I do agree that the Studio's performance (certainly in CPU) is very impressive for its size / power. This is more a welcome side effect of the chips' mobile origins (a desktop doesn't need to be that small), but it is a benefit nonetheless.
Of course. Though, Apple is more concerned with besting the prior models than they are with besting the competition. Hence, the asininely vague graphs and comparisons to "the latest/most popular PC chip". If there was a superstar x86-64 CPU that Apple could directly say "look, this outperforms this CPU in every possible way" they'd probably have done so. Instead, it's shrouded in kludgy marketing.
Obviously, you can take an M1 MacBook Air, compare it to the 2020 10th Gen Intel Y-series based MacBook Air and say "look, this thing is universally faster in every possible way." But, relative to how crappy those specific processors were, that's really not saying much.
Similarly, you take a 2020 27-inch iMac and compare it to an M1 Max Mac Studio, it's going to be sizably faster; and that's the only comparison that matters to Apple's marketing (despite the fact that a contemporary Intel processor designed for the desktop form factor that the 10th Generation Intel processors in that iMac were designed for might be more closely matched to the M1 Max).
Apple's greatest gains with Apple Silicon are always going to be in notebooks. There's no debate there. I concede that the gains won't be as great with desktops, but I'd imagine Apple will still make Mac Studio models that iterate on performance over their predecessors.
I don't think the generational improvement is quite that impressive, aside perhaps for very specific tasks (e.g. number of simultaneous 8K video streams that can be compressed). From a quick look, 25% seemed in the ballpark for average performance gains.
There are benchmarks that will show impressive gains. Just as I'm sure that there will be some that don't. Again, Apple is optimizing the performance of these SoCs to specific workloads. It stands to reason that some will be markedly more powerful while others will not.
ASi CPU performance is certainly impressive. The big issue is with GPU though.
They definitely have more room to grow there. But, it also seems as though Apple is trying to compensate for that with custom accelerators that augment specific tasks such that the GPUs themselves do not need to work anywhere near as hard on their own. I'm not saying that this negates the need, or even the benefit of a more powerful GPU. Just that it's probably the case that the workloads Apple is targeting in their optimizations are not as affected by their GPUs not being as powerful as contemporary ones made by the competition (which, inherently, have to do much more heavy lifting to accomplish similar results).
The 2019 MP may be stuck with last-gen AMD cards, but they are beefy and it can fit a lot of them in there.
I think that depends on your workflow. I know that, when Apple was re-evaluating their mistakes with the 2013 Mac Pro, one of the things they realized is that most of their customers actually didn't need multiple GPUs. They accommodated for it beautifully as an option with the 2019 model, but never forced it upon anyone. I'm just guessing, but I think that much of the results of their surveying the damage from the 2013 Mac Pro (and assessment of what was bare minimum for a viable replacement) informed the minimum bar of entry for a replacement. Incidentally, M2 Ultra IS two M2 Max chips fused onto a single package.
Not saying that a 2019 Mac Pro isn't potentially benefited by TWO Duo cards; just that the likelihood is that the number of people who saw serious benefit from that kind of configuration was REALLY small and better served by Windows workstations than Apple would ever want to do with a Mac Pro.
And of course, PCs have current-gen Nvidia, which will shred an Ultra in this regard.
Well, yeah. That's pretty much a given, at this point.
Good point. RAM capacity may have been as much of a bar as anything else. I agree that 192GB likely caters for the vast majority of customers. Even if unified RAM works differently etc., it would have been hard to get customers to (spend a lot of money) upgrading to a machine with less RAM.
I totally agree that 192GB as a maximum RAM capacity (compared to 768GB for 8-16 Core 2019 Mac Pros or 1.5TB for 24 or 28 Core 2019 Mac Pros) is bad optics. Then again, spending even 192GB of RAM in a 2019 Mac Pro would've been extremely expensive and not something a user or business would've done lightly. We also got those RAM capacities due to the Xeons supporting them more than because Apple thought it was something users needed. Again, Apple's whitepaper on the 2019 Mac Pro strongly suggests that they fully believed that 192GB was the most that would be needed on the highest end of workflows.
An Extreme would be a logical extension of the current paradigm. It will be interesting to see if the M3 gets one. Perhaps it will be more akin to a dual-socket design, with 2 Ultras, rather than a true 4-die chip linked with UltraFusion (which would seem to require a separate Max variant with additional UF links - and is there even space on its sides?).
UltraFusion would seem to be vastly preferable from the standpoint of latency between SoC components. Then again, I'm not exactly sure why they're not offering the lesser M2 SoCs on compute cards. Fine that they don't communicate as fast as if they were the same organism; why wouldn't an M2 Ultra Mac Pro with 192GB of RAM and 76 GPU cores NOT benefit from an M2 Pro with the full 12-core CPU and 19-core GPU with 32GB of RAM on a compute card doing its own tasks on the side?
At the very least though, it seems that Apple went full-steam ahead with the transition without knowing if an Extreme was really feasible. I think they were only willing to take this risk because at the end of the day, the MP isn't that important to them; if they had known for certain that it would never be possible, I doubt it would have affected their plans at all. The transition had too many other upsides to them.
Nah. They saw the 24-core CPU, the 60-76 core GPU with 7 Afterburner cards' worth of accelerator performance, and the 64-192GB of RAM and decided that it was enough to suffice for the vast majority of the customers that would need this class of machine. It's not that the Mac Pro isn't important to them. It's that, TO THEM, the Mac Pro does not need to be anything more than an Ultra variant of a Mac Studio with I/O that would only be possible in the configuration of a tower. If Apple did the research and found that most of the people needing to buy a Mac Pro really only needed what this 2023 Mac Pro has to offer, then it was merely a calculated risk as to just how many customers they'd alienate by nixing the added upgradability from the 2019 model.
Though, I don't agree with it, I'm guessing that, unlike in 2013, they actually calculated intelligently this time.
You're assuming people are buying it.
But if there are a handful, yes, they're crazy.
It sounds like there ARE people buying it. Just not anyone that wouldn't otherwise be just fine on an M2 Ultra model of Mac Studio. But, this Mac Pro was never meant for those people; rendering that point utterly moot. If you need to slap in a $7000 broadcast/capture card into a Mac, this is really the only option. If you don't have any need for PCIe cards, what are you even doing talking about this machine?
Not true. Independent pros will save up their pennies for a new monster machine --> if it's upgradable.
Upgradeability adds value. I won't deny this. However, upgradability is far from the only value to a Mac Pro and it completely pales in comparison to the value added by I/O. Yes, you have to pay Apple's RAM tax to buy more RAM up front. That part sucks. But that really won't stop this machine from being sold to anyone for whom that isn't an absolute deal-breaker.
It must be machine with longevity. This one is not.
Considering that (a) the 2010 and 2013 Mac Pros are tied at 8 years for longest-supported Mac Pro models, (b) it is extremely unlikely that the 2019 Mac Pro will have that much time being supported, and (c) that we have absolutely no metric for how long Apple Silicon Macs will last before not being able to run a macOS version, I'd say your claim has issues.
Furthermore, the only way in which this machine lacks longevity is with GPU and RAM upgrades (which never really did all that much to extend the overall life of even 2010 and 2012 Mac Pros relative to the rest of the system to begin with). Yes, you could squeeze a bit more performance out of a 2010 or 2012 Mac Pro with those upgrades, but again, you're still chipset limited. The best that 580X card in a 2012 Mac Pro would do is give you the illusion that it's still worth holding on to that 2012 Mac Pro (when it's actually a much better idea to move on).
No sane indie pro will pay premium prices for disposable computers. Fan boys won't stay in business.
Apple has made killings off of indie pros paying premium prices for disposable computers for decades. Indie pros didn't balk when Apple started soldering the SSDs into 15-inch MacBook Pros. Indie pros have never balked at the fact that the drives are not user-replaceable in a 27-inch iMac. And no one HAS EVER balked at the fact that ANY Mac Pro or Power Mac HAS EVER allowed you to swap out one generation of system board (and therefore processor and RAM type) for another which, if we're REALLY being honest here, is what truly makes ANY Mac ultimately disposable.
The Mac Pro was literally a Xeon class workstation that could run Windows as a secondary operating system.
Until Apple activates the TPM 2.0 latent on all 2018 and later Macs (give or take adding Secure Boot functionality to the 2019 iMacs) and releases Windows 11 compatible Boot Camp drivers, it's limited to Windows 10, which is only supported for another two more years. Compare that to any other 2019 era Windows workstation which can absolutely run Windows 11 with no question.
My point is that if we're talking about Xeon workstations that can run Windows in a supported configuration, a Mac Pro is subpar compared to its contemporary Xeon-based alternatives.
It was, tier for tier, component to component an equal to a PC workstation.
This is debatable. First off, there are TONS of different sizes, shapes, and feature sets across the myriad of Xeon based workstations you can buy. At the time, HP had the Z4, Z6, and Z8 workstations that all offered very different degrees of performance. Dell has a similar array of Precision towers. Other than maximum RAM capacity differences between the 16 (and fewer) core models and the 24 (or 28) core models, it was the same Mac Pro with the same features. Incidentally, I'm pretty sure that the mid-higher tier Z and Precisions had more capabilities as Windows boxes than the 2019 Mac Pro had as a Mac, let alone as a PC.
Secondly, the Xeon was one used in PCs. The UEFI and some of the other common system board components were probably similar. But, that Mac Pro had a lot of proprietary Apple nonsense in them. The T2 chip and its proprietary SSDs are a fantastic example of this.
And if you really weren't convinced you could load Windows onto the exact same hardware and run it as a Windows workstation.
Again, up to Windows 10 which only has two more years of supported life left in it and with no supported way of running Windows 11.
Again, we can quibble about Radeon vs GeForce but other than that it used _the exact same parts as a PC workstation._
The guy you are replying to is right, though. The Mac Pro was never the best Xeon workstation, let alone for its price. Lack of NVIDIA cards is a big reason, but far from the only one.
The GPU upgrade options in the Power Mac era were actually pretty good. Even compared to the 2019 Mac Pro.
They paled in comparison to the 2008-2012 Mac Pro era. They were only good compared to the 2019 Mac Pro era in that, by that point, you were pretty much limited to AMD cards, the best of which used a proprietary slot that was only good on the 2019 Mac Pro (and useless literally everywhere else).
The Power Mac G5 supported ECC memory. It was added specifically because major customers asked for it.
So I'm not sure what you're talking about.
Only the final (Late 2005/Dual-Core CPU) generation of Power Mac G5 had ECC support (didn't know this until I looked it up in MacTracker just now). All other Power Macs (G5 or otherwise) before didn't have it, which still speaks to the point of the person you replied to made.
This needs citation in a big way.
There were complaints about AMD vs Intel. But that's true of any PC workstation - that's not a Mac vs PC thing.
There were complaints about price. All of which still apply to the Apple Silicon version.
Go look up Dell Precision and HP Z workstations circa 2019. Apple offered ONE machine (technically two if you count the 8-16 core vs. 24 & 28 core split). That machine wasn't the apex predator of the Xeon workstations that were out there nor equivalent to the ones that were.
