No matter how you spin it, the M2 Mac Pro is a real disappointment. Full stop.
The only excuse I’d give Apple for this disappointment is if they felt it were more important to formally complete the Apple Silicon transition than it was to provide a proper pro machine. So, if this is a stop gap measure to hold us until this time next year, then fine.
The Mac Pro is meant to be the flagship device… the pinnacle of Mac performance. Instead, it’s a Mac Studio with PCI slots. At the very least, an M2 Extreme (2 M2 Ultra chips) is what users are expecting. Apple seems content on comparing to a 4 year old Intel Mac Pro while ignoring the current Intel / nVidia 4090 based solutions. That would address the CPU / GPU scalability concerns or at least help mute them. The other concern is memory. 192 GB is fine for most solutions, but there are very high end needs which go well beyond that.
In my workaday world, where CPU and memory rule, the 28-core 2019 Mac Pro is a hair faster running x64 code compared to the M2 Ultra running arm64 code, and there are jobs I simply can't run on the M2 Ultra due to (vastly) insufficient memory. A 3-year old 64-core Threadripper is more than 3x faster than the M2 Ultra and even supports more memory (256 GB RAM).
It's quite obvious to me that this current machine is a stopgap product. The quad M2 or whatever they were going to call it apparently either didn't work or they couldn't get it to scale. I'm sure their hope is that when the M3 becomes available, they can differentiate the Studio from the Pro. They simply had to complete their "2 year transition" to ASi, and this was their placeholder to do just that.
With Intel shipping PCI Gen 5 (double bandwidth), latest Xeon delivering +2x multicore performance, no GPU-support, the removal of ECC memory, and no way of upgrading... are we looking at the last in a series that started with Mac II?
Should Cook have axed it like Jobs discontinued Newton, servers, and printers? Why didn't he? They have been doing their best not to upgrade the Mac Pro for more than a decade. With all the benefits of Apple Silicon then be honest about "we're no longer in the modular workstation market".
With Intel shipping PCI Gen 5 (double bandwidth), latest Xeon delivering +2x multicore performance, no GPU-support, the removal of ECC memory, and no way of upgrading... are we looking at the last in a series that started with Mac II?
Should Cook have axed it like Jobs discontinued Newton, servers, and printers? Why didn't he? They have been doing their best not to upgrade the Mac Pro for more than a decade. With all the benefits of Apple Silicon then be honest about "we're no longer in the modular workstation market".
Seems short sighted. The M series chips having the equivalent of 7 Afterburner cards would be unlikely to be the case if the Afterburner cards hadn't been developed in the first place. Would Apple's GPU expertise have been what it was if they hadn't invested in MPX, and worked closely with AMD to build GPUs as close to the silicon as they could get them? The extremely high end workstation drives the development that often filters down to the low end.
Apple has repeatedly made a really big deal out of the benefits of integrating the GPU and CPU. While I could imagine a PCI card that acted as an accelerator of sorts, and a GPU could technically play such a role, it would have been surprising to see the Mac Pro support add-on GPUs aa that term is commonly understood.
As I’ve noted before, the path to having a “modular” Mac Pro — in the sense of cobbling together multiple cpus and GPUs — is to buy as many Mac minis as you want and network them.
It might be cool if apple took that to a higher level and made it possible to network multiple minis using something faster that 10 Gb Ethernet. But I think that’s the most one could reasonably hope for in terms of a future modular “Mac Pro”
Xgrid was cool and the concept remains so, but that's not quite the same and requires massively parallel jobs to get the fullest extent. Even theoretically including Thunderbolt for networking, that will still only allow 40 gigabit connections in between, and more realistically, 10gig ethernet is today's solutions. The folks I worked with for this review have all experimented with clustering, but the calculations don't lead themselves perfectly to it.
It's a good idea, don't get me wrong. It's just not for everything that a "modular" Mac Pro could do, or has done.
agreed, it's definitely a niche that can effectively use multiple networked minis. Happens to be my niche, so I find it interesting.
Really, the high-end 'pro' market is a just a collection of niches. The 2008-2012 Mac Pro served several of these niches. Now, no single machine serves them all. The Mac Studio gets some, the Mac Pro gets some, a cluster of minis can get some.
I've long thought Apple should offer an AWS-like 'iCloud Pro' to address more pro niches. I mean 'AWS-like' in a very loose sense. Using AWS isn't terribly hard but it's not super easy, either. There's plenty of room for improvement in the UI/UX, especially for smaller (but still 'pro') users. I know AWS itself offers Mac mini instances, and other providers do too, but there's a lot of room to make it easier, more transparent.
Apple doesn’t understand enterprise very well. All their apps and services are for families and individuals. Not sure if Apple wants to capture the market you’re describing.
There are excellent online creative video workflows that work with NLEs, Google, Microsoft and Amazon captured the enterprise market… I think Apple rather supports others to capture these markets and they stick what they do best.
Apple has repeatedly made a really big deal out of the benefits of integrating the GPU and CPU. While I could imagine a PCI card that acted as an accelerator of sorts, and a GPU could technically play such a role, it would have been surprising to see the Mac Pro support add-on GPUs aa that term is commonly understood.
As I’ve noted before, the path to having a “modular” Mac Pro — in the sense of cobbling together multiple cpus and GPUs — is to buy as many Mac minis as you want and network them.
It might be cool if apple took that to a higher level and made it possible to network multiple minis using something faster that 10 Gb Ethernet. But I think that’s the most one could reasonably hope for in terms of a future modular “Mac Pro”
Xgrid was cool and the concept remains so, but that's not quite the same and requires massively parallel jobs to get the fullest extent. Even theoretically including Thunderbolt for networking, that will still only allow 40 gigabit connections in between, and more realistically, 10gig ethernet is today's solutions. The folks I worked with for this review have all experimented with clustering, but the calculations don't lead themselves perfectly to it.
It's a good idea, don't get me wrong. It's just not for everything that a "modular" Mac Pro could do, or has done.
I've been beating the Xgrid drum, as I think it's an amazing opportunity — especially with so many Apple devices on our networks now — but you shed some light on the topic that I hadn't considered. Thanks.
The previous Intel Mac Pro used the “shredder” design that the new AS Mac Pro does. Not sure why you're bringing ancient devices into this.
Yes my feelings exactly. The "cheese grater" nickname was already used for the Power Mac G5 through Mac Pro 2012 era of case design. That's what I meant by "the old tower design" and by "the new design" I meant "the current design" from 2019 to present.
We need a new nickname for this new generation, when Mike wrote "It's still got the signature 'cheese grater' front and back" that to my ears is "bringing ancient devices into" the conversation, as you said.
We already had the "cheese grater", the "trash can", and now our current design should have its own nickname, I vote that it should be called the "shredder"!
Instead of $54K/$64K, you can get double the performance for $9K in a maxed out Studio or $12K for a maxed out Mac Pro if you need the PCI expansion. And this is disappointing because why?
Because a major use case for PCI expansion is for discrete GPUs, and that use case has been eliminated with Apple Silicon, at least for now.
We know that memory is handled differently on Apple Silicon vs Intel, so why is it assumed that the 192GB max is “a problem?” I’d like to see the test where this problem is actually shown.
There are many scientific applications including simulations and other things that require far more RAM than what the Mac Pro provides. Also a deficit compared to the 1.5TB available in the 2019 Mac Pro.
Also: on what video tasks is Apple’s on-board video proving to be an issue vs a separate video card?
Ever hear of 3D rendering? VFX? VR? There is an increasing need for powerful realtime graphics processing, and the Mac Pro falls way short here. The M2 Ultra is far slower than the top of the line graphics cards available to PC users, not even getting into the fact that you can run multiple cards in parallel. An M2 "Extreme" may have bridged that gap enough to make the Mac Pro somewhat competitive, but that didn't happen.
It would also be interesting to see if the larger enclosure and fan system on the Mac Pro would allow users to push it harder and longer without throttling than the Studio with its smaller enclosure and single fan.
Mabbe, but that's unlikely to make a significant difference.
I definitely hope that Apple has. MUCH more in store for the Mac Pro in the next year or two.
There was obviously a major push to launch/announce their vr headset and no doubt, that took (and is taking) the brunt of R&D from hardware and software development. There is so much that Apple can do.
While the m2 ultra soundly cleaned up m1 ultra design bottlenecks as predicted, it isn’t quite the king that we were hoping for. Now, we’ve heard inklings since M1 that M3 is where things get really serious and so I’m waiting until m3 ultra to really compare.
Yet, the m2 ultra is still quite the impressive powerhouse on a Mac Studio level.
But Mac Pro level? It’s just not.
I have a hard time believing this is some part of a masterful stroke by apple in identifying this mythical subset of users who just need a Mac Studio with more I/O and PCI-E that leaves out graphics upgrades - and that it is somehow a good thing to downgrade the Mac Pro so severely as to fit that group.
What is more likely is threefold: 1. Apple didn’t have a plan to launch - or wasn’t ready to launch - anything above max studio grade horsepower.
2. More likely - Apple wanted to get the Mac Pro ball rolling to get the market onto the Apple Silicon train. The drawbacks to this are:
a. The Mac Pro has only been out a few years and those who invested in MPX modules, etc. are completely out with no true upgrade path and now they are paying for a whole new architecture and starting from scratch.
b. These users are going from a system that when it launched was in the stratosphere of computing. They are now supposed to shell another 6-12k for Mac Studio power.
a. A system where users could upgrade their MULTIPLE processors, graphics, and ram as needed down the line - now has ONE SOC option. It looks like the m2 ultra module may be seapoable later. We will see. If that’s the case, it alleviates things a tiny bit. Especially if there is a way to add more than one SOC.
3. Apple doesn’t value exceedingly high performance computing anymore and wants to artificially limit their top tier offerings as they shift to focus on mass market consumer level offerings and projects like the headset and sneakily move their business toward that and subscription software models.
#2 sounds the most plausible. I sincerely hope it’s not #3. But I truly believe the Mac Pro isn’t so disrespected due to some genius and honest gameplay that isolated this perfect group of users who need only what it offers. It’s s strange day when a high end. PC is a better Mac Pro than the Mac Pro itself. While going PC is not an option for most Mac users, myself included, it does indeed absolutely suck that the pc world has this to hold over Mac osiers heads once again.
Instead of $54K/$64K, you can get double the performance for $9K in a maxed out Studio or $12K for a maxed out Mac Pro if you need the PCI expansion. And this is disappointing because why?
Because a major use case for PCI expansion is for discrete GPUs, and that use case has been eliminated with Apple Silicon, at least for now.
We know that memory is handled differently on Apple Silicon vs Intel, so why is it assumed that the 192GB max is “a problem?” I’d like to see the test where this problem is actually shown.
There are many scientific applications including simulations and other things that require far more RAM than what the Mac Pro provides. Also a deficit compared to the 1.5TB available in the 2019 Mac Pro.
Also: on what video tasks is Apple’s on-board video proving to be an issue vs a separate video card?
Ever hear of 3D rendering? VFX? VR? There is an increasing need for powerful realtime graphics processing, and the Mac Pro falls way short here. The M2 Ultra is far slower than the top of the line graphics cards available to PC users, not even getting into the fact that you can run multiple cards in parallel. An M2 "Extreme" may have bridged that gap enough to make the Mac Pro somewhat competitive, but that didn't happen.
It would also be interesting to see if the larger enclosure and fan system on the Mac Pro would allow users to push it harder and longer without throttling than the Studio with its smaller enclosure and single fan.
Mabbe, but that's unlikely to make a significant difference.
I’m surprised to see anyone defending the Mac Pro decision to be honest. Its quite indefensible. Nothing more or less than a massive letdown.
They really should have waited for M3 and gone with the Quad-max “Extreme” setup. Since the m3 is supposed to have the game changing GPU, that should put it up there with the grown ups and an extreme setup should lay the smack down.
The m2 ultra, while a reformer in its own right is as you said - not ideal for the many GPU workloads in these modern times.
As Apple is now entering the VR and AR space, you’d think they want to make sure that Macs are beyond everyone else in crafting 3D/AR/VR experiences.
But one thing I noticed in the VP keynote WAS THAT apple is copying HoloLens and just going with Unity, while capable, Unity is more of a low grade console, web browser, mobile friendly 3D engine. It’s nowhere near Unreal 5 (the new gold standard) or CryEngine anything like that. So hopefully Apple isn’t just not caring about the GPU as much since their creator-case is limited by Unity anyway. That would be very short-sighted.
Holding out hope that this Mac Pro is just “what we can do today” and that they have crafted it in such a way as to add much much more power down the line.
On another note, Apple really, really should have offered new upgrades for Intel Mac Pro customers. That’s two generations of Mac Pro in a row that went obsolete in 3 years time. That’s nearly MS Zune-level bad faith.
Boy do feel bad for those who actually spect 25-50k on maxed out, superpower B.A. “upgradeable/extpandable” machines only to be dropped like a bad habit shortly after with no upgrade path. That is horrible. Apple used to be much better than this.
Should Cook have axed it like Jobs discontinued Newton, servers, and printers? Why didn't he? They have been doing their best not to upgrade the Mac Pro for more than a decade. With all the benefits of Apple Silicon then be honest about "we're no longer in the modular workstation market".
Seems short sighted. The M series chips having the equivalent of 7 Afterburner cards would be unlikely to be the case if the Afterburner cards hadn't been developed in the first place. Would Apple's GPU expertise have been what it was if they hadn't invested in MPX, and worked closely with AMD to build GPUs as close to the silicon as they could get them? The extremely high end workstation drives the development that often filters down to the low end.
Workstations are not growth. For most purposes companies hire capacity with AWS and process far more at a lower/flexible cost. Why not let Final Cut etc. hook into cloud processing? Apple could absolutely still ship the Studio. Look at cars: EV, self-driving and -parking cars with big screens that auto-update? There was never a trickle-down effect from racing. It was Tesla.
Same with computing. The innovation isn't driven by archaic workstations. The requirements set by smartphones made Apple Silicon. Watch, iPhone, iPad, Vision Pro, fanless computers... I would rather see Apple engineering focus on what they do best than put a barn around Studio and add a Pro sticker. Let IBM be the king of mainframes and HP the king of workstations and calculators.
I definitely hope that Apple has. MUCH more in store for the Mac Pro in the next year or two.
There was obviously a major push to launch/announce their vr headset and no doubt, that took (and is taking) the brunt of R&D from hardware and software development. There is so much that Apple can do.
While the m2 ultra soundly cleaned up m1 ultra design bottlenecks as predicted, it isn’t quite the king that we were hoping for. Now, we’ve heard inklings since M1 that M3 is where things get really serious and so I’m waiting until m3 ultra to really compare.
Yet, the m2 ultra is still quite the impressive powerhouse on a Mac Studio level.
But Mac Pro level? It’s just not.
I have a hard time believing this is some part of a masterful stroke by apple in identifying this mythical subset of users who just need a Mac Studio with more I/O and PCI-E that leaves out graphics upgrades - and that it is somehow a good thing to downgrade the Mac Pro so severely as to fit that group.
What is more likely is threefold: 1. Apple didn’t have a plan to launch - or wasn’t ready to launch - anything above max studio grade horsepower.
2. More likely - Apple wanted to get the Mac Pro ball rolling to get the market onto the Apple Silicon train. The drawbacks to this are:
a. The Mac Pro has only been out a few years and those who invested in MPX modules, etc. are completely out with no true upgrade path and now they are paying for a whole new architecture and starting from scratch.
b. These users are going from a system that when it launched was in the stratosphere of computing. They are now supposed to shell another 6-12k for Mac Studio power.
a. A system where users could upgrade their MULTIPLE processors, graphics, and ram as needed down the line - now has ONE SOC option. It looks like the m2 ultra module may be seapoable later. We will see. If that’s the case, it alleviates things a tiny bit. Especially if there is a way to add more than one SOC.
3. Apple doesn’t value exceedingly high performance computing anymore and wants to artificially limit their top tier offerings as they shift to focus on mass market consumer level offerings and projects like the headset and sneakily move their business toward that and subscription software models.
#2 sounds the most plausible. I sincerely hope it’s not #3. But I truly believe the Mac Pro isn’t so disrespected due to some genius and honest gameplay that isolated this perfect group of users who need only what it offers. It’s s strange day when a high end. PC is a better Mac Pro than the Mac Pro itself. While going PC is not an option for most Mac users, myself included, it does indeed absolutely suck that the pc world has this to hold over Mac osiers heads once again.
What value for Apple is there in the consumer space to limit the top end machines?
3 just doesn't make a lot of sense to me they could have just killed the product and have gotten the value. 1 also seem unlikely given again they could have phoned in a last Intel Mac Pro instead of going to all the trouble they could have just made the Rackable in a Appleish looking case.
2 as you say seems most likely something just didn't work out as planned or when planned.
It seems telling to me they didn't compress the PCIe slots for the M2 MacPro they just deleted out the MPX extra connector. But with GPU's and MPX modules there just don't seem to any PCIe cards that need that extra slot width.
Similarly the M2 now has the back section that with no RAM slots is just a big empty space.
We know that memory is handled differently on Apple Silicon vs Intel, so why is it assumed that the 192GB max is “a problem?” I’d like to see the test where this problem is actually shown. Also: on what video tasks is Apple’s on-board video proving to be an issue vs a separate video card?
It has been well documented by plenty of reviewers that the insufficient memory expansion is a problem. 192GB v 1.5TB? Huge difference and apps that are memory hungry will choke with 192GB. Memory may be faster on chip, but it still has a limited cap and apps that want more memory cannot do it with 16, 24, 32, 64, 96 or now 192GB. Please explain how you think memory is handled different with Apple Silicon where less memory is preferred? The iMac went from supporting 128GB of RAM to 16GB of RAM. I used an M1 iMac and iMovie bounced in the dock 10 times before it actually opened. How is the limited RAM an improvement? People doing heavy video editing on a 32GB Mac Studio say their Pro apps choke because of the limited memory. The apps want more, but none is available.
Did you miss the article last year when Apple got busted for intentionally cropping the GPU graph of the Mac Studio with M1 Ultra? Apple tried to claim the GPU in the M1 Ultra was faster than the fastest dedicated GPU card at the time; however, Apple's graph was intentionally cut off at 200w of power consumption for the GPU. It did not take long for others to post the real GPU results of the dedicated card operating at its full power potential over 300w, and smoking the M1 Ultra GPU, limited to 200w. In a desktop Mac, power efficiency is not an issue. That's why Apple took so long to do an Apple Silicon Mac Pro. The Intel Mac Pro with the high performance GPU cards outperform the fastest on-chip GPU in Apple Silicon. Now that it is 3 years later, Apple figured they had to do something, so they stuck a Mac Studio inside the Mac Pro case, and restricted the PCI-E card expansion to exclude GPU cards. They know a dedicated GPU card will outperform the onboard GPU of the M2 Ultra.
No matter how you spin it, the M2 Mac Pro is a real disappointment. Full stop.
The only excuse I’d give Apple for this disappointment is if they felt it were more important to formally complete the Apple Silicon transition than it was to provide a proper pro machine. So, if this is a stop gap measure to hold us until this time next year, then fine.
The Mac Pro is meant to be the flagship device… the pinnacle of Mac performance. Instead, it’s a Mac Studio with PCI slots. At the very least, an M2 Extreme (2 M2 Ultra chips) is what users are expecting. Apple seems content on comparing to a 4 year old Intel Mac Pro while ignoring the current Intel / nVidia 4090 based solutions. That would address the CPU / GPU scalability concerns or at least help mute them. The other concern is memory. 192 GB is fine for most solutions, but there are very high end needs which go well beyond that.
You nailed it right there. Three years later and Apple could not finish their Apple Silicon transition, and it has become an embarrassment. Apple completed the Intel transition of ALL Macs in 270 days. They didn't know what to do with the Mac Pro because they knew the Intel/nVidia 4090 combination smoked the M1 Ultra in GPU performance. So they threw in the towel and said F'it. Just release a half-a$$ Mac Pro with a Mac Studio inside and limited card support, and no GPU card support. Why not, let's make it $1,000 more than the model it is replacing. Watch the keynote of the Blue and White G3. Steve Jobs said it best...our Pro customers want expandability and they want easy access inside to all the parts. Those days are long gone. Now we have new Macs with one year old CPUs (M2 released June 2022), and an iMac with an almost 3 year old CPU. Yeah, that's progress and better than Intel's roadmap that saw year after year of upgraded and faster Macs with new CPUs. So the new Mac Pro - $1,000 price hike, and $3,000 more for slots with limited support. Even a Pro would not buy this.
The tech world seems to forget Apple is a business. They are watched by Wall Street like no other company. The vast majority of their financial success and sales volume comes from iPhone sales at 50%. Mac sales are 10% of revenue. Mac sales are down 40% in the first quarter of 2023. Therefore the Mac Pro is a super niche product. The business incentive to create a killer Mac Pro don’t really exist. Apple maybe decided to leave the super high performance market to Intel and AMD, unable to cost justify the creation of a truly high end Mac Pro. I also think they are running into technical/scaling issues. Creating the M2 Ultra by joining two M1 Max dies together is challenging. some have suggested creating an M2 Extreme with four dies joined together - the ability to coordinate processing across four systems joined together may be too challenging or you may run into a law of diminishing returns scaling problem. The M3 will use 3 nm lithography and will therefore get a performance and energy efficiency increase which will allow Apple to create more powerful single die systems. But the technical challenges and costs of creating super high end systems may not be justified purely from a business perspective.
Seems to me the best scenario is that this is a "stopgap" machine as many have alluded, and it was done purely to say "we've completed the transition". This machine is, without question, and indefensible answer for the use case the previous generation provided.
It also begs the question of whether Apple is in over its head on silicon design. The GPUs are purely passable at this point (where's the hardware ray tracing?), and the 'revolutionary' M3 that likely should have been in this machine based on the roadmap has yet to materialize (and I don't blame TSMC). Justifying the comparative lack of memory with throughput does not the negate far fewer tasks can be completed per clock cycle when you reduce your total addressable memory by 87.5%, the separation of memory between CPU and GPU tasks further notwithstanding.
I think it's safe to say Apple abandoned the high end market for own interests (vertical integration, bottom line). This is starting to feel like the PowerPC days, and how much better a G3 was in spite of its lower clock speed.
Its target audience are companies, mostly creative ones. Tax write-offs are typically 3 years. Average replacement is 3-5 years.
I don’t see any reason even for those companies to invest in a Mac Pro. The Studio works just fine.
And for the ones who do? E.g AI development - they’ll go for a PC with CUDA powered hardware & software.
The Mac Pro has become irrelevant. And that’s fine.
Fine = “good enough.”
that’s not the Apple way.
From Inc:
Apple's current CEO, during a 2015 episode of 60 minutes. Charlie Rose, the host, asked Cook to explain why he felt Jobs was a singular figure in the tech industry. Cook's response was that Jobs was someone "who had this incredible uncanny ability to see around the corner. Who had this relentless driving force for perfection."
"Good isn't good enough," Cook continued. "It has to be great. As Steve used to say "insanely great."
Comments
The only excuse I’d give Apple for this disappointment is if they felt it were more important to formally complete the Apple Silicon transition than it was to provide a proper pro machine. So, if this is a stop gap measure to hold us until this time next year, then fine.
The Mac Pro is meant to be the flagship device… the pinnacle of Mac performance. Instead, it’s a Mac Studio with PCI slots. At the very least, an M2 Extreme (2 M2 Ultra chips) is what users are expecting. Apple seems content on comparing to a 4 year old Intel Mac Pro while ignoring the current Intel / nVidia 4090 based solutions. That would address the CPU / GPU scalability concerns or at least help mute them.
The other concern is memory. 192 GB is fine for most solutions, but there are very high end needs which go well beyond that.
Should Cook have axed it like Jobs discontinued Newton, servers, and printers? Why didn't he? They have been doing their best not to upgrade the Mac Pro for more than a decade. With all the benefits of Apple Silicon then be honest about "we're no longer in the modular workstation market".
I don’t see any reason even for those companies to invest in a Mac Pro. The Studio works just fine.
And for the ones who do? E.g AI development - they’ll go for a PC with CUDA powered hardware & software.
The Mac Pro has become irrelevant. And that’s fine.
I've been beating the Xgrid drum, as I think it's an amazing opportunity — especially with so many Apple devices on our networks now — but you shed some light on the topic that I hadn't considered. Thanks.
We need a new nickname for this new generation, when Mike wrote "It's still got the signature 'cheese grater' front and back" that to my ears is "bringing ancient devices into" the conversation, as you said.
We already had the "cheese grater", the "trash can", and now our current design should have its own nickname, I vote that it should be called the "shredder"!
There are many scientific applications including simulations and other things that require far more RAM than what the Mac Pro provides. Also a deficit compared to the 1.5TB available in the 2019 Mac Pro.
Ever hear of 3D rendering? VFX? VR? There is an increasing need for powerful realtime graphics processing, and the Mac Pro falls way short here. The M2 Ultra is far slower than the top of the line graphics cards available to PC users, not even getting into the fact that you can run multiple cards in parallel. An M2 "Extreme" may have bridged that gap enough to make the Mac Pro somewhat competitive, but that didn't happen.
Mabbe, but that's unlikely to make a significant difference.
1. Apple didn’t have a plan to launch - or wasn’t ready to launch - anything above max studio grade horsepower.
a. The Mac Pro has only been out a few years and those who invested in MPX modules, etc. are completely out with no true upgrade path and now they are paying for a whole new architecture and starting from scratch.
b. These users are going from a system that when it launched was in the stratosphere of computing. They are now supposed to shell another 6-12k for Mac Studio power.
a. A system where users could upgrade their MULTIPLE processors, graphics, and ram as needed down the line - now has ONE SOC option. It looks like the m2 ultra module may be seapoable later. We will see. If that’s the case, it alleviates things a tiny bit. Especially if there is a way to add more than one SOC.
3. Apple doesn’t value exceedingly high performance computing anymore and wants to artificially limit their top tier offerings as they shift to focus on mass market consumer level offerings and projects like the headset and sneakily move their business toward that and subscription software models.
Maybe a true modular Mac will come in a few more years
I suppose that depends on how you define modular, but I doubt it.
But one thing I noticed in the VP keynote WAS THAT apple is copying HoloLens and just going with Unity, while capable, Unity is more of a low grade console, web browser, mobile friendly 3D engine. It’s nowhere near Unreal 5 (the new gold standard) or CryEngine anything like that. So hopefully Apple isn’t just not caring about the GPU as much since their creator-case is limited by Unity anyway. That would be very short-sighted.
Same with computing. The innovation isn't driven by archaic workstations. The requirements set by smartphones made Apple Silicon. Watch, iPhone, iPad, Vision Pro, fanless computers... I would rather see Apple engineering focus on what they do best than put a barn around Studio and add a Pro sticker. Let IBM be the king of mainframes and HP the king of workstations and calculators.
3 just doesn't make a lot of sense to me they could have just killed the product and have gotten the value.
1 also seem unlikely given again they could have phoned in a last Intel Mac Pro instead of going to all the trouble they could have just made the Rackable in a Appleish looking case.
2 as you say seems most likely something just didn't work out as planned or when planned.
It seems telling to me they didn't compress the PCIe slots for the M2 MacPro they just deleted out the MPX extra connector. But with GPU's and MPX modules there just don't seem to any PCIe cards that need that extra slot width.
Similarly the M2 now has the back section that with no RAM slots is just a big empty space.
Did you miss the article last year when Apple got busted for intentionally cropping the GPU graph of the Mac Studio with M1 Ultra? Apple tried to claim the GPU in the M1 Ultra was faster than the fastest dedicated GPU card at the time; however, Apple's graph was intentionally cut off at 200w of power consumption for the GPU. It did not take long for others to post the real GPU results of the dedicated card operating at its full power potential over 300w, and smoking the M1 Ultra GPU, limited to 200w. In a desktop Mac, power efficiency is not an issue. That's why Apple took so long to do an Apple Silicon Mac Pro. The Intel Mac Pro with the high performance GPU cards outperform the fastest on-chip GPU in Apple Silicon. Now that it is 3 years later, Apple figured they had to do something, so they stuck a Mac Studio inside the Mac Pro case, and restricted the PCI-E card expansion to exclude GPU cards. They know a dedicated GPU card will outperform the onboard GPU of the M2 Ultra.
that’s not the Apple way.
Apple's current CEO, during a 2015 episode of 60 minutes. Charlie Rose, the host, asked Cook to explain why he felt Jobs was a singular figure in the tech industry. Cook's response was that Jobs was someone "who had this incredible uncanny ability to see around the corner. Who had this relentless driving force for perfection."
"Good isn't good enough," Cook continued. "It has to be great. As Steve used to say "insanely great."