Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
One modular possibility is that the MacPro has upgrades in the form of complete smaller logic boards; some boards would prioritize the CPU while other boards would prioritize the GPU with a different arrangement of cores. The boards come with the RAM but not the SSD, which would be upgradable separately. Apple could extend on UltraFusion, which no one really saw coming on the M1s, but make it connect the boards. It wouldn't be as fast as if they were all the same board, RAM would be faster because its not using DIMM connectors, but it would be really fast nonetheless.
 
Last edited:
I wish I had the $12k or so for one of these, I have a feeling this is the pinnacle of an era we'll never see again.

This is the last Mac that will ever be able to play PC games. Almost certainly the last with any kind of PCI slots. Almost certainly the last full size tower.

I hope I'm wrong about those things but I doubt it.
 
We are all curious to what the AS Mac Pro will be like. How many SoCs will it use, and how Apple allows customized builds?

A SoC -- system on a chip. If it actually is "on a chip", then if have multiple SoCs then probably going to have multiple systems.

What Apple needs for the "ultra" (and up) class SoCs is a better chiplet disaggregation. What they did with the M1 Ultra is somewhat of a kludge. It is too large and chunky to be an effective chiplet. It scales to two OK, but if tried to push past that it has a number of quirky problems that only get worse as add more. It is really a laptop chip that is being shoehorned into a role it really isn't good for the farther you get away from being a laptop.

Apple still will probably do a system on a chip ( package) . But a 'desktop' building block that is incrementally different is likely. Past 4-6 Thunderbolt controllers probably isn't useful. ( really past 4, but put some on the front and some on the back has an ease of use upside in a desktop context. But 8 or 12 ... that is just grossly gratuitous and wasteful of expensive die area. . ). If could add x16 PCIe v4 complex instead of four TB complexes (with x4 PCI-e v3 ) once gotten to baseline 4-6 TB controllers then that would scale much better as moved up.

Don't need 4 SSD controllers (for only one internal SSD) . Don't need 4 Secure elements. Don't need 16 x1 PCI-e v4 lanes for Ethernet and WiFi.

They don't need to 'blow up" the basic clustering of the CPU , GPU , and memory complexes to do that. the Secure element, TB controllers , 'low end' I/O via x1 PCI-e v4 , and SSD controller are at the outer edges of the layout anyway. Move them onto their own die via UltraFusion derivative and would have a much better building blocks to scale up with. Still inside the same package ( so still a SoC), but the aggregation much better suited for scaling up past 2.

Doubtful Apple tries to go 'toe-to-toe' with 64+ PCI-e v4 aggregate bandwidth output that Intel and AMD workstation CPUs do. But something better than 8 x1 PCI-e v4 lanes is pretty likely. ( decent chance won't be doing something "messy" like tasking a glut of TB controllers not using into provisioning internal PCI-e v3 lanes. )


TSMC N3, N2 ,etc are going to be even more expensive so wasteful area expenditures on SSD controllers , TB controllers ,etc that Apple (and end users ) are never going to use just gets more and more expensive going forward. It isn't going to add much 'value add' for customers at all.




Multiple SoCs though could be interesting though. If Apple put a M2 Pro on a PCI-e card then perhaps could deploy Macs on systems with PCI-e slots. A role reversal of the x86 compatibility cards where put x86 cards into Apple systems.



Could stuff 4-5 M2 Pro cards into a racked MP 2019. Or a new MP that had 5-6 slots. If the cards each have 2-3 TB and and Ethernet socket, then basically 'blades'.

A Mn Max - Ultra on a a double-triple width, full length , full height PCI-e card would probably work. Add a virtual Ethernet over internal PCI-e bus connection on both sides ( host and guest computer sharing the bus) and would have dual computers packed into a smaller package. No 'new' SoC really needed (somewhat just a Mini Pro or Studio without the enclosing box. ). Any "distribute compute over the local LAN cluster" software would just work ( just much faster network connections between the nodes).

P.S. if host/guest system network drivers, the host driver could be Windows or Linux. If have customers who need 3rd party display GPUs then space efficient way for them to keep around a 2nd computer that is a Mac.
 
Last edited:
  • Wow
Reactions: WiiDSmoker
Yes, this is an immensely powerful machine, but Apple f***** over its prosumer/hobbyist base with these outrageous prices.
I'd love to argue that you get what you pay for & point at the high engineering standards of the Mac Pro, and the fact that Intel's "suggested" prices for their Xeon-W processors are an arm and a leg to start with (...and the 24 core+ models use the M-suffix Xeon variants to get the massive RAM capacity, for which Intel charges twice as much).

...but the indefensible price of the wheels, the XDR display stand etc. kinda saps any goodwill towards Apple on that - although I suspect that's more them trying the luxury car add-on business model (plus, I bet that stonkingly overpriced accessories are great for sweetening deals with VIP customers - "if you order 100 Mac Pros + XDR we'll throw in 10 sets of wheels and 10 stands for the corner office guys... a $20,000 value!!!)

If I was more enterprising I'd start re-selling $15 IKEA castors bundled with some little bushes to adapt them to the Mac Pro legs (I'm sure some Chinese firm could make me a few hundred for a modest investment) - a bargain at $200 (for a pack of 3).

So, yeah, you can easily pay $6k+ for a comparable Xeon workstation from other manufacturers - but you'd probably get better base GPU, RAM and storage specs (and you could probably get AMD for a lot less). Or you could pay $3-$4k for something with worse specs on paper which only took 765G of RAM and could only accommodate two top-end workstation class GPUs. Oh, the humanity...

My impression of the Mac Pro is not so much good engineering and design as over engineering and maybe not-so-wonderful design - machine that will often be kept under a desk has ports & power switch on the top? Rackmount version has memory slots underneath? Expensive wheels have no brake? You couldn't have worked 4 boltholes for VESA mounting into that exiting industrial pattern on the back of the XDR? Optional HDD that sits directly in the exhaust from the CPU cooler? Do the PCIe slots really need cast aluminium blanking plates containing enough metal to make 50 aircraft-grade Coke cans? Might there be cheaper and reasonably attractive front grilles than that exciting bit of modern sculpture - probably costing a small fortune to carve from an aluminium block - on the front?

...none of which are deal-breakers if the alternative is $100,000+ in lost productivity and re-training to switch your MacOS only workflow to Windows or Linux - but that's about the only use-case I see for the MP, which really looks as if it was built up to a price.

At least now we have the Mac Studio - which is hardly cheap but rather more affordable than the MP, and with a decent base spec.

Why bother with the redesign, then? It’s a stopgap before the apple silicon mac pro. Too much of a hassle if they were planning on dropping the line.
Warring factions inside Apple?

Well, the redesign only exists in rumours and sketchy leaks. All Apple said at the Studio launch was that the Mac Pro was "for another day" (not "coming soon") - and I think the words at that event are always chosen very carefully. If the 2019 Mac Pro hadn't happened - or had turned out more like the 2013 Mac Pro - the Studio would be a credible Mac Pro/iMac Pro replacement.

Still, being more positive, one possible direction for a new Mac Pro would be something that could take several Mx Ultra "computing units" each contributing CPU, RAM and GPU and creating a PCIe-linked "cluster". In that case, the Mac Pro case design could be re-used with units plugging in via MPX slots (PCIe + extra power & thunderbolt routing - perfect) (you might even be able to add Apple Silicon 'compute units' to a 2019)
 
  • Like
Reactions: entropys and trusso
I do wonder if Apple is going to make a display as large again as the XDR. I have a feeling that the rumored high end 27" Studio display that is coming out is going to be the replacement for the XDR.
 
Multiple SoCs though could be interesting though. If Apple put a M2 Pro on a PCI-e card then perhaps could deploy Macs on systems with PCI-e slots. A role reversal of the x86 compatibility cards where put x86 cards into Apple systems.
I previously compared multi-CCA slotted STE's to what a future Mac Pro could be like. I also considered that these Mac Pros could act like computer clusters easily under UNIX, even to the point that a business could use in that manner with Mac Pros to Mac Pros connectivity. What we are all waiting for is what direction/technology has Apple decided to go with as it provides scalable AS Mac Pro workstation solutions. Maybe a return to the business marketplace not just creative usage.
 
Why do some people find it so hard to imagine that a big box with many slots might already be the best possible solution for a particular user group? I suspect it's because the concept is almost as old as the personal computer. Jobs was pretty spot on with his trucks metaphor; just because a big engine with a lot of empty loading space on wheels is a very old idea doesn't mean it isn't the best solution for a particular task.
you completely ignored what i said originally
 
Thunderbolt had a minor hit in 2017-2018 generation graphics cards and SSDs. Today the Thunderbolt bandwidth takes a very big hit with new graphics and SSDs and there is no announcement of a next gen Thunderbolt with X2 more or X4 more bandwidth yet.
i saw no effective difference between 3070 ti over TB3 vs pcie on my pc.

tb4 exists too
 
The shared PCI backplane is not solved with Thunderbolt. The lack of Afterburner is not solved with TB. The custom GPUs are designed with TB 3 ports on them specifically for Apple. They are Pro level cards. One hopes they order custom ASICs for RDNA 3.0 cards and imagine if they bothered with the MI 200-300 line of CDNA cards. Then add-in third party custom options and sorry but TB doesn't have a thing to do with most of this. Apple will have a dead product if it doesn't have the same slots for third party daughter cards on those secure PCIe 4.x or possibly 5.x slots.
afterburner sales are minimal i suspect. apple can build that in if they wanted the trashcan design too.
 
thunderbolt solves most of that. minor performance hit with eGPUs but works fine. and you can add more than a few. would be even easier if apple made a thunderbolt addon case. easy swap in/out and connect via TB.
Not at all…elegant. That matters you know.

Also a performance hit.
 
Last edited:
  • Like
Reactions: cookedart
Maybe Apple is giving up on both because the demand is just not there for this type of machine. I can only imagine it being a low volume seller at its current price.
It would always be a low seller. A very discrete market for buyers with specific needs. It is not for the likes of you and I. But that is not its only purpose.

It is also a halo device that demonstrates what is possible with a Mac and encourages purchases of more mainstream macs. And a test bed.

Like back in the day when car manufacturers made Supercars and monster trucks.
 
...even thinking about the environment, which is less likely to go to landfill - an old-model GPU or outdated & worn SSD removed from a Mac Pro or a complete working 3-year-old Studio?
Full disclosure, either one is going to the landfill eventually if the user doesn’t want to recycle. One is just going to the landfill by itself, the other is going to the landfill WITH the discarded old RAM, discarded old storage, discarded old GPU, etc. So, for someone that’s not going to recycle anyway, it’s best not to give them more options of things to get (and have packaged and shipped) and sent to the landfill.

For folks that recycle, nothing’s going to the landfill either way.
 
It is also a halo device that demonstrates what is possible with a Mac and encourages purchases of more mainstream macs. And a test bed.
I don’t think that’s been true for a long time. Take right now, there’s not even an Apple Silicon Mac Pro, and the efficient mobile devices that people actually desire (MB Air, MBPro) are selling quite well.
 
  • Like
Reactions: Duncan-UK
With cloud computing more accesible than ever and with individual workstations becoming more like appliances, the justification for buying these Pro towers is disappearing fast.
 
For millions of Apple fans - the new Mac Pro was a crushing disappointment. The price is out of this world.

The Mac Pro is not meant for millions of Apple fans, rather a select group of professionals. Pricing sure isn't for regular consumers but companies that require their power could afford them.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.