For what it's worth, if Apple wanted to discontinue the Mac Pro, they would've done so by now. The 2013 Mac Pro was widely regarded as a total failure and enough people complained about it (and switched to Windows as a result) that they spent several years putting out the 2019 model.
I know reaction to the 2023 model has been...well...less than desirable. But, for how much work they did with the Mac Pro customer base to try to rectify things with the 2019 model, it would seem to be far-fetched if they didn't:
(a) continue that open dialogue with that customer base
(b) use that data to determine what would be considered acceptable by the largest possible percentage of that customer base, given what they've already committed to with Apple Silicon's current system architecture in all of the other Macs
Yes, they will definitely alienate people that would've otherwise been fine with a 2019 model with many of the changes made in the 2023 model. That much is pretty much indisputable at this point. Though, whether it's enough to ultimately cause another 2013-Mac-Pro-caliber outcry among those customers or not is a different story. There will definitely be some whose needs will never be served by the 2023 Mac Pro, nor any Apple Silicon Mac Pro. But how much of a percentage of that customer base will be impacted? I think that's the critical question to ask here.
The Mac Pro may very well always just be an Ultra SoC equipped Mac Studio with PCIe slots. Hopefully they do eventually put out an Extreme chip. But even if they do not, this machine will still have people that have no choice but to buy it for its added I/O capability which is why it's still being sold today. Remove that requirement, and then there's no reason why the Ultra SoC version of Mac Studio wouldn't be enough for anyone buying a Mac Pro for performance reasons alone.
Irritatingly, they could've done (and still could yet do) 3 and 4 by socketing the SoC.
I'm wondering if they did analysis of 1 to determine that ECC didn't make that big of a real-world difference in workloads (mind you, I'm not asserting that it doesn't; but I'm wondering if Apple didn't look that over and decide that either it wasn't necessary or that M2 Ultra could make up for it in some other not-so-clearly evident fashion).
2 is definitely off the table, but it's also predicated on Apple's overall gamble that Apple engineered GPUs on Apple SoCs running Apple's OS and leveraging Apple's graphics API will be superior to how it was before. Given that M2 Ultra is a first stab at that in the specific context of the Mac Pro, I wonder if it won't get better. Certainly, the first Intel Mac Pro (MacPro1,1 and, MacPro2,1) had a slew of silly annoyances that were definitely made better with the second major release (Early 2008 or MacPro3,1), and even more that were made better with the one thereafter (Mac Pro Early 2009 or MacPro4,1). Like you said, it's very clearly a stop-gap release.
Agreed. It seems like a safe bet, assuming they don't just pull an M1 iMac and only update it infrequently. I could see them doing that. But I do at least agree that this isn't the final Mac Pro release.
Yeah, and how many of those people does has Apple truly catered to with the Mac Pro in the last 13 years? Yes, the 2019 model was a shot in the arm that harkened back to the 2010/2012/5,1 days. But even that is hamstrung by the inability to run NVIDIA cards on a modern macOS version, among several other inherently missing features.
I'll grant that every 16-inch MacBook Pro released since the 2019 model DOES seem adequately catered to its target market audience. But, those communities that you reference are very poorly served by the Mac Pro and have for at least the last 13 years.
Maybe 3D artists. That one is probably still served by Mac Pros.
But my point remains, Apple is very clearly not the best choice for those types of pro markets and hasn't been for several years now. Remarking that the 2023 Mac Pro is not "a true Mac Pro" for lack of replaceable graphics or RAM, let alone ECC RAM sort of is beside the point. It hasn't been a good platform for most of these high-end use cases for a while now.
Run of the mill Intel processors have been outpacing the Mac Pro for the last decade (due to Apple's inconsistency in releasing updates for it). That's not new. It's just lame that Apple only bothered to make sure that whatever SoC this Mac Pro came with (which ended up being M2 Ultra) only outperformed the Xeon that was in its predecessor, rather than what Apple would've otherwise released in a new model.
I've seen test results from a ton of different places that show the opposite. M2 Ultra seems to beat out any single AMD GPU on its own, but still come a fair bit short compared to those "Duo" MPX modules or multi-card GPU configurations in the 2019 model.
Then again, Apple really seems to be optimizing everything for certain workloads. So, it wouldn't surprise me if M2 Ultra's GPU destroyed the 6900XT in some workloads, but got destroyed by it in others.
Obviously, their "we're optimizing for the workloads WE THINK matter" sentiment from them is flawed in the context of a product the entire mission statement of which is "you build it for your needs".
Here's hoping. The original Mac Pro had some annoying trade-offs from the last Power Mac G5 that it replaced. Hopefully, the second Apple Silicon Mac Pro release fixes some of the ones in this one.
Apple seems to be optimizing for specific workloads. Not saying these things don't have an overall average speed to compare/benchmarch, but I think they shine in some settings and are crap (relative to what else is out there) in others.
