 2 generations of CPU's on it, and will likely have at least a 3rd generation. The motherboard doesn't need to be upgraded every generation either if it's designed with some knowledge of the CPU and feature set roadmap which apple should have since they make it all. Even on the AMD side, AM4 uses DDR4 and PCI-E 4.0 while AM5 uses DDR5 and PCI-E 5.0. So the motherboard only really needs upgrading when you have generational changes in core interfaces. We jumped PCI-E 4 to 5 pretty quickly, we were on DDR4 and PCI-E 3.0 for a pretty long time.
Caps. plastic caps. Snap to the bottom.Could even double as feet. What's that.. like a couple of cents worth of plastic?
AMD is the only company truly interested in catering to this market.
I don’t think Apple, Nvidia, or even Intel have much interest in it long term—for them, it’s a thing of the past, and the remaining market isn’t big enough to try to compete with AMD. There is just too much power and efficiency to be saved/gained from packaging everything but internal storage together before it reaches consumers or businesses.
I think the only option for a modular Mac Pro is the one that already exists. The internal storage is upgradable and removable. The only problem might be the heat sink—if the changes to the logic board require a different heat sink, that adds to the cost, but not to the level of difficulty—the heat sink has to be set aside regardless. One of the hallmarks of the Mac Pro from the beginning has always been how easy it is to work inside of it. That’s still true for the 2023 edition—it’s arguably the easiest Mac Pro ever to handle (repair guide), precisely because it has so few parts.
The problem with wanting the Mac to be more like the PC is that it would mean that the Mac was more like the PC.
True, the ATX formats for motherboards have stood the test of time. Plus, of course, they put most of the external connections on a standard-sized panel at the back of the motherboard. Power supplies too, to some extent, as long as you plan ahead and go for overkill (in terms of power and number of aux connections) - since mostly power requirements have gone up (that may be reversing now). However, that kinda reflects the fact that there's a huge market for generic, fairly conservative PCs where "cutting edge" means leaving off the PS/2 and Centronics ports. Apple do like to be a bit more creative with their designs...
(That said, I think Apple did once offer a logic board upgrade for some PPC or 68k model back in the 90s)
Yes - possibly.
Of course, that would mean that the hypothetical M5 Ultra Mac Pro board would have to have the exact same form factor & backwards-compatible headers for ports etc. as the M1 which would be a huge constraint (by Apple standards) on the design of a new M5 Mac Pro.
It's a bit moot because, currently, there is no upgrade path for the 2023 Mac Pro. If a new "M5 Ultra" MP does appear in the same case then maybe they could offer a logic board upgrade. (Flap, oink!) but the MP is overdue for a re-design or retirement.
Well, there was, anyway.
Bizarrely, some (all?) PowerBook 500s were marketed with "can upgrade to PowerPC later". And then indeed, several upgrade kits came out, including one from Apple.
But those were days when offering an upgrade kit made more sense. They were also days when Apple made a lot of shoot-in-foot design decisions. Just sell a PowerBook with the 68040. Then, when PowerPC-based PowerBooks are actually ready (shocker: it got delayed), start selling those.
Is it? The design per se is flexible enough that it should last a while. There isn't anything wrong with that case; it's just quite overengineered and now, with the SoC inside, needlessly large.
Mac Pro is 100% dead. They've been slowly killing the desktop tower form factor since 2013 TrashPro.
Best case, they will add an M5/Pro/Max/Ultra to the current Mac Pro and then by 2030 or whatever, completely kill it.
There's a very niche market of pros who will need PCIe slots for professional cards + storage (NVME) + networking, which they don't really care about anymore, it must be the 1% of users. There aren't even enough PCIe lanes with the M series Macs to fully support multiple high-bandwidth cards anyway. A lot of these manufacturers already started offering thunderbolt versions + you can use an external Thunderbolt PCIe enclosure, which I think is kind of whack, but it is what it is.
Best case, they will add an M5/Pro/Max/Ultra to the current Mac Pro and then by 2030 or whatever, completely kill it.
There's a very niche market of pros who will need PCIe slots for professional cards + storage (NVME) + networking, which they don't really care about anymore, it must be the 1% of users. There aren't even enough PCIe lanes with the M series Macs to fully support multiple high-bandwidth cards anyway. A lot of these manufacturers already started offering thunderbolt versions + you can use an external Thunderbolt PCIe enclosure, which I think is kind of whack, but it is what it is.
Last edited:
...rapidly shrinking I suspect, but still bigger than the market for something like a Mac Pro.
Also, CPU power, RAM capacity etc. was increasing more rapidly then than today and hadn't passed the "good enough" threshold for things like audio, video and graphics, so a shiny new computer was very rapidly obsolete.
I'm guessing a lot of those 2023 Mac Pros are doing audio/video production workloads - which needed specialist PCIe interface cards, like the ProTools ones - which an M1 Ultra will still eat for breakfast.
If it's quite overengineered and now, with the SoC inside, needlessly large I'd say its ripe for a re-design.
The original Macintosh had two logic board OEM upgrades. The “Macintosh 512K Memory Expansion Kit” simply replaced the original 128K logic board with a new one with 512K. Later, the “Macintosh Plus 1 Mb Logic Board Kit” replaced the 128K or 512K logic board with a new one with 1 Mb, although to upgrade all the way to a Mac Plus you also needed to replace the disk drive, via the “Macintosh Plus Disk Drive Kit” — the Mac Plus was the first Macintosh logic board with socketed memory, so you could replace 4x 256K with 4x 1 Mb for a maximum of 4 Mb!
January 16 is the 40th anniversary of the launch of the Mac Plus. Just sayin’
Even though he was forced out before it launched, the Mac Plus was Jobs’ last project at Apple (before his return). Seems worth commemorating?!
The case was clearly designed to cool a massively hot Intel CPU and two massively hot AMD GPUs, but with Apple Silicon running as cold as liquid nitrogen in comparison and no more discrete GPUs, I agree they could probably shrink the chassis a bit so that it was tall and long enough to support full-length PCIe cards.
It runs cold because it does less and uses a smaller process. The whole point of the pro is doing more. Energy efficiency is amazing in laptops, but in a workstation power and bandwidth are far more important.
That case is exactly what a workstation needs to be, and it’s the best case ever designed for the purpose. If Apple made a chip to do it justice we’d all be celebrating it. Right now they don’t even seem to have plans for that.
They evidently took a stab at it with the prototype SoC using four M1 Max, but it was so complex and yields were so low the chip alone would have run in the five figures as a BTO option.
But a power-sucking SoC is counter to Apple's compute design philosophy. I'm kind of surprised we get the Ultra SoC, to be honest.
We are a theoretical physics research institution. First of all, the thermals/noise will always be better with a large-form factor computer (we are running the machines at 100% for months doing numerics). Also, it is handy to be able to put several internal SSDs directly in the machine (besides the PCIe route there are options to put 2x 3.5” SATA HDD, i.e. 4x 2.5” SATA SSD into the mac Pro internally — of course U.2, U.3 or at least SAS would be better than only SATA). Also, you can put 25 Gbps, 100 Gbps, 200 Gbps, or these days even 800 Gbps network cards into a chassi like the Mac Pro provided macOS supported those (10gbps is really not enough and Thunderbolt is also not fast enough — there is a reason why the new nvidia AI development systems have dual 200 gbps network cards built-in). And no, we have absolutely 0 interest in making a *desktop* small like the Mac Studio — why would anyone? It’s not a portable, it’s a desktop (yes, wheels are helpful for moving the chassi around a lab but there is no need for small form-factor like the Studio). The small size seems to be more of an Apple fetish than anything else.
P.S. The max RAM of the M2 Mac Pro (128 GB) is also a serious limitation — The good old cheese grater Macpro 5,1 supported 128GB of RAM, even the last xServe supported 96 GB (the max of 512 GB of the M3 Ultra Mac Studio is also not enough at all these days)…
Last edited:
Oh, it's gotta be less. Only about 10% are desktops, and that includes the iMac, Mac mini, and Mac Studio.
(This is also why I don't really understand the iMac Pro rumors. They're not gonna suddenly sell more Macs that way. They'll sell fewer Studios and minis, so they'd have to make it up in margin.)
Yeah.
Right.
(Clearly some firewall between the Mac Pro team and the M1 team. The 2019 Mac Pro design makes some strange choices when the ARM transition is announced mere months later.)
That doesn't make sense to me. Surely separate CPU, GPU, NPU, with socketed RAM, are easier to engineer than placing all that on a single package.
And Apple has engineered chipsets before. See PowerPC Macs. Apple logic board + chipset; IBM or Motorola CPU.
The engineers who worked on PowerPC are long retired.
I wasn’t commenting on how hard it is, just that they don’t have any expertise (or apparent desire) to do it. Even their messaging is heavily SOC focussed. How could they row back from memory and storage speed and latency figures that aren’t possible with bus architectures? Every number where Apple beats competitors is due to SOC architecture and a modular design would almost certainly see them beaten so where’s the motivation to create a more expensive system for a small niche where Apple will be the underdog anyway?
I wasn’t commenting on how hard it is, just that they don’t have any expertise (or apparent desire) to do it. Even their messaging is heavily SOC focussed. How could they row back from memory and storage speed and latency figures that aren’t possible with bus architectures? Every number where Apple beats competitors is due to SOC architecture and a modular design would almost certainly see them beaten so where’s the motivation to create a more expensive system for a small niche where Apple will be the underdog anyway?
I don't think it has anything to do with expertise because the 2019 MP is an engineering marvel and that's a "recent" machine. It just has to do with the bottom dollar, a business decision to not care about the MP.