Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Teases Apple Silicon Mac Pro at Spring Event
- Thread starter MacRumors
- Start date
- Sort by reaction score
It's Apple... I wouldn't expect support for double the current Mac Pro count. Maybe 4, or 6? I'm wondering how large the fan will be. 12"?
My guess on the Mac Pro, probably the same enclosure and expansion slots, updated I/O to a degree and configurable with 1x M1 Ultra or 2x M1 Ultra Unless they have a workstation class version under development that beats the Ultra, and absolutely wipes the floor with anything Intel and AMD currently have on offer, and not just by a small margin... We'll know by Christmas, probably sooner 
There's nothing to walk back. It is still true and still applies to all other Macs with Apple Silicon. But you can only put so much into a single SoC and package. So, if you want to go bigger, you're going to have design discrete components.
I still believe that the Apple silicon Mac Pro will have M1 Ultra and what will effectively be four M1 Max interlinked ("Jade4C-Die") as Mark Gurman reported back in May 2021, but in another thread one person is adamant M1 Ultra is the end of the line and that "Jade4C-Die" will not happen.
I disagree, but if Apple could not get "Jade4C-Die" to work and had to scrap it, what is Apple to do? William Ma claims that Apple will have a version of M3 with 40 CPU cores that will release in 2023. And now the pundits are claiming that Mac Pro will not ship until late 2023. So does this mean we have to wait for this 40 core M3 before Apple releases an Apple silicon Mac Pro?
And if they do, does that mean Apple might have "rushed" Apple Studio out yesterday because they are not in a position to announce an Apple silicon Mac Pro at WWDC in June when everyone is expecting them to? Will Tim will use the WWDC keynote to talk up M1 Ultra and Mac Studio and note that "Mac Pro continues to be a product in our line-up" just as he did with the Mac mini in 2017 prior to the new model dropping the following year?
The thermals on a "Jade4C-Die" would be through the roof. The M1 Ultra is probably sitting above 200W. 2 of those on the same package? You'd need a liquid cooling system. Not to mention, that package would be enormous!
Depends on peak CPU or peak GPU I believe. Peak CPU for the Ultra is 60W. Peak GPU looks to be around 110W. Max would be 30W peak CPU and 60W peak GPU.
Maybe, just maybe the M2 Ultra Max Quad chip debuts in a Mac Pro first June 2022.... ?
No doubt the M2 is all ready to rawk.
Double the base memory,
M2 16, 32, 64.
Pro 32, 64, 128.
Max 64, 128 256.
Ultra 128, 256 512.
Ultra Pro 256, 512, 1024.
Ultra Max 512, 1024, 2048
and finally
Ultra Max Quad, 1024, 2047, 4096
4nm maybe also they'd double all the cores, Neural engines, etc
No doubt the M2 is all ready to rawk.
Double the base memory,
M2 16, 32, 64.
Pro 32, 64, 128.
Max 64, 128 256.
Ultra 128, 256 512.
Ultra Pro 256, 512, 1024.
Ultra Max 512, 1024, 2048
and finally
Ultra Max Quad, 1024, 2047, 4096
4nm maybe also they'd double all the cores, Neural engines, etc
That is interesting if everything is doubled wonder why the Ultra doesn’t draw 120W?
Not sure. It will be interesting to see what new can be found out once they’re in folk’s hands.
Umm, no it's not.
The M1 Max uses 30W for CPU and up to 60W for GPU, plus a few watts here and there for other components. So at least double that to get the Ultra's power consumption; like I said, around 200W.
If you doubled that again with an "Ultra Duo", that's 400W.
I looked up the numbers we see from stressing the CPU and GPU at the same time and saw that what’s represented on Apple’s chart IS additive, not exclusive, so my mistake. If testing for the Ultra is performed the same as the Max, and Apple correctly predicted approximately 90W when stressing CPU and GPU on the Max, then the Ultra will be at 110+60, 170W. Within 30W of 200W, yeah, but definitely not above. (Again, if the numbers they’re providing in their charts are accurate and they have been for the Max.)
I personally don’t think there’ll be an Ultra Duo. It’ll be the same Ultra with some options for internal storage or expansion. The fanciest thing I think they may do would be related to whether or not Apple considers it “worth it” to provide a system that can address 1.5 TB of memory.
There's more to the SoC than CPU and GPU; 32 ANE cores, 128GB of RAM, plus a few other blocks... While they don't consume anywhere near the CPU and GPU, they are not insignificant.
Not sure what LPDDR4 consumes, but DDR4 can consume up to 3W for every 8GB.
Your last sentence is why your prior sentences are wrong. The capacity of air cooling is determined by the surface area over which the air is circulated. Because the heat density of a Jade4C-Die would be the same as the heat density of the smaller die, you could air cool it (as long as the package is, indeed, twice the surface area, and as long as you design the heat sink to cover the entirety of that area and maintain turbulent air flow over the entirety of the fins).
But running at higher clock speeds increases temperature. Has Apple ever yet said what the clock is for the various systems? I'm sure a notebook is going running slower, cooler, leaner power wise than a desktop version of the same chip which can obviously run faster, warmer, and unbridled by battery life concerns. So if the SOC runs hotter, the memory is going to run hotter too. 'Free space' in the chassis also plays into cooling. The Mac Pro should cool better than the Studio/Mini. It's just 'bigger', and dumps more air through that cavernous enclosure than a notebook, or Studio/Mini. I'd therefore expect the new Mac Pro to run faster (hotter) than earlier iterations of the same basic technology because it can breath better, has room for bigger power supply and NEEDS to be fast, 'pedal to the metal', along with all (most) of the other stuff in the box. So size of heat sink, but also how easy it is to get cooling into the box. Air cooling a fast Ultra alone wouldn't seem to make sense. Heat sinks and fans are far more efficient at removing heat. I'd expect the new Mac Pro to have some pretty interesting thermal management gear stuffed in the box.
Apple has not, but GeekBench shows the speeds and it ranges from 2064 - 3220 MHz.
This appears to be the same for all models: M1, M1 Pro and M1 Max. Since M1 Ultra is two M1 Max, it should apply for it, as well.
Not necessarily the case at all. M1 max could have 2 interconnects. Apple "hid" the interconnect fabric from their die shots and announcements until they announced Ultra, and it is at least possible the M1 Max actually has two such interconnects in order to support a quad chip configuration. It's also possible we wont get a quad chip setup (or just additional cores) until the M2 family, but all that is conjecture at this point. Either way they sure are stretching this M1 design a lot further and longer than I thought they would, and taking a lot longer to get to M2 than I'd hoped.
g\
I wouldn't expect a second generation SoC until after all systems had made the transition.
The single interconnect and making two seem as one is a crazy amount of work from what I understand. Would probably be an order of magnitude more complicated to get four together.
No it’s not. It doesn;t matter that apple “hid” die shots. There are plenty of third-party die shots. And many of us CPU designers pointed out that the blob on the bottom of the die was an interconnect fabric to connect to one other die. There are no other hidden interconnect fabrics, and there is no way that the fabric we all knew was there can connect to multiple die without the creation of a new, very complicated, die to sit between the fabric and another fabric. That new die would be very tall, and two actually be necessary in all likelihood.