Gaming rigs (intels 13000, 4090) can be a cost efficient solution for work but it does not make it a workstation (xeon, ECC ram, high RAM count gpu). The current Mac Pro looks like video/audio production platform and fits neither of the mentioned categories.
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.
Mac Pro 2023 = Another Mac Pro 2013 crisis and failure.
- Thread starter sunny5
- Start date
- Sort by reaction score
Depends from which perspective you look at it. CPUs with long vector and dedicated matrix hardware are exclusively workstation parts, so are GPUs with 40GB+ of RAM. Lack of ECC is certainly a big minus, I/O on the other hand is sufficient given the fact that GPUs (biggest PCIe bandwidth consumers by far) are not supported.
All in all, the Mac Pro right now is a kind of a hybrid, which is also cheaper than a comparable classical workstation.
II’d say it’s also a decent platform for certain types of scientific computing (with the obvious minus being lack of ECC).
Computers like that run the world in data centers. Every time you use the cloud, you are using hardware like that. The average cost-effective computer has 64-128 cores and 4-16 GB memory per core, depending on the application. Buying the same overall performance in smaller or larger units would be more expensive.
I swear every thread involving the Mac Pro devolves into “I want a Windows PC”
Anyway I had a few relevant thoughts:
1. The 2019 apology Mac Pro was apparently made with feedback to Apple on what their Mac Pro customers wanted. I think it’s safe to assume the Apple Silicon version takes those issues into account.
2. Apple has never done well in the enterprise market, which a lot of the criticism seems to stem from. (Comparisons between workstation brands, etc.)
As someone who works in a lab now, I can guarantee that Apple isn’t gonna take over anytime soon.
3. Apple’s target market seems to be “prosumers”, people who work independently on their own projects, or in a small business rather than large enterprise.
In following that train of logic, the new MP seems fine. It seems like people were expecting some insane machine that outruns threadripper 4090 builds at $6000 max with everything socketed.
Anyway I had a few relevant thoughts:
1. The 2019 apology Mac Pro was apparently made with feedback to Apple on what their Mac Pro customers wanted. I think it’s safe to assume the Apple Silicon version takes those issues into account.
2. Apple has never done well in the enterprise market, which a lot of the criticism seems to stem from. (Comparisons between workstation brands, etc.)
As someone who works in a lab now, I can guarantee that Apple isn’t gonna take over anytime soon.
3. Apple’s target market seems to be “prosumers”, people who work independently on their own projects, or in a small business rather than large enterprise.
In following that train of logic, the new MP seems fine. It seems like people were expecting some insane machine that outruns threadripper 4090 builds at $6000 max with everything socketed.
What exactly is that you do on your workstation(s)? Multimedia, such as NLE, DAW, 3D and VFX comping on the same machine? Or you are strictly into audio or visual field? Unfortunately I cannot help you with machine learning and physics simulations.
lol, tell that to all pro users with high end specs Mac Pro for several decades. Clearly, you are defending Apple's degeneration. Even video markets for Mac is quite doomed. I gotta laugh.
Yep. I was running into GPU memory issues with my 4090. I could have fixed it with a $5,000 Quadro card. But at that price I’ll just get the Mac Studio with max memory for a bit more.
Wouldn’t make sense from a business perspective. I’m sure Apple’s partnership with Intel is some form of contract and costs some amount of money. Also, that would delay Rosetta and dropping Intel builds from macos.
I still have apps that are Intel only and I’m not happy installing Rosetta on my new devices still.
avkills
macrumors 65816
Apple is going to have Intel binaries floating around for some time; so it really isn't that big of a deal; besides it is a known fact that Apple has Mac OS compiled for multiple architectures constantly. The entire OS was designed so you could have multiple binaries; been like that since they bought NeXT.
Maybe as a short term business plan. I have a feeling this move will lose a lot of pro users forever. About the only users that have a real incentive to stay is those who use FCP and associated apps as their primary editing system.
I am holding out because maybe Apple has a big surprise with the M3 kits. But the PCIe is severely hampered on the M2 systems based on the reports I have read.
I think so too M3 is where it will be. Unfortunate real-world issues made plans a mess.
Whole purpose of PCI is to replace and upgrade GPUs. Period. Everything else is super tertiary. Do you know when I upgraded my MADI or AJA? Never. After GPUs the second most used PCI application is SSD upgrade which would benefit from PCI5 whole a lot especially if you are into video editing. Speeds between PCI4 and PCI5 almost double when working with prores/DNxHR files.
And you would be surprised how many people actually have a need for multiple GPUs or latest and greatest GPU today. Many workflows in visual creation have switched from CPU to GPU and if you are in that field you can pretty much get the job done with Intel gaming CPU and latest and greatest GPU on board. And guess what? Nuke still prefers clock speed of single thread instead of slow multi-cores and it relies on GPU more and more with each release. So today even serious compositing in both Nuke and Fusion can be done on Intel gaming CPU. The bottle neck would be RAM memory cause Fusion maxes out 128GB with an ease but then SSD comes into play. Yup that same SSD you shoved into your PCI slots.
A bit of a tangent, but was having a look at GPUs with a large amounts of memory and came across a comparison of the Ada RTX 6000 vs a 4090. The 4090 is 75% cheaper, runs faster than the RTX 6000 in most tests. What you get for your money is 50% more memory and studio drivers. Always though Nvidia Quadro cards were overpriced, but damn...
Kinda makes Apples high memory prices and the new Mac Pro seem reasonable 😀
More generally, if Apple adds decent raytracing cores to the M3 then that's pretty compelling use case for 3D rendering. As an artist workstation, the Studio is already pretty great; solid cpu cores with plenty of memory bandwidth for simulations, fast SSD for interactivity and the GPU is plenty strong enough for most viewport rendering.
TBH that's where I think Apple broadly sees the Mac Studio and Pro, at least as far as DCC goes; as an artist workstation. Competing with 64 core thread rippers with TB's of memory doesn't seem like something Apple would be overly interested in (if you're building a render farm, linux boxes are always going to be the obvious choice, and the 1.5Tb of memory on the 2019 Mac Pro was probably more of a historical anomaly due to Intels roadmap than anything else).
As it is, for they've done a pretty good job at providing a workstation for image and video editing and solid start for 3D.
There is a huge difference between keeping binaries going in the lab for a let’s see what can do if we want to swap architecture, and provoiding commercial support for those binaries.
PCIe primary purpose was to provide a standardised expansion bus/slot Type suitable for everything.
by the time of PCIe then thankfully the various ISA, EISA, VL-bus systems, and MCA (Reiner those) had gone away and left with PCI, PCI-X and AGP being around. AGP being developed as PCI starting to show limits for cards.
so would end up with high end boards where has AGP slot for Graphics card. PCI-X for 10gb NIC/Storage Controllees, PCI for sound cards, 10/100/1000 NIC, etc.
PCIe replacing all 3 slot types allows greater flexibilty in that only one expansion bus to cater for and the PCIe lanes could be divided up how the system board vendor wanted too Based on the number of lanes the controller supported.
As computers got more integrated and the built in on motherboard parts got better then less expansion cards needed for the majority of people And so ended up where the majority of PCIe cards installed by users into desktops are GPU cards but that is more the fact that people don’t have too buy there Network Cards and Sound Card seperately these days for the majority of people. Wi-Fi and BT again pretty much onboard and not expansion slot. 4 port NVME storage cards again for the majority of PC’s are not being installed.
can even find 10gbe built into higher end boards theses days rather then requiring to go out and buy a 10gb NIC card.
M2 Ultra will be competitive up to Sapphire Rapids w7-3465X, which is more than ok for the price they are asking. It’s a case of not excellent, not terrible. The M2 Ultra is certainly disappointing for someone looking at extreme workstation performanc, but will still work well enough for most users interested in that kind of machine.
As to staying in Intel, I understand your point, but ultimately, Apples decision made sense from the business standpoint. Releasing a new Intel machine means delaying the point when they can deprecate x86 altogether, and they have to do it ASAP to ensure software quality and stability going forward. Another option would have been to wait, but that would probably have caused more PR harm. So they went with a compromise.
In the end, only time will tell what their plans are for high-end desktop market. It is very clear that M1/M2 are mobile-first products that are unable to utilize the full range of the desktop thermal bracket. This could be indicative of the long-term strategy, or this could be merely the side effect of gen-one product. I think there are certain signs that this is not the end of the strategy and that future iterations of Apple Silicon might scale better for desktop, but of course, that’s just speculation.
Yup. All true you wrote there. Who can not remember good ol days of AGP video cards such as Number Nine or Matrox. We had bunch of them.
Anyway, those who stayed with Mac after OG Mac Pro cheese grater have probably wondered into PICe-less world of external devices and probably have bunch of them right now. So what is their incentive to go back to PICe with the new M2 cheese grater? Thrash can Mac Pro did solid damage to PCIe stuff on Mac. Thank God it was not that great of a seller but I do know some people who sold off dozens of OG cheese graters and all of their PCIe component in order to buy the trash can and external replacements (they even sold me PICe cards in order to complete the transition). And even some of them upgraded to new Intel cheese grater but they never returned to PCIe cause they already had external stuff they needed.
So I would assume M2 cheese grater is for the first time Mac Pro buyers or an upgrade to those who were the first time 2nd gen Intel cheese grater Mac Pro buyers. So far I don't know any of my colegues who have 2nd gen Intel cheese grater to even think about upgrading to M2 Mac Pro or Studio. There are numerous reasons why they will not upgrade anytime soon and one of them is software and hardware incompatability in audio production world. And even if they do they are already stuck with external devices from the trash can era.
I left Mac world cause I needed PCIe, I needed top of the line GPU, I needed faster lanes because of Fusion IoDrive (well that was long time ago) and in the end we wanted to keep all of our PCIe cards cause they just worked. My RME HDSPe MADI is now ten years old and it aint going nowhere anytime soon. Now those early adopters of trash can had to buy RME external devices which at the time came with lousy firewire or USB 2.0. It was not until 2016 that Thunderbolt became viable option. Then they had to buy UAD Satelite (also firwire) while I still had and still have to this day UAD PCIe in two of my audio workstations and one in my personal desktop tower. Even though UAD accelerators are now pretty much obsolete no one will go and sell their external Satelite jus so they can buy UAD2 PCIe and jam it into new M2 Mac Pro. I bet number of people doing that would be in three figures globaly.
Why would they need to sell their equipment? They can use it just fine with the external I/O of the new MP. I think you are describing a false dichotomy here. Just because the new tower has PCI-e slots doesn't mean that your old equipment becomes unusable (this was very different with the 2013 model). It just gives users more options going forward.
Well that's exactly my conclusion. They will just continue to use external hardware and there is zero incentive to upgrade to M2 Mac Pro. Maybe my rant was cluttered but bottom line is M2 Mac Pro as an upgrade makes little sense for both current M1/2 users and Intel Mac Pro users.
I can imagine that performance is a good enough incentive for many users. With many caveats, of course (like the fact that if you really need performance, there are better options on the market). But if you are on a 2013 cylinder and want to keep using a Mac, either the Studio or the new MP would be terrific upgrades. And the MP also gives you the option of internal expansion, which might be of interest to some people with aging equipment.
JoeG4
macrumors 68040
Am I disappointed? Yea of course. $7k? Why the **** is it so expensive? Why can't they make a "for the rest of us morons that insist on a tower with easily up-gradable storage" model for like $2500 or $3k or something.
(sarcasm) This thing is just for porn editors to flex on their video editor friends that get stuck using windows machines. lol.
(sarcasm) This thing is just for porn editors to flex on their video editor friends that get stuck using windows machines. lol.
I swear every thread involving the Mac Pro devolves into video editors acting like they're god's gift to mankind and Apple should only cater to them.
