Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I wanted to add to this thread with a YouTube video made by Jon Rettinger.


The first few seconds of this video .... hmmmm.

Looks like a vertically oriented IBM PCjr expansion bus.

ibm-pcjr-expansion-sidecars_1_522b15c392a416258bf868b3f882120b.jpg


While the jr had a horizontal / stackable bus. This design leverages the same concept in a vertical chassis.

If this is the Apple's approach... I expect over priced, under powered, over visually designed add-on components.

Legacy_Front_thumbnail.jpg
 
Last edited:
  • Like
Reactions: crjackson2134
All your criticisms of the idea have opposing arguments in their favour that are no less valid.

Maybe, but I suspect those arguments come from fringe use cases rather than the the bulk of actual potential buyers.

My reasoning isn't so much that the stackable concept couldn't be made to work, it is that Apple would not want to make it. It just doesn't fit their aesthetic or their obsessive control over how things look even when in the hands of users.

I'm of the mind that it'll be tool-less internal modular upgrades only. Perhaps also making it easier to link a bunch of Macs together to share the load of some tasks.

Minimalist box for sure.
 
Maybe, but I suspect those arguments come from fringe use cases rather than the the bulk of actual potential buyers.

The Mac Pro's only userbase, is Fringe customers. It's such a small product, I don't think it's possible to identify any "majority" use case. The largest use case, will still only be a small minority of its total users.

You've suggested inventory would be a problem, but a stackable system can offer more customer configs, with fewer standard manufacturing SKUs.

Look at the iMac Pro - every CPU, times Every GPU.
Vs every CPU , plus every GPU.

I think stackable, and embracing the stackability as a part of the aesthetic, (eg styling the outside with horizontal banding, almost like a heat sink, so the joins are hidden) is AS likely as building a tower that's big enough for any sort of internal parts swaps or upgrades.
 
Last edited:
  • Like
Reactions: nerdynerdynerdy
I strongly disagree, but I've already laid out my reasoning in the main thread. No point in going over it again.

Ultimately neither of us know, but whoever is right when they reveal it can point the Nelson meme.


(I think the user base is largely content creation)
 
  • Like
Reactions: mattspace
Aren't all workstations fringe products for a very niche audience?

Example... Lenovo and HP are the top two computer vendors in the world. Nearly half of all computers sold are from just those two companies.

But I'd imagine they sell more IdeaPad and Pavilion consumer machines than they do big powerful ThinkStations and Z8 workstations.

The point is... workstation-class PC sales are tiny in the grand scheme of things... no matter who the vendor is. But there's still an audience for them.

The Mac Pro is unique in that it's the only workstation that runs MacOS. Users who need to run MacOS on a powerful "headless" workstation can only do that with a Mac Pro.

Meanwhile... you can swap out a Windows-running Lenovo ThinkStation for an HP Z8 or Dell Precision and no one will really notice. :p
 
Wouldn't a stack require some pretty unique cabling solutions to make the modules all work at the required bandwidths and generally avoid a spider's nest of cabling? Not to mention reliability issues from wonky connections, machines being moved, etc.

I recall those SCSI tower setups where hooking up devices in a certain order was crucial to get the most out of them. Ok, perhaps not so different from having to be aware of which Thunderbolt plug corresponds to what controller inside the machine. ;)
 
  • Like
Reactions: Flint Ironstag
Seeing all these concepts makes me wonder... does the desktop workstation need an overhaul?

For years... everyone made workstations in a standard tower form-factor... Dell, HP, Lenovo and everyone else in that market... including Apple. And there were very few complaints.

Then... for some reason... Apple decided to get all "clever" and make a workstation in a cylinder. Well... we know how that turned out... as even Apple themselves admitted the problems with that.

But while Apple was experimenting with cylinders... everyone else stuck to the tried-and-true tower design. And notice how the other guys didn't have a 5 year lapse in workstation updates. I think we've seen how the tower is a proven design.

So do we really think Apple will try another crazy experiment?

These concepts sound cool... in theory. Stackable boxes, different modules for various functions, etc. You can build exactly what you want.

But that capability already existed in the tower form-factor. Albeit internally. And yet... people were fine with that. I haven't heard anyone say "aargh I can't add XYZ to this HP Z8 tower..."

My point is... I don't think the MacOS workstation market can handle another experiment.
 
  • Like
Reactions: haralds
It does not. The existing form factor of the cMP is fine. It just needs an updated logicboard/CPU board.
Actually, IMHO it needs a lot more than that.
  • The cMP is enormous. Almost 50 litres. The worst thing is that those painfully sharp (and fragile) handles make it 20.1" high, so that it can't fit in the 17.5" width of a standard rack. (Actual size of a 19" rack shelf)
    • Z4 is 29 litres
    • Z6 is 35 litres
    • Z8 is 53 litres, but supports 56 cores, 3 TiB of RAM, 48 TB of spinners, and 3 double-wide GPUs (but still fits in a rack)
  • The storage options are so 2005. Four 3.5 SATA bays are good, but four to twelve M.2 NVMe slots should also be standard. (And don't start the usual whine about "not enough PCIe lanes" - PCIe switches will let you connect as many NVMe drives as you need - my HPE servers come with x8 to six by x4 switches for the NVMe drives. About the only place where one-to-one PCIe lanes means anything is running artificial disk tests like Black Magic - most real applications don't care - they spend some time calculating.)
 
Last edited:
Modular AND looks - done a long time ago already. This is what I expect Apple will do:

Modules accessible on the back:


Front view:
o2-02.jpg


Enjoy those custom form factor parts. Priced like unobtanium. ;)

o2-02.jpg

Yes, years ago I worked on one of these along with just about every unix box on the market. There were a LOT at that time. At the same time I was also working on DEC VAX's with VMS OS.

Long before Apple "Changed the World" from Beige to Color with this:

imac_flowershot.jpg



SGI had this:

indigo.01.big.jpg


Around the same timeframe I ran an Apple produced port of System 7 on an HP 9000 Workstation. Anyone remember that ?

UNIX was very big in the mid-1980's. It still is today. Apple also had their own version of UNIX running I believe it was called A/UX (1988?).
 
Last edited:
I have a question.

The mMP could really use pcie 4.0 for obvious reasons.

AMD Epyc “Rome” second gen server processors will support pcie 4.0 already in 2019.

Could the mMP be based on AMD Epyc?

If an iMac Pro tops at 18 cores already, why not a 64-core MacPro?

If an iMac Pro can be configured up to almost 20k$, why not a MacPro that starts at 20K$ like it’s a RED camera or something?
 
To me the key point of a Mac Pro is to leverage the much larger PC installed base with high end products - just like my wonderful cheese grater.

This means an old style bus based system. Not more custom style Apple product, it misses the point. Apple modules would be quickly out of date because they cannot compete with the scale of the PC ecosystem.

It is not brain surgery, folks!
[doublepost=1553303284][/doublepost]
To me the key point of a Mac Pro is to leverage the much larger PC installed base with high end products - just like my wonderful cheese grater.

This means an old style bus based system. Not more custom style Apple product, it misses the point. Apple modules would be quickly out of date because they cannot compete with the scale of the PC ecosystem.

It is not brain surgery, folks!

ps: If Apple comes out with another Vader on your desktop, I will go Hackintosh.
 
What if modules are empty boxes with a pcie backplane where you can put whatever expansion card or gpu? (provided you have drivers)

They would connect via internal “draft thunderbolt 4”...itself based on pcie 4.0...
 
I have a question.

The mMP could really use pcie 4.0 for obvious reasons.

AMD Epyc “Rome” second gen server processors will support pcie 4.0 already in 2019.

Could the mMP be based on AMD Epyc?

If an iMac Pro tops at 18 cores already, why not a 64-core MacPro?

If an iMac Pro can be configured up to almost 20k$, why not a MacPro that starts at 20K$ like it’s a RED camera or something?
AMD Epyc MAY just let them offer an CPU / RAM add on module that stacks on the pci-e bus.

So
MAIN BOX <----> EXP PCI-E X64 <-----> CPU 2 IN PCI-E X64 <-----> PCI-E X64 stack out.

Yes 128 lanes with 1 or 2 cpus and the CPU to CPU LINK seems to be pci-e based.
 
I want to believe in an AMD EPYC mac pro.

Fully knowing I will never be able to afford it

But it would be a sight to behold...and honestly one of the few ways to avoid overlapping with the iMac Pro...
 
What if modules are empty boxes with a pcie backplane where you can put whatever expansion card or gpu? (provided you have drivers)

They would connect via internal “draft thunderbolt 4”...itself based on pcie 4.0...
Could look slick, but would be more expensive. You need additional casing, power supplies, etc.
I actually find a single floor standing unit much more practical than a bunch of boxes strung together.
 
From what I can see so far the T2 chip is gonna be a pain in the butt for the new Mac pro. You can't just put your own parts it has to be "approved" by the T2 chip. Great caution here.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.