A fantasy for me would be an ASI Mac Pro, with the ability to insert a PCIe (or external Thunderbolt) x86 "card" for those that want x86 compatibility on some level.
Just like external GPU boxes, if someone could figure out get full, native x86 processing on an ASI Mac, I would seriously consider "retiring" my 2020 iMac, but as I need x86 to run Windows I'm joined at the hip with Intel for the long, foreseeable future.
I think Apple should name the desktop-class M1 chip the “M1 Meta” … (LOL, but, if not for the Facebook thing, I actually do think it would have been a good name, if maybe a bit weird or pretentious)
Seriously, though, I strongly dislike suggestions like “M1 Ultra” and “M1 Extreme” — for me, they don’t differentiate clearly enough from M1 Max. I think I’d argue for “M1 Pro-W” and “M1 Max-W” …
Boring, but effective, and it sends the right competitive message by adopting the same differentiator Intel uses for Xeon-W. Plus, it opens the door to other sorts of differentiators for other, even more specialized, SOC configurations in the future.
Mt recollection was that the previous generation Mac Pro (aka the trash can design) was dropped in favor of the cheese grater because of thermal limitation in the former. Now with M1-line power efficiencies solving that, one wonders if that design (a beautiful one, albeit limited by physics) if it will make a reappearance. Can't innovate? My ass....
Let us all hope that Apple will sell ARM motherboards that will fit into the 2019 Mac Pro, along with driver support for at least the 1st party dGPU modules, HDD modules, Afterburner modules, etc. They owill hopefully sell 1st party 64 and 128 GPU core MPX modules too, for all the customers who just need more GPU compute.
If the rumored half sized Apple Silicon Mac Pro starts at $2000, a full sized Mac Pro (2019 model) starts at $5000, and a Mac mini starts at $700, Apple will have a full headless desktop lineup spanning all the price tiers since 2006, just before the PPC to Intel transition. Even more so than back then even. That "half size Mac Pro" has to hit $2000 though.
Hope away. Not gonna happen. Sell Apple Silicon mother boards that the user can swap out in the current Intel Mac Pro? Hilarious. Ridiculous.
You can join along lkrupp, and hope with us! We need all the help we can get.
The 2019 Mac Pro is as modular a design as there is. As Lego as it can get. There aren’t any wires in it save for aux power for high Watt 3rd part dGPUs or PCIe cards. (Well, there is the speaker too). Owners don’t need any tools to take it apart, with everything accessible by hand with easy access everywhere. A machine to last for a decade or more.
So if there is ever machine where Apple sells a motherboard kit, the 2019 Mac Pro is it.
Not. Going. To. Happen. Apple Silicon is a totally different architecture. It's literally a different way of building a computer. It doesn't matter that the 2019 MP is Lego-like. They would have to ensure that the board was compatible with everything from Afterburner cards to discreet RAM to GPUs. While some of these may be options on a redesigned MP with Apple Silicon, they sure as hell aren't going to cannibalize new MP sales by offering a board that makes purchasing a new one unnecessary. It would also be unprecedented for Apple. Did they do anything of the kind when they went from G5 to Intel? Of course not. We're talking about an even greater transition here.
Apple really shouldn't worry about cannibalization. The boards aren't going to be cheap, and they shouldn't worry about cannibalizing sales of the half sized Apple Silicon Mac Pro at all as maybe half to three quarters if the Mac Pro userbase wouldn't upgrade to the rumored half sized Apple Silicon Mac Pro. Most of them want the 2019 Mac Pro sized machine for reasons afforded by a big box. They like the internal storage capacity. They like the 8 PCIe slots. They like the memory capacity. They like being able to do 1000 Watts worth of GPU compute. If Apple GPUs are 3x more power efficient, it only means they can pack 3x more GPU perf in the box. They like that it comes in a rack mount case. Heck, Apple had the front of 80 "rack mount Mac Pros" as wall decorations in their MBP intro videos in the "secret Apple Silicon studio" where Srouji talks. So, they like it too! They may pick up more "prosumer" sales with the half sized Mac Pro, but if it doesn't have the capacity of the Mac Pro, I don't think those people will switch.
So, it really is necessary to have both classes of headless desktop: a workstation and prosumer oriented mid-sized desktop. They could have even more form factors too, but having an actual 3 tiers of headless desktop is a nice big step. They really are monied enough to give buyers choices here. Even in the laptop space they can have more choices.
The problem is you are starting from a very questionable assumption: That the new Mac Pro won't be geared towards true professionals. While you make some decent points after that, it's all fruit from the poison tree. Whatever the form factor, the Mac Pro is going to be Apple's workstation-class machine. There is really no reason they would release a prosumer desktop. Almost anyone who wants that will buy the new iMac and its eventual "pro" version. There just isn't a large enough market for what you obviously want (which, as member here for almost 22 years, I can tell you is nothing new...there is always a small contingent of members who want prosumer headless machine).
So, it really is necessary to have both classes of headless desktop: a workstation and prosumer oriented mid-sized desktop.
This is where you really go off the rails. It's not even close to necessary.
They could have even more form factors too, but having an actual 3 tiers of headless desktop is a nice big step.
They could. But they won't. Any it's a nice big step towards...what?
They really are monied enough to give buyers choices here. Even in the laptop space they can have more choices.
They are "monied enough?" Oh boy. It's no wonder you are making the predictions and arguments you are. Product development doesn't work this way. Apple isn't going to say "hey, we've got enough money that we should give buyers more choices." No, they will look at the market and determine the right innovations for the segment, like they always do. That doesn't mean they will always be right. But it also doesn't mean they are going to release a third headless desktop to make a small % of the desktop customer base happy. I also still find it highly unlikely they are going to release an Apple Silicon board, though I agree there is a better chance of that than a third headless device.
Mt recollection was that the previous generation Mac Pro (aka the trash can design) was dropped in favor of the cheese grater because of thermal limitation in the former. Now with M1-line power efficiencies solving that, one wonders if that design (a beautiful one, albeit limited by physics) if it will make a reappearance. Can't innovate? My ass....
I hope not. A Mac mini with an M1 Max gets them 90% of the way there for a compact or SFF desktop, and I hope that's as far as they go.
A design like the 2013 Mac Pro can only work if Apple actually updates it on a regular basis, provides a way for them to go into racks, and makes it easy for users and software to use them in a cluster, even in a desk through TB networking. Even so, a compact form factor just isn't a good fit for the potential buyers who are looking for a workstation.
I'm really hoping this rumored half-size Mac Pro still supports at least 1 MPX module, plus 2x wide PCIe card. And, there still needs to be a way for them to be easily put into racks. And, there needs to be an Apple Silicon version of the 2019 Mac Pro.
Mt recollection was that the previous generation Mac Pro (aka the trash can design) was dropped in favor of the cheese grater because of thermal limitation in the former. Now with M1-line power efficiencies solving that, one wonders if that design (a beautiful one, albeit limited by physics) if it will make a reappearance. Can't innovate? My ass....
I hope not. A Mac mini with an M1 Max gets them 90% of the way there for a compact or SFF desktop, and I hope that's as far as they go.
A design like the 2013 Mac Pro can only work if Apple actually updates it on a regular basis, provides a way for them to go into racks, and makes it easy for users and software to use them in a cluster, even in a desk through TB networking. Even so, a compact form factor just isn't a good fit for the potential buyers who are looking for a workstation.
I'm really hoping this rumored half-size Mac Pro still supports at least 1 MPX module, plus 2x wide PCIe card. And, there still needs to be a way for them to be easily put into racks. And, there needs to be an Apple Silicon version of the 2019 Mac Pro.
You know, there is an important decision here for Apple designers. There has been a coordinated effort on most macs to move away from upgradable designs. Maybe this is driven by the engineering, but the idea of modular parts that can be upgraded by owners seems to be going away - at least in consumer macs.
The problem is you are starting from a very questionable assumption: That the new Mac Pro won't be geared towards true professionals. While you make some decent points after that, it's all fruit from the poison tree. Whatever the form factor, the Mac Pro is going to be Apple's workstation-class machine. There is really no reason they would release a prosumer desktop. Almost anyone who wants that will buy the new iMac and its eventual "pro" version. There just isn't a large enough market for what you obviously want (which, as member here for almost 22 years, I can tell you is nothing new...there is always a small contingent of members who want prosumer headless machine).
I'm mostly talking about it from a perspective of gaining more sales. I don't think you do that in the workstation market by going with a 2013 Mac Pro type form factor, whether it is a cylinder or a box. If there is a lesson to be learned from the 2013 Mac Pro debacle, it is that it's not a form factor that a lot of buyers want, especially at its price tiers. It ended up as mostly a workstation for FCP users, not as a generic workstation that a lot of different types of users can use. Whatever its form factor benefits were, not enough people wanted them. They seemingly learned this and created the 2019 Mac Pro. Pricey, but at least it has a larger market of buyers for it.
So, if you hear rumors that the Apple Silicon Mac Pro is going to be about half the size, we really should be skeptical of it. If Apple ships it, and it has no internal expansion capacity, how is it going to be different from the 2013 Mac Pro? What are they going to do to achieve better sales than they were before? And the 2013 Mac Pro wasn't the first go around. The first go around was the 2000 Power Mac G4 Cube, and that failed too.
Maybe the 3rd time is the charm, but why bother putting yourself in such a situation? What are they gaining? Why make the bet? What will be different this time?
sdw2001 said:They are "monied enough?" Oh boy. It's no wonder you are making the predictions and arguments you are. Product development doesn't work this way. Apple isn't going to say "hey, we've got enough money that we should give buyers more choices." No, they will look at the market and determine the right innovations for the segment, like they always do. Thatit doesn't mean they will always be right. But it also doesn't mean they are going to release a third headless desktop to make a small % of the desktop customer base happy. I also still find it highly unlikely they are going to release an Apple Silicon board, though I agree there is a better chance of that than a third headless device.
Well, "monied enough" is probably a wrong choice of words then. I want them to have more Mac sales, enough such that they have a permanence in the market with a near equivalent ecosystem of software to Windows & Linux. I think that is going to be about 15% of PC market sales, especially if it is the money spending customers. I don't think they will get there with their current lineup. They aren't 2001 Apple anymore, when they needed to limit what products they could offer. 2021 Apple is "monied enough" to have more ambition, to try to double or triple sales. So, they need to try to get more sales in the workstation markets, the gaming market, on office worker desks, just in as many markets as possible.
A MBA15? Yes. A mid-tier headless desktop box? Yes. An $800 MBA13 with an A15? Yes. Buy a game studio? Yes. Get more engineering apps on the platform? Yes. Develop some competitive applications in specific markets? Yes. Similar thing for iPads too. Lots of work to do.
Remember, they are spending billions annually on Apple TV+ content. What is the strategic value of that? Is it a long game to provide content for the rumored AR/VR hardware? How is it different or better then offering more types of hardware and more software for its current platforms? Part if it is to prevent the big streaming incumbents from shafting them, but that's also true of other markets and niches too.
Mt recollection was that the previous generation Mac Pro (aka the trash can design) was dropped in favor of the cheese grater because of thermal limitation in the former. Now with M1-line power efficiencies solving that, one wonders if that design (a beautiful one, albeit limited by physics) if it will make a reappearance. Can't innovate? My ass....
I hope not. A Mac mini with an M1 Max gets them 90% of the way there for a compact or SFF desktop, and I hope that's as far as they go.
A design like the 2013 Mac Pro can only work if Apple actually updates it on a regular basis, provides a way for them to go into racks, and makes it easy for users and software to use them in a cluster, even in a desk through TB networking. Even so, a compact form factor just isn't a good fit for the potential buyers who are looking for a workstation.
I'm really hoping this rumored half-size Mac Pro still supports at least 1 MPX module, plus 2x wide PCIe card. And, there still needs to be a way for them to be easily put into racks. And, there needs to be an Apple Silicon version of the 2019 Mac Pro.
All the MPX slots of the NewMacPro are in the lower half with the power supply. Almost exactly half if you get 4 double-wide PCIe or 2 MPX 1 that could mirror the all the slots and power supply inside the box.
The back chamber of the tower is mostly empty space with memory and more air movement. That to me looks to be a vast amount of space for the M1 Max or 2 to sit with heaps of room for a heat spreader and assuming memory will be limited to what is unified with the SOC(s) there should be room for 2 if not more. Wonder if this is how they get a half-height tower. Use a board sandwich like an iPhone to get M's on the back PCIe in the front. The same board Repeats upwards in a full-height tower with a second power supply at the top of the case and a high-speed bridging cable.
The problem is you are starting from a very questionable assumption: That the new Mac Pro won't be geared towards true professionals. While you make some decent points after that, it's all fruit from the poison tree. Whatever the form factor, the Mac Pro is going to be Apple's workstation-class machine. There is really no reason they would release a prosumer desktop. Almost anyone who wants that will buy the new iMac and its eventual "pro" version. There just isn't a large enough market for what you obviously want (which, as member here for almost 22 years, I can tell you is nothing new...there is always a small contingent of members who want prosumer headless machine).
I'm mostly talking about it from a perspective of gaining more sales. I don't think you do that in the workstation market by going with a 2013 Mac Pro type form factor, whether it is a cylinder or a box. If there is a lesson to be learned from the 2013 Mac Pro debacle, it is that it's not a form factor that a lot of buyers want, especially at its price tiers. It ended up as mostly a workstation for FCP users, not as a generic workstation that a lot of different types of users can use. Whatever its form factor benefits were, not enough people wanted them. They seemingly learned this and created the 2019 Mac Pro. Pricey, but at least it has a larger market of buyers for it.
So, if you hear rumors that the Apple Silicon Mac Pro is going to be about half the size, we really should be skeptical of it. If Apple ships it, and it has no internal expansion capacity, how is it going to be different from the 2013 Mac Pro? What are they going to do to achieve better sales than they were before? And the 2013 Mac Pro wasn't the first go around. The first go around was the 2000 Power Mac G4 Cube, and that failed too.
Maybe the 3rd time is the charm, but why bother putting yourself in such a situation? What are they gaining? Why make the bet? What will be different this time?
The problem in 2013 was that there was still an upgrade path beyond the 2x 3.5TFLOPs of GPU power and they locked it at that level without the option to upgrade the GPUs. On top of that, hardly anyone buys these types of computer any more and the same is true of the 2019 model.
What's different this time is there's not going to be GPU upgrades when it uses unified memory and there's not the same kind of upgrade path needed as manufacturers start to hit limits at 3nm and the power at these levels meets the needs of pretty much every workflow today.
There might be a need for PCIe peripherals in some cases but very few and they can go outside the main box through a PCIe connector.
The Mac Pro only reverted to a PCIe enclosure because of PCIe GPUs and those aren't needed any more.
Another option is that they make a dual-mode iMac. The iMac enclosure has enough cooling power to handle 4x M1 Max. If they made a larger iMac like the XDR display, they could sell the same internals in a server blade for people who want it rack-mounted and if people want the headless setup, they can buy the server model too.
The server version would look closer to the Xserve pictured below the Mac Pro here:
It would have metal all around instead of a display and because it's flat, it allows direct access for increasing storage. It wouldn't use an XDR display panel but a display similar to what's in the MBP.
This iMac Pro could start at $1999 with M1 Pro (16GB min), going up to $3099 for M1 Max (32GB min), $4999 for M1 Max Duo (Extreme, 64GB min), $7999 for M1 Max Quad (Ultra, 64GB min). Up to 256GB memory total.
If they get returns and need to refurb, they can turn the iMac into a server model or vice versa.
Long-term, some people would be unhappy that they can't upgrade GPUs but there's no way to allow this anyway and the options should be cheaper so that doing a full upgrade isn't so bad.
If you’re going to compare the highest end New MacBook Pro with 32 core GPU and 10 Core M1Max and 16 core Neural Engine, then compare it to the Mac Pro with Duo W6900X with Afterburner at 28 cores and 384GB DDR4 memory, at least.
Show people why Apple demonstrated a fully loaded Mac Pro w/ 1.5TB DDR4, 28 Core Xeon and peak Duo GPGPU Logic Pro on stage, or at the very least cite what a fully loaded latest offerings Mac Pro performance can do and how far the new MacBook Pro had to go to even be in the ballpark.
Studios by the Mac Pro for music production and post production never mind 3D Modeling and Engineering because that expansion will be viable for the next 7 years and pay for itself tenfold.
The problem in 2013 was that there was still an upgrade path beyond the 2x 3.5TFLOPs of GPU power and they locked it at that level without the option to upgrade the GPUs. On top of that, hardly anyone buys these types of computer any more and the same is true of the 2019 model.
What's different this time is there's not going to be GPU upgrades when it uses unified memory and there's not the same kind of upgrade path needed as manufacturers start to hit limits at 3nm and the power at these levels meets the needs of pretty much every workflow today.
There might be a need for PCIe peripherals in some cases but very few and they can go outside the main box through a PCIe connector.
The Mac Pro only reverted to a PCIe enclosure because of PCIe GPUs and those aren't needed any more.
I think your comments here are basically the same as the responses for the 2013 Mac Pro, and I don't think the prospective buyers of a Mac Pro would like it all. For the 2013 Mac Pro, Apple said it had a lot of expansion due to 6 TB2 ports. If you wanted more storage, use an external TB RAID or an external USB drive. If you wanted to use a a PCIe card, get a TB2 PCIe breakout box.
This is the antithesis of Appleness. Its buyers want clean minimalism on their desks, so not something with a lot of wires, not something with lot of external boxes, not something with a lot of plugs is something they do not want, and it's something workstation market does not want if the 2013 model is anything to go by. Then, if there is not any dGPU support due to Apple's UMA, that forgoes people who want more than 128 GPU cores in the desktop. The Mac Pro currently offers 4 to 6 dGPUs. They will need to offer a solution for people who just want more. They essentially want to put as much compute power in a box at 1400 Watts as possible.
Regarding dGPUs and UMA, UMA doesn't prevent having dGPUs. Why would it? It would be just like today, where GPU assets have to be moved back and forth from main memory to GPU memory. UMA also doesn't imply memory has to be in-package either. It could all be in DIMM slots. How Apple gets to 1 TB of memory will be interesting to see.
Apple probably would have been more successful with the 2013 Mac Pro model if they updated it, year by year, with more performant CPUs and GPUs, but I bet they knew it wasn't going to sell enough units for many reasons, and stopped development on successive models within a year in favor of the then iMac Pro. And, they had to realign again as they eventually figured out that most of the market didn't want the iMac Pro either and had to pre-announce the iMac Pro 8 months in advance and the Mac Pro 2.5 years in advance.
There better be a book about Apple's Mac decisions from 2011 to 2017. It is absolutely baffling how they could have arrived at the decisions they ended up with, with a lot of decisions in strategic opposition to other decisions. They still are afflicted by those decisions today.
Another option is that they make a dual-mode iMac. The iMac enclosure has enough cooling power to handle 4x M1 Max. If they made a larger iMac like the XDR display, they could sell the same internals in a server blade for people who want it rack-mounted and if people want the headless setup, they can buy the server model too.
The server version would look closer to the Xserve pictured below the Mac Pro here:
It would have metal all around instead of a display and because it's flat, it allows direct access for increasing storage. It wouldn't use an XDR display panel but a display similar to what's in the MBP.
This iMac Pro could start at $1999 with M1 Pro (16GB min), going up to $3099 for M1 Max (32GB min), $4999 for M1 Max Duo (Extreme, 64GB min), $7999 for M1 Max Quad (Ultra, 64GB min). Up to 256GB memory total.
If they get returns and need to refurb, they can turn the iMac into a server model or vice versa.
Long-term, some people would be unhappy that they can't upgrade GPUs but there's no way to allow this anyway and the options should be cheaper so that doing a full upgrade isn't so bad.
Hey, no problems from me on this, I want more product offerings after all. Remember, they have been free to do this, but have chosen not to. They don't view themselves as a computing company, but more like consumer technology company. So, "big iron" computing isn't something they are going to do, or offer a wide range of computing products even in consumer markets. I wonder what is powering Xcode Cloud though. This defacto limits how many Mac units they can sell because they simply don't offer a product in this or that niche, and just as important, work to get software for those niches.
I hope they have a big goal of market permanence, where, if someone buys a Mac, they don't have to give a second thought on whether it will have the software they need to do whatever they want to do. That's not true today. I don't think grabbing a big chunk of the consumer market will help with various niche software, and it leaves them under threat.
The Mac mini needs to have M1 Pro and M1 Max SoC options. The iMac 24 needs to have an M1 Pro option at least. The large iMac needs to have M1 Pro, M1 Max and a 16 p-core + 64 GPU core SoC options, and the Mac Pro is needs have 16+64 and 32+128 options, with rack mount and PCIe slots. Then, they need to have an MBA15 at $1600 or so, and a modular box at $2000. Still think a $800 A15 powered 13" laptop is a good idea. $1000 and $2000 Apple monitors will hopefully be announced sooner rather than later.
The problem in 2013 was that there was still an upgrade path beyond the 2x 3.5TFLOPs of GPU power and they locked it at that level without the option to upgrade the GPUs. On top of that, hardly anyone buys these types of computer any more and the same is true of the 2019 model.
What's different this time is there's not going to be GPU upgrades when it uses unified memory and there's not the same kind of upgrade path needed as manufacturers start to hit limits at 3nm and the power at these levels meets the needs of pretty much every workflow today.
There might be a need for PCIe peripherals in some cases but very few and they can go outside the main box through a PCIe connector.
The Mac Pro only reverted to a PCIe enclosure because of PCIe GPUs and those aren't needed any more.
I think your comments here are basically the same as the responses for the 2013 Mac Pro, and I don't think the prospective buyers of a Mac Pro would like it all. For the 2013 Mac Pro, Apple said it had a lot of expansion due to 6 TB2 ports. If you wanted more storage, use an external TB RAID or an external USB drive. If you wanted to use a a PCIe card, get a TB2 PCIe breakout box.
This is the antithesis of Appleness. Its buyers want clean minimalism on their desks, so not something with a lot of wires, not something with lot of external boxes, not something with a lot of plugs is something they do not want, and it's something workstation market does not want if the 2013 model is anything to go by. Then, if there is not any dGPU support due to Apple's UMA, that forgoes people who want more than 128 GPU cores in the desktop. The Mac Pro currently offers 4 to 6 dGPUs. They will need to offer a solution for people who just want more. They essentially want to put as much compute power in a box at 1400 Watts as possible.
Regarding dGPUs and UMA, UMA doesn't prevent having dGPUs. Why would it? It would be just like today, where GPU assets have to be moved back and forth from main memory to GPU memory. UMA also doesn't imply memory has to be in-package either. It could all be in DIMM slots. How Apple gets to 1 TB of memory will be interesting to see.
Apple probably would have been more successful with the 2013 Mac Pro model if they updated it, year by year, with more performant CPUs and GPUs, but I bet they knew it wasn't going to sell enough units for many reasons, and stopped development on successive models within a year in favor of the then iMac Pro. And, they had to realign again as they eventually figured out that most of the market didn't want the iMac Pro either and had to pre-announce the iMac Pro 8 months in advance and the Mac Pro 2.5 years in advance.
There better be a book about Apple's Mac decisions from 2011 to 2017. It is absolutely baffling how they could have arrived at the decisions they ended up with, with a lot of decisions in strategic opposition to other decisions. They still are afflicted by those decisions today.
The Mac mini needs to have M1 Pro and M1 Max SoC options. The iMac 24 needs to have an M1 Pro option at least. The large iMac needs to have M1 Pro, M1 Max and a 16 p-core + 64 GPU core SoC options, and the Mac Pro is needs have 16+64 and 32+128 options, with rack mount and PCIe slots. Then, they need to have an MBA15 at $1600 or so, and a modular box at $2000. Still think a $800 A15 powered 13" laptop is a good idea. $1000 and $2000 Apple monitors will hopefully be announced sooner rather than later.
One issue with Thunderbolt expansion was relying on 3rd parties who made really expensive and usually not well designed peripherals and they weren't available globally. It's only a small amount of people who need those but for them it's a poorer compromise over an integrated unit. An Apple designed expansion box would be different though and it wouldn't have to use Thunderbolt, there could be a PCIe style connector that connects to the box and the peripherals behave like they do internally. The box could also include Thunderbolt connections.
The easiest route would be to cut the 2019 model in half and leave some slots but I don't see them having PCIe GPU support. This machine was also designed around Intel/AMD parts and the setup has changed completely with Apple Silicon.
I don't think they would do unified memory via DIMMs because it compromises the memory bandwidth. The Mac Pro will have in the region of 1TB+/s memory bandwidth, I don't see them achieving this with plug-in modules. Nvidia used their own NVLink to get around PCIe limits in their servers connecting memory to the GPUs. With such fast SSDs, the need for larger capacities like 1TB of memory is reduced, the use cases for over 256GB are very few. 64GB HBM per chip x4 chips would be enough.
I'd say the decisions made around the Mac Pro over the years make a lot of sense. The first thing is that the larger form factor wasn't selling well in 2012 and hadn't been selling well for a while, they had plans to cancel it way before 2012. The more expensive a computer is, the less frequent the upgrade cycle is. People hang onto them for 10 years. So a radical design and marketing campaign was tried to boost sales in 2013 but with no way to upgrade the GPUs and the higher entry price point (it used to start at $2199-2499), it only appealed to a portion of an already small and diminishing market. The iMac Pro gave them more room to expand thermal capacity (500W vs 300W) but again it had a high price point and no ability to upgrade GPUs. I believe they also had some input/pressure from some of their pro users, likely in the film industry to just make another box like before. While the 2019 model satisfies this group, it cuts out a bigger portion of the much larger volume of buyers at a lower price point.
Today things are different. The latest Macbook Pro is beating the Mac Pro in some workflows. Marques Brownlee tested it out against his $50k Mac Pro (16:50):
He said he used to cart around a 27" iMac to be able to handle the jobs and would now be able to do it on a MBP and on battery. This is for editing and exporting 4K/8K footage. Raw power is the main thing that people have needed and it's not a need that will keep increasing for every use case. There's very few use cases that a maxed out 2019 Mac Pro wouldn't handle even in 10 years time and the same will apply to an Apple Silicon Mac Pro with integrated graphics running at 40TFLOPs+.
I would also like to see some of those kind of Mac models suggested there, especially the larger Air, they used to have 12" and 14" Macbooks. An $800 Air would boost sales a bit but I wonder how they'd hit that price point without reducing storage/RAM. I don't think the M1/A15 chip would make that price difference. They sell an education Air for $899 by lowering margins. The entry iPad is $479 with 256GB storage and 4GB RAM and the Air is $749 so there's probably a way to make a cheaper laptop while maintaining margins. I think they just don't want to chase the lower price points in the PC space.
The more that computers develop, the need for modular computers is going away. Modular over the years has meant RAM, storage and PCIe GPUs. RAM makes more sense being integrated, storage is nice to be able to upgrade and one of the only parts left that can fail with use. Integrated graphics are the way forward.
Mt recollection was that the previous generation Mac Pro (aka the trash can design) was dropped in favor of the cheese grater because of thermal limitation in the former. Now with M1-line power efficiencies solving that, one wonders if that design (a beautiful one, albeit limited by physics) if it will make a reappearance. Can't innovate? My ass....
AI wrote: "AppleInsiderran the benchmark on a 32-core 16-inch MacBook Pro and saw a score of 1,769 for the single-core test, 12,308 for the multi-core version." [emphasis added]
The classical Geekbench benchmark uses the CPU cores of which there are 10 in the M1 MAX (8 performance cores + 2 efficiency cores).
AI wrote: "AppleInsiderran the benchmark on a 32-core 16-inch MacBook Pro and saw a score of 1,769 for the single-core test, 12,308 for the multi-core version." [emphasis added]
The classical Geekbench benchmark uses the CPU cores of which there are 10 in the M1 MAX (8 performance cores + 2 efficiency cores).
32 GPU cores is to what it's referring. The 16" MBPs only come with 10 CPU cores, so the differentiator is the number of GPU cores for the main silicon, which can be 16, 24, or 32.
The easiest route would be to cut the 2019 model in half and leave some slots but I don't see them having PCIe GPU support.
Yes. This is what I'm hoping for with the rumored half sized Apple Silicon Mac Pro. Basically reduce the height of the box by 1 of its axial fans, and reduce the width by the height of the blower fan and DIMM slots. It's pretty darn close to half size by volume. This will leave room for 1 MPX module and 3 1-wide PCIe slots (total of 5). And hopefully, it means it will start out at half price with an M1 Pro or M1 Max.
Being able to share MPX modules, both 2-wide and 4-wide varieties, with the 2019 Mac Pro will increase the amount and kinds of MPX modules available for both models. And, either there is an Apple Silicon version of the 2019 Mac Pro model, or that find a way to rack the half size model.
It seems all but inevitable for them to come up with a solution for people who want more GPU performance on a desk. Whether it is through PCIe slots or clustering, they will need to offer a solution. The usefulness of internal expansion is important for a rather large subset of the workstation market. It's not just PCIe dGPUs. I love the RAID MPX module. I'd like to see 16 TB, 32 TB SSD PCIe solutions. There are lots of niche audio, accelerator and I/O cards.
So, an Apple Silicon Mac Pro like the 2019 model but half sized, absolutely. A machine like the 2013 model would be a mistake.
I don't think they would do unified memory via DIMMs because it compromises the memory bandwidth. The Mac Pro will have in the region of 1TB+/s memory bandwidth, I don't see them achieving this with plug-in modules. Nvidia used their own NVLink to get around PCIe limits in their servers connecting memory to the GPUs. With such fast SSDs, the need for larger capacities like 1TB of memory is reduced, the use cases for over 256GB are very few. 64GB HBM per chip x4 chips would be enough.
It's really quite the mystery what they will do. I hope someone will tear down the 16 GB LPDDR5 modules in the M1 Max/Pro to see how the DRAM chips are arranged. It's non-standard packaging and looks bigger than typical DRAM packages. Maybe it is just two sets of 8 GB LPDDR5 chips, which are in mass production, inside the package. I don't think the 16 GB LPDRR5 packages are in mass production yet, so curious what they are doing with those RAM packages.
If they have in-package memory like the M1 Max, they probably will have to come up with a custom stacking solution to get to 1 TB capacities. 4 M1 Max SoCs in a package would make for 16 LPDDR5 packages, with each package having a stack of up to 4 LPDDR packages. 64 LPDDR5 packages in 16 stacks for 1.6 TB/s of bandwidth. Uh, sounds sporty to have package with 16 packages/SoCs insides. 16 channels of DDR5, only 800 GB/s, in 16 DIMM slots doesn't sound great either, but they'll be able to achieve 2 TB memory capacity.
We will have to wait and see whether they and customers determine if 256 GB of HBM is good enough. For a half sized box, probably enough, given that competitor boxes of the size will be about the same. For a 2019 Mac Pro sized box, probably not, especially rack mount versions, as those competitor boxes will have 1+ TB of RAM capacity.
I'd say the decisions made around the Mac Pro over the years make a lot of sense. The first thing is that the larger form factor wasn't selling well in 2012 and hadn't been selling well for a while, they had plans to cancel it way before 2012. The more expensive a computer is, the less frequent the upgrade cycle is. People hang onto them for 10 years. So a radical design and marketing campaign was tried to boost sales in 2013 but with no way to upgrade the GPUs and the higher entry price point (it used to start at $2199-2499), it only appealed to a portion of an already small and diminishing market. The iMac Pro gave them more room to expand thermal capacity (500W vs 300W) but again it had a high price point and no ability to upgrade GPUs. I believe they also had some input/pressure from some of their pro users, likely in the film industry to just make another box like before. While the 2019 model satisfies this group, it cuts out a bigger portion of the much larger volume of buyers at a lower price point.
The 2013 Mac Pro has a 450 W power supply, just a little bit less than the 2017 iMac Pro. Most of the extra 50 W in the iMac Pro went to powering its display, and the two have about the same power envelope for the computing bits.
For the Mac Pro, I see it as 2 major mistakes. Not only did they make a mistake with the 2013 model, they made it again with the iMac Pro. After 6 years from late 2013 to late 2019, Apple rounded back to the beginning with a big desktop box with modular components. We know they realized the mistakes after 4 years as they had that Mac Pro PR tete-e-tete with journalists in 2017, where they discuss their pro Mac plans. Since when does Apple ever talk about unannounced products?
If it is a diminishing market, they wouldn't be selling a 2019 Mac Pro, right? It seems more that they didn't understand the workstation market, outside of a narrow video editor one. They didn't anticipate machine learning, I think they gave up on CAD and 3D in that 2012 to 2015 period. Computational codes (anything involving large sets of differential equations) didn't interest them that much? There is a market out there, where computational requirements are basically infinite, so my expectation is the workstation market is here to stay for the foreseeable future. And, Apple will need to race to get computational codes to use its Metal API to even have a chance at all these potential buyers. UMA on Apple Silicon is a big advantage, giving these codes 20, 30, 40 GB of memory, and a highly threaded GPU, with presumably, 100 to 200 GB of memory in the future. The time to push to get software on Metal is now.
They also trot out the "Pro workflow team" term in its statements probably as way to assure customers that they know what they are doing now. I guess that is ok, but if you look at it glass half empty, it means Apple's marketing team is so bereft of high end computing knowledge that they need a team of "pros" to tell them what makes a good high end computer. Hopefully it's the former (assurance), not the latter (absence of knowledge).
This is just the Mac Pro! Every Mac product had issues from 2012 to 2019. Something went wrong with the marketing team or the decision makers for Macs in that time frame. Not having a a successor to the Thunderbolt display was also a crazy pants decision. Someone has got be dying to write a book.
Today things are different. The latest Macbook Pro is beating the Mac Pro in some workflows. Marques Brownlee tested it out against his $50k Mac Pro (16:50) ... There's very few use cases that a maxed out 2019 Mac Pro wouldn't handle even in 10 years time and the same will apply to an Apple Silicon Mac Pro with integrated graphics running at 40TFLOPs+.
It's great that the 2021 MBP machines have Mac Pro performance, but that doesn't mean it ends there. It only means people want 2x to 10x the performance of the 2021 MBP in a desktop box. My perspective is that there are a lot of use cases that have continually increasing compute needs, and therefore, those customers will continually want the most compute performance that can be placed in a desktop box with a 1000 W of power. Like I said above, Apple has to sponsor a lot of codes to convert from CPU compute to GPU compute (Metal), than have Metal hardware capable of running an ecosystem high performance computing codes.
I would also like to see some of those kind of Mac models suggested there, especially the larger Air, they used to have 12" and 14" Macbooks. An $800 Air would boost sales a bit but I wonder how they'd hit that price point without reducing storage/RAM. I don't think the M1/A15 chip would make that price difference. They sell an education Air for $899 by lowering margins. The entry iPad is $479 with 256GB storage and 4GB RAM and the Air is $749 so there's probably a way to make a cheaper laptop while maintaining margins. I think they just don't want to chase the lower price points in the PC space.
The more that computers develop, the need for modular computers is going away. Modular over the years has meant RAM, storage and PCIe GPUs. RAM makes more sense being integrated, storage is nice to be able to upgrade and one of the only parts left that can fail with use. Integrated graphics are the way forward.
The iPad Air has an A15, 4 GB RAM, 64 GB NAND and 11" display at $600. The iPad Pro has an M1, 8 GB RAM, 128 GB NAND and 11" display for $800. It's not a stretch for a MBA13 with A15, 8 GB RAM, 256 GB NAND and 13" display to cost $800 including desired margins, especially in 2022.
I think the time is now to have an $800 laptop model, with Edu discount to $700. Trying to sell more units downmarket also means they have to get a self sustaining games ecosystem for macOS. Getting more units in the market is vitally important to establishing that gaming ecosystem, so a cheaper Mac is a necessary first step. A Mac mini with M1 Pro for $800 and with an M1 Max for $1000 would be part of a goal of getting more units out there. They will need to buy a few game developers, and become a publisher. On the pro software, easiest path is probably like what they did with Blender, sponsor more open source software, get a lot of these pro apps on Mac. A first class Matlab implementation. Creo. Computational codes. Possibly ship their own set of pro software for STEM fields, etc.
Comments
Seriously, though, I strongly dislike suggestions like “M1 Ultra” and “M1 Extreme” — for me, they don’t differentiate clearly enough from M1 Max. I think I’d argue for “M1 Pro-W” and “M1 Max-W” …
The problem is you are starting from a very questionable assumption: That the new Mac Pro won't be geared towards true professionals. While you make some decent points after that, it's all fruit from the poison tree. Whatever the form factor, the Mac Pro is going to be Apple's workstation-class machine. There is really no reason they would release a prosumer desktop. Almost anyone who wants that will buy the new iMac and its eventual "pro" version. There just isn't a large enough market for what you obviously want (which, as member here for almost 22 years, I can tell you is nothing new...there is always a small contingent of members who want prosumer headless machine).
This is where you really go off the rails. It's not even close to necessary.
They could. But they won't. Any it's a nice big step towards...what?
They are "monied enough?" Oh boy. It's no wonder you are making the predictions and arguments you are. Product development doesn't work this way. Apple isn't going to say "hey, we've got enough money that we should give buyers more choices." No, they will look at the market and determine the right innovations for the segment, like they always do. That doesn't mean they will always be right. But it also doesn't mean they are going to release a third headless desktop to make a small % of the desktop customer base happy. I also still find it highly unlikely they are going to release an Apple Silicon board, though I agree there is a better chance of that than a third headless device.
A design like the 2013 Mac Pro can only work if Apple actually updates it on a regular basis, provides a way for them to go into racks, and makes it easy for users and software to use them in a cluster, even in a desk through TB networking. Even so, a compact form factor just isn't a good fit for the potential buyers who are looking for a workstation.
I'm really hoping this rumored half-size Mac Pro still supports at least 1 MPX module, plus 2x wide PCIe card. And, there still needs to be a way for them to be easily put into racks. And, there needs to be an Apple Silicon version of the 2019 Mac Pro.
So, if you hear rumors that the Apple Silicon Mac Pro is going to be about half the size, we really should be skeptical of it. If Apple ships it, and it has no internal expansion capacity, how is it going to be different from the 2013 Mac Pro? What are they going to do to achieve better sales than they were before? And the 2013 Mac Pro wasn't the first go around. The first go around was the 2000 Power Mac G4 Cube, and that failed too.
Maybe the 3rd time is the charm, but why bother putting yourself in such a situation? What are they gaining? Why make the bet? What will be different this time?
Well, "monied enough" is probably a wrong choice of words then. I want them to have more Mac sales, enough such that they have a permanence in the market with a near equivalent ecosystem of software to Windows & Linux. I think that is going to be about 15% of PC market sales, especially if it is the money spending customers. I don't think they will get there with their current lineup. They aren't 2001 Apple anymore, when they needed to limit what products they could offer. 2021 Apple is "monied enough" to have more ambition, to try to double or triple sales. So, they need to try to get more sales in the workstation markets, the gaming market, on office worker desks, just in as many markets as possible.
A MBA15? Yes. A mid-tier headless desktop box? Yes. An $800 MBA13 with an A15? Yes. Buy a game studio? Yes. Get more engineering apps on the platform? Yes. Develop some competitive applications in specific markets? Yes. Similar thing for iPads too. Lots of work to do.
Remember, they are spending billions annually on Apple TV+ content. What is the strategic value of that? Is it a long game to provide content for the rumored AR/VR hardware? How is it different or better then offering more types of hardware and more software for its current platforms? Part if it is to prevent the big streaming incumbents from shafting them, but that's also true of other markets and niches too.
The back chamber of the tower is mostly empty space with memory and more air movement. That to me looks to be a vast amount of space for the M1 Max or 2 to sit with heaps of room for a heat spreader and assuming memory will be limited to what is unified with the SOC(s) there should be room for 2 if not more. Wonder if this is how they get a half-height tower. Use a board sandwich like an iPhone to get M's on the back PCIe in the front. The same board Repeats upwards in a full-height tower with a second power supply at the top of the case and a high-speed bridging cable.
What's different this time is there's not going to be GPU upgrades when it uses unified memory and there's not the same kind of upgrade path needed as manufacturers start to hit limits at 3nm and the power at these levels meets the needs of pretty much every workflow today.
There might be a need for PCIe peripherals in some cases but very few and they can go outside the main box through a PCIe connector.
The Mac Pro only reverted to a PCIe enclosure because of PCIe GPUs and those aren't needed any more.
Another option is that they make a dual-mode iMac. The iMac enclosure has enough cooling power to handle 4x M1 Max. If they made a larger iMac like the XDR display, they could sell the same internals in a server blade for people who want it rack-mounted and if people want the headless setup, they can buy the server model too.
The server version would look closer to the Xserve pictured below the Mac Pro here:
It would have metal all around instead of a display and because it's flat, it allows direct access for increasing storage. It wouldn't use an XDR display panel but a display similar to what's in the MBP.
This iMac Pro could start at $1999 with M1 Pro (16GB min), going up to $3099 for M1 Max (32GB min), $4999 for M1 Max Duo (Extreme, 64GB min), $7999 for M1 Max Quad (Ultra, 64GB min). Up to 256GB memory total.
If they get returns and need to refurb, they can turn the iMac into a server model or vice versa.
Long-term, some people would be unhappy that they can't upgrade GPUs but there's no way to allow this anyway and the options should be cheaper so that doing a full upgrade isn't so bad.
Show people why Apple demonstrated a fully loaded Mac Pro w/ 1.5TB DDR4, 28 Core Xeon and peak Duo GPGPU Logic Pro on stage, or at the very least cite what a fully loaded latest offerings Mac Pro performance can do and how far the new MacBook Pro had to go to even be in the ballpark.
Studios by the Mac Pro for music production and post production never mind 3D Modeling and Engineering because that expansion will be viable for the next 7 years and pay for itself tenfold.
This is the antithesis of Appleness. Its buyers want clean minimalism on their desks, so not something with a lot of wires, not something with lot of external boxes, not something with a lot of plugs is something they do not want, and it's something workstation market does not want if the 2013 model is anything to go by. Then, if there is not any dGPU support due to Apple's UMA, that forgoes people who want more than 128 GPU cores in the desktop. The Mac Pro currently offers 4 to 6 dGPUs. They will need to offer a solution for people who just want more. They essentially want to put as much compute power in a box at 1400 Watts as possible.
Regarding dGPUs and UMA, UMA doesn't prevent having dGPUs. Why would it? It would be just like today, where GPU assets have to be moved back and forth from main memory to GPU memory. UMA also doesn't imply memory has to be in-package either. It could all be in DIMM slots. How Apple gets to 1 TB of memory will be interesting to see.
Apple probably would have been more successful with the 2013 Mac Pro model if they updated it, year by year, with more performant CPUs and GPUs, but I bet they knew it wasn't going to sell enough units for many reasons, and stopped development on successive models within a year in favor of the then iMac Pro. And, they had to realign again as they eventually figured out that most of the market didn't want the iMac Pro either and had to pre-announce the iMac Pro 8 months in advance and the Mac Pro 2.5 years in advance.
There better be a book about Apple's Mac decisions from 2011 to 2017. It is absolutely baffling how they could have arrived at the decisions they ended up with, with a lot of decisions in strategic opposition to other decisions. They still are afflicted by those decisions today.
Hey, no problems from me on this, I want more product offerings after all. Remember, they have been free to do this, but have chosen not to. They don't view themselves as a computing company, but more like consumer technology company. So, "big iron" computing isn't something they are going to do, or offer a wide range of computing products even in consumer markets. I wonder what is powering Xcode Cloud though. This defacto limits how many Mac units they can sell because they simply don't offer a product in this or that niche, and just as important, work to get software for those niches.
I hope they have a big goal of market permanence, where, if someone buys a Mac, they don't have to give a second thought on whether it will have the software they need to do whatever they want to do. That's not true today. I don't think grabbing a big chunk of the consumer market will help with various niche software, and it leaves them under threat.
The Mac mini needs to have M1 Pro and M1 Max SoC options. The iMac 24 needs to have an M1 Pro option at least. The large iMac needs to have M1 Pro, M1 Max and a 16 p-core + 64 GPU core SoC options, and the Mac Pro is needs have 16+64 and 32+128 options, with rack mount and PCIe slots. Then, they need to have an MBA15 at $1600 or so, and a modular box at $2000. Still think a $800 A15 powered 13" laptop is a good idea. $1000 and $2000 Apple monitors will hopefully be announced sooner rather than later.
The easiest route would be to cut the 2019 model in half and leave some slots but I don't see them having PCIe GPU support. This machine was also designed around Intel/AMD parts and the setup has changed completely with Apple Silicon.
I don't think they would do unified memory via DIMMs because it compromises the memory bandwidth. The Mac Pro will have in the region of 1TB+/s memory bandwidth, I don't see them achieving this with plug-in modules. Nvidia used their own NVLink to get around PCIe limits in their servers connecting memory to the GPUs. With such fast SSDs, the need for larger capacities like 1TB of memory is reduced, the use cases for over 256GB are very few. 64GB HBM per chip x4 chips would be enough.
I'd say the decisions made around the Mac Pro over the years make a lot of sense. The first thing is that the larger form factor wasn't selling well in 2012 and hadn't been selling well for a while, they had plans to cancel it way before 2012. The more expensive a computer is, the less frequent the upgrade cycle is. People hang onto them for 10 years. So a radical design and marketing campaign was tried to boost sales in 2013 but with no way to upgrade the GPUs and the higher entry price point (it used to start at $2199-2499), it only appealed to a portion of an already small and diminishing market. The iMac Pro gave them more room to expand thermal capacity (500W vs 300W) but again it had a high price point and no ability to upgrade GPUs. I believe they also had some input/pressure from some of their pro users, likely in the film industry to just make another box like before. While the 2019 model satisfies this group, it cuts out a bigger portion of the much larger volume of buyers at a lower price point.
Today things are different. The latest Macbook Pro is beating the Mac Pro in some workflows. Marques Brownlee tested it out against his $50k Mac Pro (16:50):
He said he used to cart around a 27" iMac to be able to handle the jobs and would now be able to do it on a MBP and on battery. This is for editing and exporting 4K/8K footage. Raw power is the main thing that people have needed and it's not a need that will keep increasing for every use case. There's very few use cases that a maxed out 2019 Mac Pro wouldn't handle even in 10 years time and the same will apply to an Apple Silicon Mac Pro with integrated graphics running at 40TFLOPs+.
I would also like to see some of those kind of Mac models suggested there, especially the larger Air, they used to have 12" and 14" Macbooks. An $800 Air would boost sales a bit but I wonder how they'd hit that price point without reducing storage/RAM. I don't think the M1/A15 chip would make that price difference. They sell an education Air for $899 by lowering margins. The entry iPad is $479 with 256GB storage and 4GB RAM and the Air is $749 so there's probably a way to make a cheaper laptop while maintaining margins. I think they just don't want to chase the lower price points in the PC space.
The more that computers develop, the need for modular computers is going away. Modular over the years has meant RAM, storage and PCIe GPUs. RAM makes more sense being integrated, storage is nice to be able to upgrade and one of the only parts left that can fail with use. Integrated graphics are the way forward.
The classical Geekbench benchmark uses the CPU cores of which there are 10 in the M1 MAX (8 performance cores + 2 efficiency cores).
Being able to share MPX modules, both 2-wide and 4-wide varieties, with the 2019 Mac Pro will increase the amount and kinds of MPX modules available for both models. And, either there is an Apple Silicon version of the 2019 Mac Pro model, or that find a way to rack the half size model.
It seems all but inevitable for them to come up with a solution for people who want more GPU performance on a desk. Whether it is through PCIe slots or clustering, they will need to offer a solution. The usefulness of internal expansion is important for a rather large subset of the workstation market. It's not just PCIe dGPUs. I love the RAID MPX module. I'd like to see 16 TB, 32 TB SSD PCIe solutions. There are lots of niche audio, accelerator and I/O cards.
So, an Apple Silicon Mac Pro like the 2019 model but half sized, absolutely. A machine like the 2013 model would be a mistake.
It's really quite the mystery what they will do. I hope someone will tear down the 16 GB LPDDR5 modules in the M1 Max/Pro to see how the DRAM chips are arranged. It's non-standard packaging and looks bigger than typical DRAM packages. Maybe it is just two sets of 8 GB LPDDR5 chips, which are in mass production, inside the package. I don't think the 16 GB LPDRR5 packages are in mass production yet, so curious what they are doing with those RAM packages.
If they have in-package memory like the M1 Max, they probably will have to come up with a custom stacking solution to get to 1 TB capacities. 4 M1 Max SoCs in a package would make for 16 LPDDR5 packages, with each package having a stack of up to 4 LPDDR packages. 64 LPDDR5 packages in 16 stacks for 1.6 TB/s of bandwidth. Uh, sounds sporty to have package with 16 packages/SoCs insides. 16 channels of DDR5, only 800 GB/s, in 16 DIMM slots doesn't sound great either, but they'll be able to achieve 2 TB memory capacity.
We will have to wait and see whether they and customers determine if 256 GB of HBM is good enough. For a half sized box, probably enough, given that competitor boxes of the size will be about the same. For a 2019 Mac Pro sized box, probably not, especially rack mount versions, as those competitor boxes will have 1+ TB of RAM capacity.
The 2013 Mac Pro has a 450 W power supply, just a little bit less than the 2017 iMac Pro. Most of the extra 50 W in the iMac Pro went to powering its display, and the two have about the same power envelope for the computing bits.
For the Mac Pro, I see it as 2 major mistakes. Not only did they make a mistake with the 2013 model, they made it again with the iMac Pro. After 6 years from late 2013 to late 2019, Apple rounded back to the beginning with a big desktop box with modular components. We know they realized the mistakes after 4 years as they had that Mac Pro PR tete-e-tete with journalists in 2017, where they discuss their pro Mac plans. Since when does Apple ever talk about unannounced products?
If it is a diminishing market, they wouldn't be selling a 2019 Mac Pro, right? It seems more that they didn't understand the workstation market, outside of a narrow video editor one. They didn't anticipate machine learning, I think they gave up on CAD and 3D in that 2012 to 2015 period. Computational codes (anything involving large sets of differential equations) didn't interest them that much? There is a market out there, where computational requirements are basically infinite, so my expectation is the workstation market is here to stay for the foreseeable future. And, Apple will need to race to get computational codes to use its Metal API to even have a chance at all these potential buyers. UMA on Apple Silicon is a big advantage, giving these codes 20, 30, 40 GB of memory, and a highly threaded GPU, with presumably, 100 to 200 GB of memory in the future. The time to push to get software on Metal is now.
They also trot out the "Pro workflow team" term in its statements probably as way to assure customers that they know what they are doing now. I guess that is ok, but if you look at it glass half empty, it means Apple's marketing team is so bereft of high end computing knowledge that they need a team of "pros" to tell them what makes a good high end computer. Hopefully it's the former (assurance), not the latter (absence of knowledge).
This is just the Mac Pro! Every Mac product had issues from 2012 to 2019. Something went wrong with the marketing team or the decision makers for Macs in that time frame. Not having a a successor to the Thunderbolt display was also a crazy pants decision. Someone has got be dying to write a book.
It's great that the 2021 MBP machines have Mac Pro performance, but that doesn't mean it ends there. It only means people want 2x to 10x the performance of the 2021 MBP in a desktop box. My perspective is that there are a lot of use cases that have continually increasing compute needs, and therefore, those customers will continually want the most compute performance that can be placed in a desktop box with a 1000 W of power. Like I said above, Apple has to sponsor a lot of codes to convert from CPU compute to GPU compute (Metal), than have Metal hardware capable of running an ecosystem high performance computing codes.
The iPad Air has an A15, 4 GB RAM, 64 GB NAND and 11" display at $600. The iPad Pro has an M1, 8 GB RAM, 128 GB NAND and 11" display for $800. It's not a stretch for a MBA13 with A15, 8 GB RAM, 256 GB NAND and 13" display to cost $800 including desired margins, especially in 2022.
I think the time is now to have an $800 laptop model, with Edu discount to $700. Trying to sell more units downmarket also means they have to get a self sustaining games ecosystem for macOS. Getting more units in the market is vitally important to establishing that gaming ecosystem, so a cheaper Mac is a necessary first step. A Mac mini with M1 Pro for $800 and with an M1 Max for $1000 would be part of a goal of getting more units out there. They will need to buy a few game developers, and become a publisher. On the pro software, easiest path is probably like what they did with Blender, sponsor more open source software, get a lot of these pro apps on Mac. A first class Matlab implementation. Creo. Computational codes. Possibly ship their own set of pro software for STEM fields, etc.