How about something new and exciting from Apple? No, I don't know what that would be, maybe
- Folding iPhone (multiple folding feature)
Pretty sure that's coming, but as always they'll take their time to get it right. Except, unless their AR/VR headsets are their idea of the display of the future.
- Screens that you can use your fingers on
Yeah they released that in 2010. It's called iPad
- Cool tech gear like that being shown on "Expanse"
I'd like to comment on that but that show is still only on my "to watch but haven't watched yet" list.
- How about a cool iPad or laptop that the keyboard slide down exposing another screen under it, now giving you a screen that is twice as big. And the screen automatically knowns if you have slide the keyboard down, and expands the screen or makes it standard size as needed when pushed back up
That sounds intriguing. Although I think Apple really aren't fans of that many moving parts.
I really think Apple's ideas for the displays of the future are Macs that don't have displays at all. Imagine a laptop that is only the bottom half of today's laptops, with no display.
Instead you're wearing their AR headset. It's been slimmed down to something no more obtrusive than any normal glasses/sunglasses of today. It has incredibly high resolution (both density and number of pixels) transparent displays in the lenses. And it knows where your no-screen laptop is** and projects a virtual display over that. Of course it can adjust the transparency of the pixels in the lenses so that for all intents and purposes it can appear to you exactly like today's laptop displays do. Or even better... how about a MacBook Air sized no-screen laptop, with one, or any number of, virtual displays any size you like (from say 10 inches up to the equivalent of a 32" XDR hovering in the air wherever you want it -- or more).
The only thing this doesn't cater to, at least in the short term, is when you want to show it to someone(s) else... at least until Apple's offering here is one of at least a few, and we're all wearing these things, similar to how not having a smartphone these days is extremely unusual. And eventually they'll shrink it down to contact lens size.
Where I mentioned pixel density, it doesn't need to be incredibly high resolution throughout, since our peripheral vision is pretty low "resolution". Looking at my 5K display right now, the retina resolution is great, but at any given moment I'm only seeing retina resolution for a few degrees of angular vision around the point on the display where my eyes and head are actually pointing at. So something that's high resolution in an area centered on the lens, and lower resolution round the edges would be plenty.
How might this replicate the iPad experience? We could still have a device that might be little more than a 13", 11", 8", whatever Magic Trackpad, with the eyewear projecting a virtual display onto it. Maybe there's even some potential for a smaller touchpad device to have a larger virtual display, where the physical touch device and the virtual display move around each other to allow physical touch anywhere on the virtual display. Then again, maybe we don't need the touch device at all -- instead just pressing on my other hand or wrist or even pressing thumb and fingers together in different ways or whatever. It's right up Apple's alley to come up with new yet super intuitive ways to interact with our devices.
So yeah, to "how about something new and exciting from Apple...?" I'd be willing to bet they're heading down a road something like that.
----
**with standard keyboard, trackpad input until they figure out how to read our brain waves and think our input. Don't scoff. There's at least one neurologist I'm aware of who is making some serious headway into that. It's no more than a few years away from being a reality, and not long after that coming to market.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro. And even then Apple won't provide anything standard M.2, any more than they do the current MP.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro.
They might. Plenty of people said Apple would never bring HDMI or SD back to the MacBook. Plenty of people said Apple wouldn't make their machines user serviceable again. Apple have been doing plenty of u-turns recently.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro.
They might. Plenty of people said Apple would never bring HDMI or SD back to the MacBook. Plenty of people said Apple wouldn't make their machines user serviceable again. Apple have been doing plenty of u-turns recently.
That's a good point. I feel pretty strongly about this one, but now I'm asking myself why. I feel like my own personal track record for predicting these things (just conceptually from what I know of Apple's history etc., no real insights) is usually pretty good. And in this case I honestly can't see it.
But I admit, I was one of those "plenty of people" -- I really couldn't see them bringing those ports back and they did. So I guess you're right -- who knows at this point.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro.
They might. Plenty of people said Apple would never bring HDMI or SD back to the MacBook. Plenty of people said Apple wouldn't make their machines user serviceable again. Apple have been doing plenty of u-turns recently.
That's a good point. I feel pretty strongly about this one, but now I'm asking myself why. I feel like my own personal track record for predicting these things (just conceptually from what I know of Apple's history etc., no real insights) is usually pretty good. And in this case I honestly can't see it.
But I admit, I was one of those "plenty of people" -- I really couldn't see them bringing those ports back and they did. So I guess you're right -- who knows at this point.
One things for certain, they certainly won't add m2 unless people are asking for it. That's why I find lkrupp's attempts to shut down any discussion (I honestly have no idea what they think message boards are for, they seem to hate every conversation) with snark so objectionable. If no one talks about what they want then nothing gets better.
For the record, I'd be very surprised if they added m2 slots. I don't think it's at all likely to happen. But they might, and I'd like it very much if they did.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro.
They might. Plenty of people said Apple would never bring HDMI or SD back to the MacBook. Plenty of people said Apple wouldn't make their machines user serviceable again. Apple have been doing plenty of u-turns recently.
That's a good point. I feel pretty strongly about this one, but now I'm asking myself why. I feel like my own personal track record for predicting these things (just conceptually from what I know of Apple's history etc., no real insights) is usually pretty good. And in this case I honestly can't see it.
But I admit, I was one of those "plenty of people" -- I really couldn't see them bringing those ports back and they did. So I guess you're right -- who knows at this point.
One things for certain, they certainly won't add m2 unless people are asking for it. That's why I find lkrupp's attempts to shut down any discussion (I honestly have no idea what they think message boards are for, they seem to hate every conversation) with snark so objectionable. If no one talks about what they want then nothing gets better.
For the record, I'd be very surprised if they added m2 slots. I don't think it's at all likely to happen. But they might, and I'd like it very much if they did.
If they do add them to a future MBP, I hope they'll add two, and not compromise them in any way, so they can be striped. Internally I care about performance. I can accommodate redundancy externally.
I gave it some more thought...
To add them is more than just whacking a couple of slots on a board, obviously. Among other things, there's PCIe lanes at least to consider and how those are distributed across, well, almost everything now. (I believe the SD card and the HDMI aren't the latest specs because of PCIe limitations..?)
So it's a significant design decision. And therefore one reason I really think they won't is because if they were ever going to anytime soon -- at least for MacBook Pros (where arguably they'd be the most useful, if they are in any Mac) -- then they would have done so already (while they made, as you say, a significant number of other reversals).
Perhaps future M chips will have more lanes or whatever else is needed.
[…] if you're mentioning eGPU because you fear a new, smaller, ASi Mac Pro won't be as expandable, I'll suggest otherwise. Two reasons:
1. Less expandability required:
Today's most powerful Mac Pro graphics cards take up half or more of the 2019 Mac Pro's expansion capabilities.
I know people have doubts -- but I don't, after what they were able to squeeze into the new MBP's -- that even the Mac Pro version of ASi is going to be integrated (SoC) graphics with enormous amounts of unified memory, and insanely powerful (and efficient) CPU and GPU. They'll keep up with, if not run rings around, the fastest PCIe GPUs. In fact I suspect Apple won't even support third party GPUs in an ASi Mac Pro. […]
That means we don't need as many PCIe slots in an ASi Mac Pro to still be as expandable as the 2019 Mac Pro.
[…] Apple really surprised us with how insanely expandable the 2019 Mac Pro was, and restored our faith in the idea that they're not exclusively all about being so tightly integrated that you can't upgrade anything. So with any luck they'll maintain that, and if so -- if there's enough demand for RAM and GPU upgrades in an ASi Mac Pro, and if Apple can acknowledge that -- then I'll wager we'll be able to swap out the entire SoC, meaning we'll be able to upgrade RAM, GPU and now even CPU, albeit all at once, not one piece at a time. I seriously doubt they're going to compromise a future Mac Pro with anything less than their SoC/unified design now. […]
2. Less space required
Let's not underestimate what they can likely do with a half-size Mac Pro. A lot of the space in the 2019 Mac Pro is thermals, most of which isn't necessary with ASi. Less thermals required, less space required.
The differences in both performance and thermal issues between the last generation Intel MBPs and these new ASi ones speaks loudly to that -- especially how the 14 can be specced to the same level as the 16, […]
I suspect they're going to deliver a box that's significantly smaller than the 2019 Mac Pro box, but still has as much expandability, simply because there's just so much more efficiency, etc. in ASi.
This vision for the ASi modular Mac Pro is reasonable and compelling.
The basic concept brings to mind Apple’s original OEM upgrades: the “Macintosh 512K Memory Expansion Kit,” which simply replaced the entire logic board in a Macintosh 128K with a new one with 512K. No mess, no fuss. Later, the “Macintosh Plus 1 Mb Logic Board Kit” did the same thing and was just as easy to install, although to get to a Mac Plus you also needed to replace the disk drive, via the “Macintosh Plus Disk Drive Kit” …
I have been thinking of waiting for the second gen M-series iPad Pros. But if they aren't going to arrive until fall, or even sprint '23, I think I'm going to have to jump.
I'm holding out. Thought I would get a 2021 iPP12.9, but I'm going to stretch. If there is an M2 MBA in Spring/Summer, I'm betting an M2 iPP will arrive in Fall of 2022. The home button on my iPP10.5 is getting flaky though.
I didn’t. My new iPP should be here at the end of the week along with an Apple Pencil 2 and a Brydge keyboard case.
Congrats! I have couple of family members that got the 2021 iPP12.9, and it is a very very nice machine. Just have to stretch another year.
I suspect they're going to deliver a box that's significantly smaller than the 2019 Mac Pro box, but still has as much expandability, simply because there's just so much more efficiency, etc. in ASi.
If the Apple Silicon Mac is half the size of the 2019 Mac Pro, by physical constraints, it won't have the same expandability. It's really not that big because of its cooling system per se. It's that big because it can support a Xeon, probably two socket if needed, and 4 250 W GPUs. Moreover, I think a lot of folks are looking at this backwards. Yes, Apple's compute performance/Watt will be about 2x that of x86 systems. That really shouldn't mean that Apple should make the Mac Pro half the size for comparable performance to x86 systems. It means that they should keep the same size box and offer 2x the compute performance. They aren't competing on the size and lightness of the box. It's all computing metrics.
So, if the M1 Max Duo has 20 TFLOPs of GPU compute (about 130k in GB5 Metal) at 120 W, they should offer a system with 8 of them in a box at 1000 W, or 4 M1 Max Quads at 1000 W. Like with the 2019 Mac Pro, have a system at 1400 W, which is a practical limit for typical 110V 15A circuits (1650 W) in the USA, and have it support as many CPUs and GPUs that can fit in 1400 W, minus margin for other stuff (storage, ports, et al).
1400, 1500 W is basically a high gate for workstation computer systems as the vast majority of power circuits in the USA are at 1700 to 1800 W or so. To get more will require 220/240V circuits, and represents a whole new level of safety concerns and cooling capacity. Not cooling for the box, but cooling for the room that the box is in. x86 workstation systems will be designed for 1000 to 1400 W. That's the top of the line. Year over year, x86 systems will be at that power level, incrementally increasing compute performance. 40 TFLOPS will be more normal in 2022. In a year or two, it's going to be 80 TFLOPS. Apple will need 256 g-core GPUs in 2023, and not only just one, but offer a box that has 2 to 4 of them.
I think the half sized box for the Apple Silicon Mac Pro can work. They would have to actually update them at a yearly cadence. This is absolutely vital. They have to update on a yearly cadence so that there will be enough of a secondary market for people to go through buy-sell-buy/trade cycles where they can get increasing performance year on year, and people can get the systems used at a nice discount. They would have to offer a way to make them rack-able. They need to make accessories for them, like a RAID HDD box, a PCIe breakout box, who knows what else. They can not rely on someone else to do it. Either this, or pay someone to do it for them. They should offer TB5 networking so clustering 2, 3, 4 boxes on a desk can be easy. The boxes should be neatly stackable to minimize footprint.
This begs the question of why go through all this effort, when using the 2019 Mac Pro chassis accomplishes the same thing. Just design a motherboard that is compatible with the existing Mac Pro, like they did with the M1 MBA, MBP, Mac mini.
Since Apple is likely removing the option to add RAM to their systems, how about throwing us a bone and giving us access to a M2 SSD slot or two? SSDs are in rapid development right now. The last one I bought (for my PS5 which does have a M2 slot) was slightly more than 6 GB per second read bandwidth. It would give the Mac a longer use life if it could be upgraded with faster and larger SSDs in the future as they become available and more affordable. M2 SSDs are about as small as a stick of RAM and even easier to install. Most users could do it with a screw driver.
Unlikely. Apple solders storage chips, which have the benefit of being faster and more reliable than slotted components.
They're perfectly able to do that and have m2 slots.
Yeah but they won't. At least not in anything other than a Mac Pro.
They might. Plenty of people said Apple would never bring HDMI or SD back to the MacBook. Plenty of people said Apple wouldn't make their machines user serviceable again. Apple have been doing plenty of u-turns recently.
That's a good point. I feel pretty strongly about this one, but now I'm asking myself why. I feel like my own personal track record for predicting these things (just conceptually from what I know of Apple's history etc., no real insights) is usually pretty good. And in this case I honestly can't see it.
But I admit, I was one of those "plenty of people" -- I really couldn't see them bringing those ports back and they did. So I guess you're right -- who knows at this point.
Since Apple has their own controllers, a complete SSD is out of the question, so if there's an M.2 slots, it won't be compatible, which is pointless.
Someone could argue for a removable module, while this could work, I don't think most will find it "cheap" or"easy". It's a proprietary solution with tons of firmware protection, plus that hefty OEM price. Kinda defeats the purpose.
I don't quite get the rumors of a sort of Mac mini Pro that will replace the 2018 space-gray mini. It doesn't make sense to me. It's not unwelcome, but it steps all over the Mac Pro, whatever they do with that. I don't really get it.
The big unknown is the Mac Pro. There really haven't been any solid rumors, other than some mysterious, cutting-edge silicon on TSMC's roadmap, maybe.
There are a couple of routes they can go. The only parts they need are shown here in green:
They only need 4x M1 Max to reach the GPU performance level of the Intel model, 2x for CPU. Any M2/M3 iteration will exceed it.
- 4x M1 Max - up to 256GB unified RAM - some SSD slots - 500W PSU - single large cooling fan
Apple Silicon is faster than the Afterburner card so no need for this any more. No external GPUs so that takes out the AMD cards. Unified memory so I don't see them offering memory slots.
That won't fit into a Mac mini as it needs a much bigger power supply. It would fit into an enclosure around the size of the 2013 Mac Pro but it can use the 2019 design or the iMac Pro or a new mini tower design.
Apple knows from the 2019 model what spec people have been buying, they only need to match this.
One route they could go is to top out Apple Silicon at the M1 Max Duo. This would be equivalent to the current 28-core Mac Pro with a W6900X, which with the Afterburner card costs $20,599. In an iMac Pro, I'd estimate this model to be priced at $4999. By the time it gets to 2nm in 2025, the Max Duo chips would be around 3x faster again. Apple can easily keep selling the current Intel Mac Pro for a couple of years, EOL it in 2023 and offer maintenance and support for 7 years after that.
If Apple offered an M1 Max Quad, I'd expect it to be priced around $7,999 and it could either be in an iMac Pro or mini tower. I'm not sure that the unit volume will justify making these at all. It's tens of thousands of units vs millions for every other model and it's not 100x the profit. I would like to see either the Cylinder form factor again paired with 32" XDR or 32" iMac Pro.
One thing is for sure, no more than 4 dies. That means 32+8 for M1 and 48+16 for M3. Not the 96- or 128-Core Threadripper level, if there's still any. Don't be betting that ASi will top out every x86 chip, that's not their goal (Unless Apple somehow makes their own server chips dedicated to internal use).
@tht said: Detnator said: If the Apple Silicon Mac is half the size of the 2019 Mac Pro, by physical constraints, it won't have the same expandability. It's really not that big because of its cooling system per se. It's that big because it can support a Xeon, probably two socket if needed, and 4 250 W GPUs. Moreover, I think a lot of folks are looking at this backwards. Yes, Apple's compute performance/Watt will be about 2x that of x86 systems. That really shouldn't mean that Apple should make the Mac Pro half the size for comparable performance to x86 systems. It means that they should keep the same size box and offer 2x the compute performance. They aren't competing on the size and lightness of the box. It's all computing metrics. So, if the M1 Max Duo has 20 TFLOPs of GPU compute (about 130k in GB5 Metal) at 120 W, they should offer a system with 8 of them in a box at 1000 W, or 4 M1 Max Quads at 1000 W. Like with the 2019 Mac Pro, have a system at 1400 W, which is a practical limit for typical 110V 15A circuits (1650 W) in the USA, and have it support as many CPUs and GPUs that can fit in 1400 W, minus margin for other stuff (storage, ports, et al). 1400, 1500 W is basically a high gate for workstation computer systems as the vast majority of power circuits in the USA are at 1700 to 1800 W or so. To get more will require 220/240V circuits, and represents a whole new level of safety concerns and cooling capacity. Not cooling for the box, but cooling for the room that the box is in. x86 workstation systems will be designed for 1000 to 1400 W. That's the top of the line. Year over year, x86 systems will be at that power level, incrementally increasing compute performance. 40 TFLOPS will be more normal in 2022. In a year or two, it's going to be 80 TFLOPS. Apple will need 256 g-core GPUs in 2023, and not only just one, but offer a box that has 2 to 4 of them. I think the half sized box for the Apple Silicon Mac Pro can work. They would have to actually update them at a yearly cadence. This is absolutely vital. They have to update on a yearly cadence so that there will be enough of a secondary market for people to go through buy-sell-buy/trade cycles where they can get increasing performance year on year, and people can get the systems used at a nice discount. They would have to offer a way to make them rack-able. They need to make accessories for them, like a RAID HDD box, a PCIe breakout box, who knows what else. They can not rely on someone else to do it. Either this, or pay someone to do it for them. They should offer TB5 networking so clustering 2, 3, 4 boxes on a desk can be easy. The boxes should be neatly stackable to minimize footprint. This begs the question of why go through all this effort, when using the 2019 Mac Pro chassis accomplishes the same thing. Just design a motherboard that is compatible with the existing Mac Pro, like they did with the M1 MBA, MBP, Mac mini.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
It's already been said. Desktop computers have a high gate to cross at around 1500 Watts power draw. Most of the USA's power circuits in offices and desks are going to be limited by 1650 Watt power circuits. The 2019 Mac Pro uses a 1400 W power supply. So, design an Apple Silicon desktop around 1200 to 1400 W, and fill in the chips accordingly. Have it idle at less than 50 W, and at full power draw, around 1000 to 1200 W. Don't leave performance off the desk in the workstation Mac.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
It's already been said. Desktop computers have a high gate to cross at around 1500 Watts power draw. Most of the USA's power circuits in offices and desks are going to be limited by 1650 Watt power circuits. The 2019 Mac Pro uses a 1400 W power supply. So, design an Apple Silicon desktop around 1200 to 1400 W, and fill in the chips accordingly. Have it idle at less than 50 W, and at full power draw, around 1000 to 1200 W. Don't leave performance off the desk in the workstation Mac.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
There’s a point where extra performance doesn’t matter. You will get something twice as expensive and hot with marginal gain. A 4-die is plenty to compete with a 56-core Xeon (assuming 48+16 M3) and Zen4 alike, most likely much better. There’s no law that states a workstation has to match 1,500 watt, which is overkill for an Apple silicon system. Apple is no AMD, they don’t need to build a show piece just to claim a title. We know what the M1 can do.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
It's already been said. Desktop computers have a high gate to cross at around 1500 Watts power draw. Most of the USA's power circuits in offices and desks are going to be limited by 1650 Watt power circuits. The 2019 Mac Pro uses a 1400 W power supply. So, design an Apple Silicon desktop around 1200 to 1400 W, and fill in the chips accordingly. Have it idle at less than 50 W, and at full power draw, around 1000 to 1200 W. Don't leave performance off the desk in the workstation Mac.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
There’s a point where extra performance doesn’t matter. You will get something twice as expensive and hot with marginal gain. A 4-die is plenty to compete with a 56-core Xeon (assuming 48+16 M3) and Zen4 alike, most likely much better. There’s no law that states a workstation has to match 1,500 watt, which is overkill for an Apple silicon system. Apple is no AMD, they don’t need to build a show piece just to claim a title. We know what the M1 can do.
We are getting to a point where the performance doesn't matter so much. There was someone here editing 12k raw footage on the entry M1 chip:
M1 Max is 2x CPU/4x GPU faster than this and M1 Max Quad would be 4x again. For some use cases, the power can always be more but usually these use cases are offline rendering/computing, which can be done with multiple machines.
Apple also has the option to boost clock speeds in the Mac Pro so it could be M1 Max Quad but running at 25-50% higher clocks for CPU/GPU so effectively up to 6x faster than M1 Max.
There are some things that will inherently stop needing improvement like 64-bit computing, 10-bit displays, we probably won't go beyond 8k resolution for authoring, maybe 16k in some cases. Storage is at 8TB internally, while there is some need for more, the mass market needs are well below this, same with 64GB RAM. An M1 Max Quad with 256GB RAM is close to the end-game needs for a personal workstation and it still has room to improve 3x going to 2nm.
For Apple it would make more sense to sell a 40-60TFLOP M1 Max Quad on 5nm, then a 55-80TFLOP M2 Max Quad on 4nm, 70-100TFLOP M3 on 3nm etc. Nvidia/AMD are at least a year behind Apple and Intel is planning to catch up in 2025.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
It's already been said. Desktop computers have a high gate to cross at around 1500 Watts power draw. Most of the USA's power circuits in offices and desks are going to be limited by 1650 Watt power circuits. The 2019 Mac Pro uses a 1400 W power supply. So, design an Apple Silicon desktop around 1200 to 1400 W, and fill in the chips accordingly. Have it idle at less than 50 W, and at full power draw, around 1000 to 1200 W. Don't leave performance off the desk in the workstation Mac.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
There’s a point where extra performance doesn’t matter. You will get something twice as expensive and hot with marginal gain. A 4-die is plenty to compete with a 56-core Xeon (assuming 48+16 M3) and Zen4 alike, most likely much better. There’s no law that states a workstation has to match 1,500 watt, which is overkill for an Apple silicon system. Apple is no AMD, they don’t need to build a show piece just to claim a title. We know what the M1 can do.
We are getting to a point where the performance doesn't matter so much. There was someone here editing 12k raw footage on the entry M1 chip:
M1 Max is 2x CPU/4x GPU faster than this and M1 Max Quad would be 4x again. For some use cases, the power can always be more but usually these use cases are offline rendering/computing, which can be done with multiple machines.
Apple also has the option to boost clock speeds in the Mac Pro so it could be M1 Max Quad but running at 25-50% higher clocks for CPU/GPU so effectively up to 6x faster than M1 Max.
There are some things that will inherently stop needing improvement like 64-bit computing, 10-bit displays, we probably won't go beyond 8k resolution for authoring, maybe 16k in some cases. Storage is at 8TB internally, while there is some need for more, the mass market needs are well below this, same with 64GB RAM. An M1 Max Quad with 256GB RAM is close to the end-game needs for a personal workstation and it still has room to improve 3x going to 2nm.
For Apple it would make more sense to sell a 40-60TFLOP M1 Max Quad on 5nm, then a 55-80TFLOP M2 Max Quad on 4nm, 70-100TFLOP M3 on 3nm etc. Nvidia/AMD are at least a year behind Apple and Intel is planning to catch up in 2025.
The real issue is that workstation is changing, as GPU or even HSA acceleration are having more roles to play, what used to rely on CPU cores are moving away, hence why Intel never competes for something like the 3990X/5990WX. In fact, I believe the latter is still delayed. Whatever the M3 quad is competing, will be where performance is just at the right amount. AMD can still build some 96- or even 128-core TR just to claim that "performance leader" crown, sure, they need that to sell their hardware. Apple only build chips to serve products that need the most.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
They currently sell such a system in the 2019 Mac Pro. It has room for a 300W CPU, 4 250W GPUs, 2 3.5" HDD, 4 1-wide PCI cards, and over 1 TB of memory. It allows for 10 3.5" HDD drives to be put inside, PCIe storage at all price points, additional ports, capture cards, so on and so forth. The two NAND storage cards are not worth mentioning for a box size argument. So, there is a value in it, even at its $6000 starting price. The modularity is a feature a lot of buyers pay for even if they don't do much with it.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
Okay, there's no such thing as "the most performance", we all need to make that decision. I find it ironic that people who say that are always the 1st to make that decision, because faster computers always exist until you hit TOP100, well then where are you going to stop? I guess we can define that everything less than TianHe 1 is non-pro? That's nuts.
It's already been said. Desktop computers have a high gate to cross at around 1500 Watts power draw. Most of the USA's power circuits in offices and desks are going to be limited by 1650 Watt power circuits. The 2019 Mac Pro uses a 1400 W power supply. So, design an Apple Silicon desktop around 1200 to 1400 W, and fill in the chips accordingly. Have it idle at less than 50 W, and at full power draw, around 1000 to 1200 W. Don't leave performance off the desk in the workstation Mac.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
There’s a point where extra performance doesn’t matter. You will get something twice as expensive and hot with marginal gain. A 4-die is plenty to compete with a 56-core Xeon (assuming 48+16 M3) and Zen4 alike, most likely much better. There’s no law that states a workstation has to match 1,500 watt, which is overkill for an Apple silicon system. Apple is no AMD, they don’t need to build a show piece just to claim a title. We know what the M1 can do.
There are applications that can make use of the compute performance, especially GPU compute applications. Basically any STEM fields using numerical methods (3D, CFD et al) have infinite compute requirements. If what used to be jobs needed to be done on a compute cluster can be done on a desktop, it will increase efficiency and quality of work by giving analysts more time and data to assess their problems. It's a continuum of classes of problems whose analytical fidelity only improves with more compute power.
There aren't any laws about 1500 W, but it's a practical limit for how much power is available to desks in offices, whether at work or at home. Workstation vendors design to it be becausd it's a practical limit. The Mac Pro is Apple's workstation machine whose job "is to challenge what we think a computer can do and do things that no computer has ever done before, be more and more powerful and capable so that we need a desktop because of its capabilities,” says Schiller. “Because if all it’s doing is competing with the notebook and being thinner and lighter, then it doesn’t need to be.” It's interesting that Schiller made this statement before saying the 2019 Mac Pro will be modular. I would have thought they learned that being modular is part of the job of a big desktop. It's not just compute performance. Being able to put 50+ TB of storage into the box is part of it. Being able to run a PCIe card for some job that Apple didn't anticipate is part of it.
There are trades regarding whether a user should just buy some compute hardware in the cloud or have it in a server room and run their jobs from a thin client, but it isn't a binary one way of the other. Some are fine with a thin client with a server room. Some really need all the compute and storage in a single box that they can get. Apple really needs a continuum of machines, more than they have today.
Comments
Pretty sure that's coming, but as always they'll take their time to get it right. Except, unless their AR/VR headsets are their idea of the display of the future.
Yeah they released that in 2010. It's called iPad
I'd like to comment on that but that show is still only on my "to watch but haven't watched yet" list.
That sounds intriguing. Although I think Apple really aren't fans of that many moving parts.
I really think Apple's ideas for the displays of the future are Macs that don't have displays at all. Imagine a laptop that is only the bottom half of today's laptops, with no display.
Instead you're wearing their AR headset. It's been slimmed down to something no more obtrusive than any normal glasses/sunglasses of today. It has incredibly high resolution (both density and number of pixels) transparent displays in the lenses. And it knows where your no-screen laptop is** and projects a virtual display over that. Of course it can adjust the transparency of the pixels in the lenses so that for all intents and purposes it can appear to you exactly like today's laptop displays do. Or even better... how about a MacBook Air sized no-screen laptop, with one, or any number of, virtual displays any size you like (from say 10 inches up to the equivalent of a 32" XDR hovering in the air wherever you want it -- or more).
The only thing this doesn't cater to, at least in the short term, is when you want to show it to someone(s) else... at least until Apple's offering here is one of at least a few, and we're all wearing these things, similar to how not having a smartphone these days is extremely unusual. And eventually they'll shrink it down to contact lens size.
Where I mentioned pixel density, it doesn't need to be incredibly high resolution throughout, since our peripheral vision is pretty low "resolution". Looking at my 5K display right now, the retina resolution is great, but at any given moment I'm only seeing retina resolution for a few degrees of angular vision around the point on the display where my eyes and head are actually pointing at. So something that's high resolution in an area centered on the lens, and lower resolution round the edges would be plenty.
How might this replicate the iPad experience? We could still have a device that might be little more than a 13", 11", 8", whatever Magic Trackpad, with the eyewear projecting a virtual display onto it. Maybe there's even some potential for a smaller touchpad device to have a larger virtual display, where the physical touch device and the virtual display move around each other to allow physical touch anywhere on the virtual display. Then again, maybe we don't need the touch device at all -- instead just pressing on my other hand or wrist or even pressing thumb and fingers together in different ways or whatever. It's right up Apple's alley to come up with new yet super intuitive ways to interact with our devices.
So yeah, to "how about something new and exciting from Apple...?" I'd be willing to bet they're heading down a road something like that.
----
**with standard keyboard, trackpad input until they figure out how to read our brain waves and think our input. Don't scoff. There's at least one neurologist I'm aware of who is making some serious headway into that. It's no more than a few years away from being a reality, and not long after that coming to market.
Yeah but they won't. At least not in anything other than a Mac Pro. And even then Apple won't provide anything standard M.2, any more than they do the current MP.
Third party M.2 in a Mac Pro certainly possible via PCIe. Eg. You'll almost certainly still be able to put this insane (in a good way) thing in it: https://eshop.macsales.com/shop/ssd/owc-accelsior-8m2
But I admit, I was one of those "plenty of people" -- I really couldn't see them bringing those ports back and they did. So I guess you're right -- who knows at this point.
For the record, I'd be very surprised if they added m2 slots. I don't think it's at all likely to happen. But they might, and I'd like it very much if they did.
I gave it some more thought...
To add them is more than just whacking a couple of slots on a board, obviously. Among other things, there's PCIe lanes at least to consider and how those are distributed across, well, almost everything now. (I believe the SD card and the HDMI aren't the latest specs because of PCIe limitations..?)
So it's a significant design decision. And therefore one reason I really think they won't is because if they were ever going to anytime soon -- at least for MacBook Pros (where arguably they'd be the most useful, if they are in any Mac) -- then they would have done so already (while they made, as you say, a significant number of other reversals).
Perhaps future M chips will have more lanes or whatever else is needed.
The basic concept brings to mind Apple’s original OEM upgrades: the “Macintosh 512K Memory Expansion Kit,” which simply replaced the entire logic board in a Macintosh 128K with a new one with 512K. No mess, no fuss. Later, the “Macintosh Plus 1 Mb Logic Board Kit” did the same thing and was just as easy to install, although to get to a Mac Plus you also needed to replace the disk drive, via the “Macintosh Plus Disk Drive Kit” …
Everything old is new again!
So, if the M1 Max Duo has 20 TFLOPs of GPU compute (about 130k in GB5 Metal) at 120 W, they should offer a system with 8 of them in a box at 1000 W, or 4 M1 Max Quads at 1000 W. Like with the 2019 Mac Pro, have a system at 1400 W, which is a practical limit for typical 110V 15A circuits (1650 W) in the USA, and have it support as many CPUs and GPUs that can fit in 1400 W, minus margin for other stuff (storage, ports, et al).
1400, 1500 W is basically a high gate for workstation computer systems as the vast majority of power circuits in the USA are at 1700 to 1800 W or so. To get more will require 220/240V circuits, and represents a whole new level of safety concerns and cooling capacity. Not cooling for the box, but cooling for the room that the box is in. x86 workstation systems will be designed for 1000 to 1400 W. That's the top of the line. Year over year, x86 systems will be at that power level, incrementally increasing compute performance. 40 TFLOPS will be more normal in 2022. In a year or two, it's going to be 80 TFLOPS. Apple will need 256 g-core GPUs in 2023, and not only just one, but offer a box that has 2 to 4 of them.
I think the half sized box for the Apple Silicon Mac Pro can work. They would have to actually update them at a yearly cadence. This is absolutely vital. They have to update on a yearly cadence so that there will be enough of a secondary market for people to go through buy-sell-buy/trade cycles where they can get increasing performance year on year, and people can get the systems used at a nice discount. They would have to offer a way to make them rack-able. They need to make accessories for them, like a RAID HDD box, a PCIe breakout box, who knows what else. They can not rely on someone else to do it. Either this, or pay someone to do it for them. They should offer TB5 networking so clustering 2, 3, 4 boxes on a desk can be easy. The boxes should be neatly stackable to minimize footprint.
This begs the question of why go through all this effort, when using the 2019 Mac Pro chassis accomplishes the same thing. Just design a motherboard that is compatible with the existing Mac Pro, like they did with the M1 MBA, MBP, Mac mini.
Someone could argue for a removable module, while this could work, I don't think most will find it "cheap" or"easy". It's a proprietary solution with tons of firmware protection, plus that hefty OEM price. Kinda defeats the purpose.
And such a system will be humongous, costing an arm & a leg and no software that can fully utilize it. It won't sell. That's the same story for the 64-core Threadripper. Every "top-end" hardware is more or less a showpiece, more of "we did it!" rather than being practical, hence why there are no Xeon workstations to follow.
A 4-die M3 will challenge them at the practical level, likely offering more performance & less heat, that's all.
Then, Apple really haven't demonstrated that they will have lower prices with Apple Silicon. They are using all the same price tiers give or take 10%. So this new smaller Mac Pro could start at $6000, and have very little to zero modularity. There's the rumor that Apple will continue on with the 2019 Mac Pro with Xeon W-3300 series processors (up to 38 cores), and presumably new AMD GPUs. That Intel update may result in an even higher price point ($8000 for 12c model?) to make room for an Apple Silicon Mac Pro at $6000. It would be amazing that they would do this as the Xeon W-3300 series uses a different socket compared to the Xeon W-3200 series in the current model. So a whole new Intel motherboard to design and qualify. Odd rumor as they could use all those resources on an Apple Silicon motherboard that could fit in the 2019 Mac Pro.
Lastly, what is a buyer of this smaller Mac Pro really buying if it is a little smaller than a G4 cube? I don't think being quiet and using less energy is high on the list of features for the buyers of this type of machine. They want the most performance, most storage, most flexibility in a box volume that can fit on a desk, and in today's world, in a rack is hugely important too. Intel competitors will be able to match Apple Silicon performance at higher power consumption, and they will have more storage and more modularity, and will be at similar prices. These types of features are more important in this market than being small and quiet. Hard to see how this will not be a repeat of the G4 cube and the 2013 Mac Pro.
x86 workstations are basically limited by these circuits, but they all will edge to the line. They will have 1 or 2-socket Xeon boards with support for 3 to 4 GPUs hitting somewhere between 1200 to 1500 W. And, they will have room for lots of HDD storage and other cards.
M1 Max is 2x CPU/4x GPU faster than this and M1 Max Quad would be 4x again. For some use cases, the power can always be more but usually these use cases are offline rendering/computing, which can be done with multiple machines.
Apple also has the option to boost clock speeds in the Mac Pro so it could be M1 Max Quad but running at 25-50% higher clocks for CPU/GPU so effectively up to 6x faster than M1 Max.
There are some things that will inherently stop needing improvement like 64-bit computing, 10-bit displays, we probably won't go beyond 8k resolution for authoring, maybe 16k in some cases. Storage is at 8TB internally, while there is some need for more, the mass market needs are well below this, same with 64GB RAM. An M1 Max Quad with 256GB RAM is close to the end-game needs for a personal workstation and it still has room to improve 3x going to 2nm.
For Apple it would make more sense to sell a 40-60TFLOP M1 Max Quad on 5nm, then a 55-80TFLOP M2 Max Quad on 4nm, 70-100TFLOP M3 on 3nm etc. Nvidia/AMD are at least a year behind Apple and Intel is planning to catch up in 2025.
There aren't any laws about 1500 W, but it's a practical limit for how much power is available to desks in offices, whether at work or at home. Workstation vendors design to it be becausd it's a practical limit. The Mac Pro is Apple's workstation machine whose job "is to challenge what we think a computer can do and do things that no computer has ever done before, be more and more powerful and capable so that we need a desktop because of its capabilities,” says Schiller. “Because if all it’s doing is competing with the notebook and being thinner and lighter, then it doesn’t need to be.” It's interesting that Schiller made this statement before saying the 2019 Mac Pro will be modular. I would have thought they learned that being modular is part of the job of a big desktop. It's not just compute performance. Being able to put 50+ TB of storage into the box is part of it. Being able to run a PCIe card for some job that Apple didn't anticipate is part of it.
There are trades regarding whether a user should just buy some compute hardware in the cloud or have it in a server room and run their jobs from a thin client, but it isn't a binary one way of the other. Some are fine with a thin client with a server room. Some really need all the compute and storage in a single box that they can get. Apple really needs a continuum of machines, more than they have today.