Wednesday, March 9th 2022

Apple M1 Ultra Chip Uses Multi-Chip Module Design to Create a Massive Software Agnostic Processor

Apple yesterday announced its M1 Ultra processor. It is designed to be one of the most powerful solutions ever envisioned for desktop users, and it leverages some of the already existing technologies. Essentially, the M1 Ultra chip combines two monolithic dies containing M1 Max designs. They are stitched together to create one massive chip behaving in a rather exciting way. To pair the two M1 Max dies together, Apple has designed a package called UltraFusion, which is a die-to-die interposer with more than 10,000 signals. It provides 2.5 TB/s low latency inter-processor bandwidth and enables seamless sharing of information across two dies.

What is more interesting is that this approach, called multi-chip module (MCM) design philosophy, allows the software to view these two dies as a single, unified processor. Memory is shared across a vast pool of processor cache and system memory in a single package. This approach is software agnostic and allows hardware to function efficiently with loads of bandwidth. Apple notes that no additional developer optimization is required for the new processor, and the already-existing stack of applications for M1 Max works out-of-the-box. Talking about numbers, the M1 Ultra chip has a potential main memory bandwidth of 800 GB/s, with up to 128 GB of unified system memory. We are yet to see how this design behaves as the first Mac Studio units start shipping, so we have to wait for more tests to check these claims out.
Add your own comment

18 Comments on Apple M1 Ultra Chip Uses Multi-Chip Module Design to Create a Massive Software Agnostic Processor

#1
TheLostSwede
Apple notes that no additional developer optimization is required for the new processor
I guess this is the advantage Apple has, as they control the software and hardware, so they can make things like this with minimal extra effort.
Impressive nonetheless.
Posted on Reply
#2
Denver
Apple and its marketing team can say anything. The reality is another story...
Posted on Reply
#3
Tigger
I'm the only one
stitched together, another term for glued :p

But, going by Apple’s promotional videos and mockup animations, it looks like they’re using a small, silicon bridge of some sort. Which would make this similar in implementation to Intel’s EMIB technology or Elevated Fanout Bridge (EFB) technology. Both of these are already on the market and have been used for years, so Apple is far from the first vendor to use the technology.

Posted on Reply
#4
BArms
How do the GPUs get presented to the OS, as a single unit or more akin to SLI?
Posted on Reply
#5
FeelinFroggy
Tiggerstitched together, another term for glued :p

But, going by Apple’s promotional videos and mockup animations, it looks like they’re using a small, silicon bridge of some sort. Which would make this similar in implementation to Intel’s EMIB technology or Elevated Fanout Bridge (EFB) technology. Both of these are already on the market and have been used for years, so Apple is far from the first vendor to use the technology.

I am definitely not an expert on how CPU's operate or are manufactured, but reading about it sounds a little like AMD's infinity fabric.
Posted on Reply
#7
MKRonin
Talking about numbers, the M1 Ultra chip has a potential main memory bandwidth of 800 GB/s
That number seems high. Any more details on memory type and # of channels?
Posted on Reply
#8
Patriot
MKRoninThat number seems high. Any more details on memory type and # of channels?
LPDDR5-6400 32 half channels, or 16 channels in ddr4 terms.

The block diagram is out confirming dual m1 max so yeah... this is known
512/64 = 8 on each side, ddr5 operates in dual channel per dimm but this is soldered sooo.
In anycase the numbers work.

Mostly, 820GT/s
Posted on Reply
#9
MKRonin
PatriotLPDDR5-6400 32 half channels, or 16 channels in ddr4 terms.

The block diagram is out confirming dual m1 max so yeah... this is known
512/64 = 8 on each side, ddr5 operates in dual channel per dimm but this is soldered sooo.
In anycase the numbers work.

Mostly, 820GT/s
Cool, thx for posting.
Posted on Reply
#10
mechtech
Is the non-ultra M1 in iPads,iPhones, regular macs, etc. anything else??
Posted on Reply
#11
TheLostSwede
mechtechIs the non-ultra M1 in iPads,iPhones, regular macs, etc. anything else??
The new iPad Air, iPad Pro, the 24-inch iMac and the MacBook Air.
BArmsHow do the GPUs get presented to the OS, as a single unit or more akin to SLI?
Supposedly as a singel GPU, but it's kind of unclear at the moment.
DenverApple and its marketing team can say anything. The reality is another story...
It's not all marketing.
www.techpowerup.com/292789/apples-brand-new-mac-studio-with-the-m1-ultra-cpu-gets-first-benchmark-figures
Posted on Reply
#13
timta2
DenverApple and its marketing team can say anything. The reality is another story...
I've been hearing that for decades and have been waiting that long for someone to prove them wrong.
Posted on Reply
#14
watzupken
Nike_486DXpretty sure 3990wx is still ahead
I think it really depends on the kind of workload. The 3990WX is really just a brute force CPU, with no GPU. So if the workload runs well with GPUs, the extra cores may not be useful. Similarly if the workload uses the AI cores, etc, then it may do better. Also, if you look at power consumption, I don’t believe X86 chips will be able to match the efficiency of ARM based SOCs. The former is generally built for high power devices in mind, while the latter is built for low power compact and mobile devices. Not that X86 chips cannot be used for mobile devices, but you can tell that even those ultra Low power processors generally require quite a substantial amount of power to allow them to stretch their legs. If you limit them strictly to 15 or 25W, its not going to perform well.
DenverApple and its marketing team can say anything. The reality is another story...
I feel Apple’s marketing is probably the least aggressive. You can say that they cherry pick benchmarks/ tests to get the results, but isn’t it the same for AMD and Intel, or any other companies? If you product is strong in some aspect, you will definitely use that for marketing.
Posted on Reply
#16
jimmyxxx
This is an impressive feat. While I do hate some of Apple's practices, and most of their BS marketing. Their hardware team has been working very diligently and achieving very interesting milestones, and the one that I'm more impressed of is energy efficiency. For years the CPU and GPU market are increasingly expanding the power budgets of the desktop PCs and trickle down from there to achieve smaller power envelopes for mobile processors. However apple started the other way around, they had a very powerful and efficient processor already for their mobile platform and moved it to the desktop space. It is a refreshing change of paradigm.
Posted on Reply
#17
TheLostSwede
DenverThis benchmark has updated over time to favor apple over x86, apple's biggest advantage is the accelerators/Asic for specific tasks..
Did you at all read the news post? I specifically wrote that Geekbench isn't great for comparisons across platforms.
Unfortunately, it's the only numbers we have right now.
Posted on Reply
Add your own comment
Jul 2nd, 2022 07:28 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts