MP 7,1 What must Apple include in a new Mac Pro that we can’t live without?

[mattspace]

They didn’t have control of the CPU and GPU?

And?

The problem has never been that Apple didn't control things, the problem has always been that the vast, vast majority of consumers, which includes professionals, instinctively do not see compactness as a "pay for it" feature of a desktop system. Nice as a bonus if it doesn't cost you anything in return, sure, but not a feature they buy.

This has been a problem Apple has had throughout its history. It's a deep part of Apple's DNA - the things Apple values in its products, are not the things its customers value in its products, and the only reason this has not bitten them harder, is because the entire company is set up to shield its products from meaningful competition, that would demonstrate this disconnect.

We saw this with the G4 Cube - Apple tried to pitch compactness as a paid feature, which you exchanged for expansion slots, and it bombed catastrophically - eventually it went into stock clearance sales. The Cube was *just* a Powermac without utility, but for the same price as the proper one, as far as consumers are concerned.

The 2013 had the same problem - it gave up all the utility, but kept the price and offered compactness. The result - Apple became a laughing stock within the workstation space.

I never bought the cube or 2013 Mac, but will definitely buy this compact pro if the specs are as predicted.

The only prediction I would bet on, is it's going to cost somewhere between the highest end iMac, and the entry level current Mac Pro, and frankly I wouldn't be surprised if it didn't start with a higher price than the current Mac Pro, and was just "The New Mac Pro"

This is of course assuming Apple can get chip supplies at all. Given the number of cars being de-contented, and stories of iPad production lines being cannibalised for parts to keep the iPhone production rolling, I'd be shocked if Apple actually launches the machine this calendar year.

But, in the spirit of the thread, and not to be purely derailing, what I can't live without:

1. The ability to have displays driven by a standard retail PCI GPU (with the choice of either consumer or workstation cards), in a full-fat slot driven by whatever is the current fastest spec PCI version, with displays plugged directly to the card, and no involvement of thunderbolt in the display path.
2. Operating System support for whatever display resolutions and number the hardware is capable of. No 8k60/10bit = no deal.
3. The ability to recover and restore the system without an internet connection, OR another Apple device - there must be no form of software bork possible that requires anything more than a thumb drive for recovery.
4. Networking handled by a user-replacable card.
5. Wireless on a user-replacable card.
6. CPU to be replaceable.
7. RAM to be replaceable, at least 128GB installable, and standard ECC DIMM format.

None of the things Apple touts as being "advantages" of the integrated paradigm, provide a performance benefit worth more than the flexibility and reassurance benefit of a proper modular system. Slower, but modular, is better than fractionally faster (potentially, but not actually in any real world situations) and integrated.

 

[TrevorR90]

So many people want the mac pro to jump on the Apple silicone bandwagon but also fail to realize the mistake apple made with the 2013 Mac Pro...

 

[mikas]

While the real ASi tower would be fantastic, I don't believe in it, not now in this situation.

Because of that I think I would settle for SoC design and probably integrated ram too, lots of it of course, just if only they put in one or two standard M.2 slots and a standard PCIe slot or two, with some room for modern gigantic GPUs inside the case. You can keep this proprietary flash storage of yours too in it, if you absolutely need it for some reason.

And if not PCIe slots are possible for some obscure reason, you should support eGPUs at least please. And absolutely both the red and the green ones thank you.

Better throw that MPX out allready, there is no real need for that. It's for lockdown and unnecessary price bump only. It's for software to use all cards in parallel, and power feed by cables is pretty convenient, standard, cheap and more flexible for different installations too.

But yeah. Cube 3.0 is coming, just wait..

edit. Ok now I see it. It's gonna be the Lube 1.0. You can't even get hold of it, let alone upgrade it.

 

[edanuff]

A compact, minimally expandable "Pro" computer was the rationale for the 2013 Mac Pro. It was a market failure, because there is no significant market for an expensive, small, minimally upgradable computer.

On a unit basis, the 2013 Mac Pro outsold the previous generations. It was a technical failure in that they couldn't make the thermals work and fixing the design didn't make sense once they'd decided to move to Apple Silicon. If you want to toss around terms like "market failure", the most fair statement you might be able to assert is that entire Mac Pro business has been a small niche business but the idea that they were selling huge amounts of the old Mac Pro that suddenly cratered with the 2013 nMP was not actually what happened and is not what Apple stated in interviews nor what analyst reports confirmed. I have not looked into how the 7,1 has done but suspect that past the initial surge of sales around release has probably sold within the range of the previous models. I'd be super curious to know how much mainstream users overall have done serious expansion of their 7,1 models in that timeframe since Nvidia isn't supported on the 7,1 and the growing cost and limited supplies for components overall.

 

[GlynH]

For me;

1. Another 5,1 / 7,1 aluminium cheese-grater full size case. You know to put things in!

2. Proper GPU support and by that I mean include NVIDIA. FFS put your Big Boy pants on Apple and don't cut your customers nose to spite your own face similar to how you behaved with Sony over BluRay licensing.

3. PCIe slots. As many as possible. As fast as possible. I'd rather have a couple of empty slots than be left drooling over the latest & greatest PCIe card wondering which existing card I will have to sacrifice just to keep up with the times.

4. By all means make a smaller/cheaper(?) cube for those who want to buy into the Mac Pro Members Club and don't want expansion but please don't compromise those of us who do.

5. Easy path to run Windows for those who need to be compatible with the rest of the World or who want to run MSFS!

6. No more Apple proprietary hardware tricks please. You know the ones that make it difficult->impossible to upgrade components in future without resorting to hacks.

I could go on but I'll stop there as this is just a personal wish list and I have not bought an Apple computer since 2013 when IMHO Sir Jony was finally overcome with his obsession for thinness, started removing ports and put form over function in most Apple devices.

I am under no illusion that Apple will ever manufacture such a computer but I have started to become interested again these last few years since they admitted their mistake over the trashcan replacing it with cheese-grater MKII and the resurgence of things like Magsafe, SD Card slots etc. that never should have been removed in the first place IMHO and that of friends of mine who once were solely Apple aficionados.

And before anyone asks no I don't miss floppy disks and parallel ports but I would gladly swap just a single USB-A port for a bagful of dongles...?

-=Glyn=-

 

[4wdwrx]

I think most of the premises in your post are incorrect and invalid, but this is the one I'll quote. You believe there is very little use for the replaced parts. That's not been my personal experience thus far. For instance when I upgraded to the W6800X Duo card, I sold my OEM MPX module in nearly no time at all. Folks will buy the previous generation GPUs, specially if you'd specced out an up-rated version.

As for your PC/tower comment: again very wrong from my own experience. I'd had the same motherboard since summer, 2017(!) I've swapped the CPU once, and the GPUs twice, I think (might be three times). All in the same tower, and all with the same system otherwise.

Modularity is a big deal. And having swapping GPUs in the AS Mac Pro will be very welcome. It just doesn't sound like it'll be possible.

I understand there are that do upgrade, but I am referring to as a whole market. As an enthusiast, you may upgrade as well as others like us in the subset of the Mac Pro forums, but 99% in the world does not.

The Mac Pro takes like 10x the resources to make than a Mini and 10x the cost to purchase.

You upgrade your PC, which is great thing and helps ewaste and reuse. In your signature you must have a 7,1 Mac Pro and Newer Macbook Pro, what is the reason to upgrade to a completely new system, would that not just be additional waste since the 2017 PC was perfectly usable? What use is upgradeability of a system, if one still need to purchase additional systems?

Upgrading in the computing world is like manual transmission in car world. I would like to see all upgradeable computers and every car have MT options, but in reality, is a like more than need. How many here drive manual transmission?

And mainly the issue why the upgradeable computer and manual transmission is going away, because most in the forums says they want the feature, but very few actually buy, therefore the market for it continues to shrink, since companies are seeing low sales, ie. meaning low demand.

The next Mac Pro will likely be modular, but probably proprietary upgrade.

The current Mac Pro is upgradeable, but not much are actually "upgradeable", since one cannot get a new motherboard, power supply, fan, even graphics cards without modification of some sort. Multi SSD PCIE cards mostly do not work since, the motherboard does not have bifurication like PC motherboards.

I actually bought another MPX 580 Pro to use (I might have bought yours ) to repurpose those getting rid of them for crossfire / multigpu usage. It does not work in bootcamp unfortunately, so forced to sell it also. I thought about tearing it apart to tinker with. Who know what the next person uses it for, might be door jam.

 

[mattspace]

On a unit basis, the 2013 Mac Pro outsold the previous generations.

Independent source?

More importantly, as compared to competitors? Did the 2013 increase Apple's Pro Workstation market share, or decrease it? If it didn't increase, or even preserve their market share in its target market, it's a failure by that metric alone. I personally do not believe a single HPZ customer moved to, and stayed with the Mac Pro because of the 2013. It is well documented that plenty of BIG Mac Pro multi-machine installations dumped Apple's gear when the 2013 was their only Workstation Mac option.

It was a technical failure in that they couldn't make the thermals work and fixing the design didn't make sense once they'd decided to move to Apple Silicon.

It was apparent the design was a failure within 6 months of release, as a result of studios burning them out over and over.

The 6 years after that was ~2-3 years of figuring out a throttling solution and waiting for extended warranties to expire, ~2 years of designing the iMac Pro as a strategy to replace the Mac Pro entirely, and 6 months of "oh crap, no one wants this, better dump it to the market as an orphan at birth, and do something else", and then 18 months of "how can we engineer a reason for Thunderbolt to exist on a Workstation".

Apple Silicon was not a forgone conclusion for the Mac at any time of the 2013's reasonable lifespan. Hell, given what a janked & kludged hurried re-boxing of iPad hardware the first gen AS Macs were, AS was barely a strategy for the Mac when the 7,1 was introduced.

If you want to toss around terms like "market failure", the most fair statement you might be able to assert is that entire Mac Pro business has been a small niche business but the idea that they were selling huge amounts of the old Mac Pro that suddenly cratered with the 2013 nMP was not actually what happened and is not what Apple stated in interviews nor what analyst reports confirmed.

A reliably working product is pretty much the number 1 requirement of a "Pro" market. If the product isn't reliable, that's a failure right there. Every external-expansion-paradigm small-form-factor "Pro" computer Apple has made, has been an unreliable piece of garbage, subject to recall and repair programmes.

For some reason, people keep wanting Apple to do something they've demonstrated over and over, they have no talent for doing.

It doesn't matter how well they can make a laptop, or a full size slotbox, or an (overpriced) mini PC, every time they've tried to make a compact, powerful system, it's been a disaster.

The collapse of Apple's reputation and presence in the Pro Workstation field, says pretty much everything about how "successful" the 2013 and its entire paradigm was.

As for Apple's claims about what "actually happened", Apple has significant form for gaslighting & outright lying in interviews and executive statements. You can dismiss their version of history without a second thought.

I have not looked into how the 7,1 has done but suspect that past the initial surge of sales around release has probably sold within the range of the previous models. I'd be super curious to know how much mainstream users overall have done serious expansion of their 7,1 models in that timeframe since Nvidia isn't supported on the 7,1 and the growing cost and limited supplies for components overall.

Apple went to the trouble of manufacturing multiple revisions of MPX GPUs - more of less keeping an annual release schedule for generational updates: 580/VegaII, 5x00, 6800/6900. So at least they think enough of their customers are doing it to make it worth their while.


But again, not to be a derailing killjoy, the thing I've posted here a couple of times - I think Apple should basically do an AS version of Intel's Beast Canyon NUC - which is a 3 slot backpane, one of which has a whole PC (with socketed storage and ram) on a PCI card, then space for a single 3080ti class GPU, then a third PCI card, for whatever the user wants it to be.

But, I think it should be done with 3 double-width MPX slots (basically the PCI side of the 7,1 behind the bottom 2 front fans), so the user can have 3x AS processor MPX modules for a desktop cluster, or 1 processor module, with 2 MPX or PCI GPUs, or a GPU and some other PCI card, or no PCI graphics, and just use the iGPU in the AS module.

Point being, I don't begrudge an AS GPU if it's really good for computational stuff, I just don't want to have my monitor & display options limited by a most expensive, and non-upgradable part of the system.

 

[theluggage]

So many people want the mac pro to jump on the Apple silicone bandwagon but also fail to realize the mistake apple made with the 2013 Mac Pro...

Trouble is, the Apple Silicon chips we have seen so far are perfect for making a better trashcan/cube (they'll never re-visit that shape), but not so good for a big box o' slots like the 2019 Mac Pro.

Thing is, I think there is zero interest in the 2019 Mac Pro from anybody who isn't deeply and expensively committed to some MacOS-only workflow. People I know who work with high-end PC systems see the Mac Pro as a joke (and although it shouldn't be important, Apple really played into that with their $1k display stand and $800 wheels). Outside of the ability to run MacOS, Apple have a really hard job distinguishing a Mac Pro tower from any other Xeon/Ryzen workstation. A lot of the "top trumps" features of the MP - insane number of PCIe lanes, RAM capacity - came from the new version of Xeon W they adopted, not anything clever Apple did - and these pushed it out of the price range of the middle-range Mac users who just wanted a modestly powerful and expandable headless system. ($6k for a system no faster than a top-end iMac until you add another $6k worth of RAM and PCIe cards).

At that end of the market the Mac Pro isn't just in competition with what you can spec up on HPDelnovo's online product configurator - there are plenty of places that will make a PC workstation to spec from a huge range of available components, including specialist stuff like purpose-designed GPU computing setups, scalable multi-Xeon systems with even more cores than the Xeon W, proper rackmount systems designed for high-density computing (not just a Mac Pro tower transplanted into a rackmount case). Apple dropped out of the rackmount/server/high-density market when the dropped the XServe - and that means that - if you need any of those things - you can't go full-stack-MacOS and (e.g.) your render farm or distributed computing setup will have to be based on Windows/Linux software.

Ultimately, the pool of users tied to MacOS because the cost of software re-tooling makes a Mac Pro economical is only going to shrink - staff will turn over, new brooms will sweep, Windows/Linux versions of 3rd party software will catch up and overtake Mac versions. Mac's role in the media production/scientific market started when Mac hardware/software was in a completely different league to the PCs of the day. Its future relies on the Mac being able to do things that Windows and Linux can't.

So, for the mid/long-term, Apple has to find some new model of computing that will make Macs attractive. The Trashcan was an opportunity to do that - but constrained by the fact that it was still just a (differently-shaped) PC with Xeon processors and AMD GPUs, so the only thing it could do differently to PCs was run MacOS. With Apple Silicon, they can potentially do better by reaping the performance benefits of powerful on-chip GPUs, unified RAM etc. and also by making sure that there's a CPU/GPU update path for new models. The downside is that those advantages will probably rely on optimised, MacOS-only software rather than ports of generic Windows/Linux code. The upside is that those are already the only non-legacy applications that justify buying a higher-end Mac.

So I think the way forward for Apple probably is making a compact "appliance" for FCPX/Logic Pro and any 3rd party apps from developers willing to invest in Apple Silicon optimisation, and ceding the really high-end "general computing" market to Windows and Linux. Their most prolific target market could be one-man-bands and hobbyists rather than film studios - a market that should have mushroomed since 2012 with all those YouTubers and self-produced rock stars out there.

...and the existing intel Mac Pro still has some years of viability left in it - and a far better update/upgrade path than the trashcan - so they don't have to throw the high-end customers under the bus.

 

[AndreeOnline]

Apple Silicon chips we have seen so far are perfect for making a better trashcan/cube (they'll never re-visit that shape), but not so good for a big box o' slots like the 2019 Mac Pro.

Yes—to the first part.

And they SHOULDN'T revisit the shape. "Small and circular" were things no one asked for of a workstation. Just like "keep all connectivity external", was never uttered.

That said, keeping the part of the Mac Pro small and silent where it can be, leads to the most efficient footprint in the end. They should update the design to allow for the cover to come off without removing all the cables, though.

[The Mac's] future relies on the Mac being able to do things that Windows and Linux can't.

Or... won't?

There are users who believe "free" is better than paid and "more is more". Personally, since I understand you can't really deliver products and services to the masses for free (but I am a huge fan of Open Source and community-driven projects), I much prefer paid goods and services. It's much more transparent.

The new MacBook Pros (like my M1 Max) are a great representation of what a Mac could be. They are fairly priced and much appreciated. They strike a great balance between power, usability, style, and price.

So, for the mid/long-term, Apple has to find some new model of computing that will make Macs attractive.

I think they have. But here's the kicker: they need to make Macs attractive to Apple! To themselves.

Mac Pros used to be a 'PC case blessed to run MacOS'. It was filled with stuff from other manufacturers. Sure, Apple applied their margins, but once they were in the possession of customers... all bets were off.
I think my own adventures with the Pixla Mac Pro were a good example of that. How does that benefit Apple?

If Apple can get their juices flowing thinking about future Mac Pro projects.... well, we all stand to benefit.

 

[ZombiePhysicist]

Slots or it’s not a real Mac Pro.

 

[JMacHack]

The people who want 3rd party gpus might as well get out your pitchforks and torches because it’s not happening. Mojave killed NVidia gpu support and the writing has been on the wall since then.

Prediction:
The apple silicon mac pro will be a “dumb” motherboard full of mpx slots and the processor proper will be on daughter boards that plug in (similar to the mpx modules). Up to four “m1 max modules” will be able to slot in. Along with pci compatible slots for things like avid cards or other accelerators like afterburner.

Looking at the 2019 Mac Pro and the architecture of Apple Silicon, I feel this makes the most sense.

I really doubt that Apple would be so dense as to try two small form factors that failed spectacularly again, especially after making an uncharacteristic apology for it.

 

[theluggage]

The apple silicon mac pro will be a “dumb” motherboard full of mpx slots and the processor proper will be on daughter boards that plug in (similar to the mpx modules). Up to four “m1 max modules” will be able to slot in. Along with pci compatible slots for things like avid cards or other accelerators like afterburner.

Maybe even with M1 Max MPX modules that could be fitted into an existing Mac Pro. The "bottleneck" would be communication between the processor modules (and the RAM and GPU cores would be distributed between these) over something like PCIe or Thunderbolt c.f. the ultra-fast, ultra-short on-chip/on-package links within a 2x or 4x MaxMax package. It would need some clever software to use it efficiency (...somewhere between regular multithreading/SMP and the old XGrid system...)

Given that the ARM transition must have been more than a twinkle in Apple's eye by the time they were designing the Mac Pro, it's a bit odd that they put the effort into developing the MPX slot idea for something that was likely to be a one-and-done product. However, MPX would make a lot of sense as a way of plugging Apple Silicon "compute" modules into a rack...

I really doubt that Apple would be so dense as to try two small form factors that failed spectacularly again, especially after making an uncharacteristic apology for it.

The problems with the trashcan seemed to be:
(a) no internal expansion,
(b) thermal constraints limiting GPU and CPU power
(c) too reliant on software being optimised to share the load between 2 separate GPUs and the CPU.
(d) Apple never updated it - maybe because Intel and AMD never made suitable chips.
(e) The old Mac Pro had been left to rot for years (and had been discontinued in Europe a year before) so there was no alternative for people who needed a big box o' slots.

But with Apple Silicon:
(a) Not sure I'm converted to all-external expansion - but Thunderbolt is a few generations better and faster now, and a 4xM1 Max will have a shedload of Thunderbolt ports.
(b) Apple Silicon's Unique Selling Point is vastly improved power/thermal efficiency.
(c) Apple have already bet the house on software being optimised for Apple Silicon/neural engine/Metal etc.
(d) It's up to Apple to make sure that processor updates are available in time
(e) The 2019 Intel Mac Pro is reasonably up-to-date and might even get some updates

...as I've said, I don't think Apple would dare re-visit the "triangle of boards around a common heatsink" concept because of the bad karma, but the more general idea of a sealed appliance for FCPX/Logic Pro and other Mac-optimised apps would probably show Apple Silicon in its best light.

...especially in the light of the rumours that there will be two "Mac Pros", large and small, with the large one being more like the current model.

 

[goMac]

Up to four “m1 max modules” will be able to slot in. Along with pci compatible slots for things like avid cards or other accelerators like afterburner.

The design of M1 Max isn't really compatible with being put onto a module. They're chiplets, which means they need to be physically right next to each other to be linked.

 

[theluggage]

And they SHOULDN'T revisit the shape. "Small and circular" were things no one asked for of a workstation. Just like "keep all connectivity external", was never uttered.

As Henry Ford ( apocryphally) said - "If I'd asked people what they wanted, they'd have said 'faster horses'".

10 years ago, Apple tried something "courageous" and struck out. That doesn't mean it's not worth another try. What I was trying to say is that they won't copy the "3 boards around a triangular chimney" concept.

Heck, what if they'd said "Touch screen smartphone? Nah - we tried the Newton 15 years ago and it tanked..."

 

[Joe The Dragon]

The design of M1 Max isn't really compatible with being put onto a module. They're chiplets, which means they need to be physically right next to each other to be linked.

But multi socket ?

 

[theluggage]

The design of M1 Max isn't really compatible with being put onto a module. They're chiplets, which means they need to be physically right next to each other to be linked.

I think you'd be talking about complete M1 Max x2 or x4 packages on each board and boards working together as something closer to a computing cluster with super-fast PCIe communications instead of network links.

Apple already have some form here: https://en.wikipedia.org/wiki/Xgrid

...and the sort of "if you need to ask the price you don't need one" Mac Pros these would be going into are already in competition with things like render farms and (especially now) cloud computing, which already rely on scheduling tasks on computing clusters...

 

[TrevorR90]

The people who want 3rd party gpus might as well get out your pitchforks and torches because it’s not happening. Mojave killed NVidia gpu support and the writing has been on the wall since then.

Prediction:
The apple silicon mac pro will be a “dumb” motherboard full of mpx slots and the processor proper will be on daughter boards that plug in (similar to the mpx modules). Up to four “m1 max modules” will be able to slot in. Along with pci compatible slots for things like avid cards or other accelerators like afterburner.

Looking at the 2019 Mac Pro and the architecture of Apple Silicon, I feel this makes the most sense.

I really doubt that Apple would be so dense as to try two small form factors that failed spectacularly again, especially after making an uncharacteristic apology for it.

You forgot AMD? I have a 6900xt in my Mac. Works like a charm in both bootcamp and Mac. I even have a 3090 in my mac for the bootcamp side and as a professional data scientist, actually use it.

 

[AndreeOnline]

As Henry Ford ( apocryphally) said - "If I'd asked people what they wanted, they'd have said 'faster horses'".

I'm all for innovation and I'm against focus groups.

Why the hell would you ask the general public what they think? If they were such great innovators, they'd be the ones running successful companies—so that part Ford got right.

For all I care, a workstation can be the size of a matchbox, completely silent and run off of the light that seeps through a small window on a cloudy day.

Just don't make those design choices at the cost of accepted compatibility*, power and flexibility.

I my mind, the Trashcan made a large number of exceptionally unnecessary trade-offs, after Apple had tied their own shoelaces together.

People were expecting faster horses, but Ford brought them a car. He didn't, though, bring them an wheelbarrow pulled by two ponies, while trying to convince them how convenient it is with "horses" that eat so little and **** in small, manageable piles.


*= but I'm generally pro cutting the chord of ageing technology sooner, rather than later.

 

[blackadde]

A tiny desktop workstation made no sense in the Cube, the Trashcan, or now.

Think about the use cases for most 'big iron' workstations. What they need is:

a) more power than the market alternatives while having
b) the ability to solve the the same computational problems and
c) a lower total cost of ownership over the lifetime of the machine

Nobody anywhere is going to buy Apple workstations at a serious premium and wire them into a render farm because your costs explode. And because the uptake for professional Apple systems has been in a death spiral since at least 2013, few vendors are willing to build out extremely specialized big-iron software that only works on the Apple software stack. These are not consumer products where blue-sky possibilities are exciting and the risks are minimal. I have a job to do with the box under my desk, and I need a good box to do that job or else someone will eat my lunch.

The problems in the workstation space are very very well defined. If your tiny desktop box doesn't actually solve these problems better than the alternatives, nobody is going to buy it except the Mac OS ride or die nerds that are willing to give up GPU compute flexibility and many thousands of dollars to avoid Windows or Linux. And there are very few of those left.

For real though, who is a tiny only-externally-expandable desktop powerhouse for? A/V carts? Is that it?

 

[mattspace]

Prediction:
The apple silicon mac pro will be a “dumb” motherboard full of mpx slots and the processor proper will be on daughter boards that plug in (similar to the mpx modules). Up to four “m1 max modules” will be able to slot in. Along with pci compatible slots for things like avid cards or other accelerators like afterburner.

That's pretty much what I've been advocating with a "MacStation" idea - an MPX version of the Intel NUC paradigm.

Something to consider... the MPX format is basically a 16x PCI slot, with a second slot for power delivery and Thunderbolt back & forth... So perhaps the AS mac Pro could see MPX2, or "MPX Extreme" - a second slot ~1cm above the existing PCI & MPX slots, that provides a stupid-wide connection between two MPX2 modules, so a single MPX2 bay can accept a PCI card, or a MPX card, OR an MPX2 card (which could use all three connectors), which could for example be a second, or third Mac Station module. The MPX2 slots would function in a similar way to the Infinity Fabric bridges in the current MPX video cards.

I could see a developer narrative that you can work on one module, and deploy to the other as your test environment etc...

And the porcine flocks wheel pinkly with abandon across the blue sky...

 

[Boil]

I think the key to having multiple SoCs on multiple cards comes down to ultra high-speed bandwidth & ultra low latency...?

Daughtercards on a backplane, smaller desktop box with two or three cards, larger deskside/rackmount box that holds eight or nine cards (or more, or less)...?

You get the Mac Pro Cube, then pull the cards to upgrade, moving the pulled cards to the server box...?

 

[mattspace]

The problems in the workstation space are very very well defined. If your tiny desktop box doesn't actually solve these problems better than the alternatives, nobody is going to buy it except the Mac OS ride or die nerds that are willing to give up GPU compute flexibility and many thousands of dollars to avoid Windows or Linux. And there are very few of those left.

The problem is that there are people who see the slotbox as in some way lazy, or a compromise - the sort of thing that gets made when people don't just try harder to get to the nub of the thing, and make a specialised solution.

These are the sort of people who think the phrase "A camel is a horse designed by committee" is pithy, instead of brutally ignorant. Or, who forget that the most common reacton to the 2019 was "oh thank god now I don't have to use breakout boxes any more, I can put everything inside the machine"

The slotbox is the pinnacle of evolution for a general purpose machine. Like crabs and weasels, it is the platonic ideal of desktop computing

 

[blackadde]

The problem is that there are people who see the slotbox as in some way lazy, or a compromise - the sort of thing that gets made when people don't just try harder to get to the nub of the thing, and make a specialised solution.

I think so too. The slotbox is the specialized solution.

 

[basehead617]

I predict they will make intel based mac pros for at least another 5 years

 

[goMac]

I think you'd be talking about complete M1 Max x2 or x4 packages on each board and boards working together as something closer to a computing cluster with super-fast PCIe communications instead of network links.

Apple already have some form here: https://en.wikipedia.org/wiki/Xgrid

...and the sort of "if you need to ask the price you don't need one" Mac Pros these would be going into are already in competition with things like render farms and (especially now) cloud computing, which already rely on scheduling tasks on computing clusters...

…XGrid?

I mean sure, if you’re still one of the two people running XGrid, I guess that could work.

XGrid required very special specifically tuned software and very special use cases. It wasn’t just “Your Mac has multiple Macs now.” Think more render farm, less multiple CPUs running the same software.

An actual render farm would make way more sense. Scales much higher. Hardware is cheaper and more commodity. Shared amoung multiple users.

I don’t think there will be multi socket options. A quad chiplet M1 Max will be the high end. Period. No cards.

The quad chiplet version is likely to be $15000 anyway so I’m not sure why people are trying to work cards into it. A system that complicated with that many cards and high bandwidth connections would be many multiples of $10k.

A lot of the this seems to be based on how Apple will do replaceable GPUs when the easiest answer is: They won't. There won't be replaceable GPU boards. And so there isn't a reason to imagine CPU linked boards when the easiest and most logical answer is there won't be removable/swappable boards.

Which you don't have to go far to imagine because this is all exactly what they did in 2013.