Apple’s new M1 Ultra chip isn’t a PC killer, it’s just a sign of great things to come

Apple’s new M1 Ultra chip isn’t a PC killer, it’s just a sign of great things to come

Nvidia, AMD, and Intel were in something of a multi-core GPU race. Theoretically, if you take one powerful chip and seamlessly glue it to another, you’ll get something twice as good. Simple, right? Well, it’s not that easy, and while AMD has managed to make this concept work for its MI200 high performance supercomputing accelerator, no one else has anything else to share just yet.

M1 Ultra is another revolutionary product for Apple Silicon that will shake the PC industry again.

                  Johnny Srouji, Senior Vice President, Apple

Well, until Apple introduced its new M1 Ultra System-on-Chip (SoC) system-on-a-chip.

Combining two SoC M1 Max, released at the end of last year, new Apple M1 ultra combines their many CPU and GPU cores into one package. This means a 20-core Arm-based CPU, a 64-core GPU, and a 32-core Neural Engine are all under the same roof. This amounts to a chip with 114 billion transistors. You can then configure up to 128 GB of storage on the side.

For reference, the Nvidia GeForce RTX 3090 has a total of 28.3 billion transistors. Sure, the Apple M1 Ultra is CPU, GPU, and I/O all in one package and then doubled through the interconnect, but in fact, Apple threw a lot of transistors at a computational problem to solve it.

UltraFusion

Key to this is the Ultra chip, which Apple calls “UltraFusion”; its new packaging architecture. It is essentially a strong bond of 10,000 signals along the edge of each of the chips, which is placed there during the packaging process. This enables high-speed communication between two connected chips up to 2.5TB/s. Which is a big number for any understanding.

Interconnect itself is not a completely new concept and Intel and AMD have their own high bandwidth interconnects, but Apple’s version definitely shows that it is doing its best to keep up to date with the latest news from other big players in the chip design space. .

“M1 Ultra is another breakthrough for Apple Silicon that will shake up the PC industry once again. By combining two M1 Max dies with our UltraFusion packaging architecture, we can scale Apple Silicon to unprecedented new heights,” said Johnny Srouji, Apple’s senior vice president of hardware technology. “With a powerful processor, a powerful GPU, an incredible Neural Engine, ProRes hardware acceleration, and a massive amount of combined memory, the M1 Ultra completes the M1 family as the world’s most powerful and feature-packed personal computer chip.”

Now the Apple M1 Ultra chip is not a game changer in the sense of changing games, really. Of course, you can run games on an Apple machine, but this GPU is in no way designed for that.

Noncommittal

The company is also once again incredibly noncommittal about the exact tests it used to show its relative performance/watt here – all we know is that it used “select standard tests” and that its “performance data from popular discrete GPUs tested.” from Core i9-12900K with DDR5 memory and GeForce RTX 3060 Ti. Highest discrete GPU performance data tested on Core i9-12900K with DDR5 memory and GeForce RTX 3090.”

However, Apple claims that this chip is capable of outperforming Nvidia’s GeForce RTX 3090 – technically Nvidia’s best card since the RTX 3090 Ti is not currently shown – under certain conditions and with much lower power consumption.

M1 Max

That’s a hell of a big statement, but as we saw with the M1 Max, which was supposed to be about as good as Nvidia’s GeForce RTX 3080, the reality is that there are caveats to everything. This is especially true if you’re a PC gamer looking for gaming performance. While Apple’s chip will be pretty darn good at a lot of things, gaming isn’t really what it’s designed for. Whereas the Nvidia Ampere architecture is more or less true.

Even with a more general performance metric, TFLOP, the M1 Ultra is still slightly behind the RTX 3090 with 35.58 TFLOP FP32. The M1 Max was roughly rated at 10.4 TFLOPs, and if you were to exactly double that (as is the case with two M1 Max M1 Ultra dies spliced ​​together), you would reach 20.8 TFLOPs. Slightly lower, even considering that TFLOPs are not a direct measure of actual performance.

However, this energy efficiency is very impressive. Apple is once again implementing TSMC’s 5nm process here, which is another advantage for the company and is sure to take it into new energy efficiency territory. Intel, AMD, and Nvidia have yet to use a comparable technology node at scale.

Features of the Apple M1 family

Features of the Apple M1 family
M1 M1 Pro M1 Max. M1 Ultra
transistors 16B 33.7 billion 57B 114B
Process node 5 nm 5 nm 5 nm 5 nm
CPU Cores (High Performance + High Efficiency) 4+4 Up to 8+2 8+2 16+4
GPU cores up to 8 up to 16 up to 32 up to 64
GPU ALU 1024 2048 4096 8192

And if Apple can get its SoC with two GPUs to be treated as a single chip by the system, that’s impressive too. Therein lies the real difficulty in creating a multi-chip GPU. It was extremely difficult to make these discrete chips look like a single unit in the system and did not require special programming. At least for anything that doesn’t just perform raw computational tasks.

We don’t need another SLI/CrossFire situation here – where game developers or Nvidia/AMD are largely responsible for multiple GPUs working in tandem – multi-GPU GPUs should be treated as one and work as one in every sense. and purposes.

In terms of CPU performance, Intel and Apple now have the equivalent of a blood feud company. So you can imagine neither side has lost the love. Apple focused on comparing it to the Intel Core i9 12900K, with its unspecified benchmark results being about as useful as a lead bouncy castle. But claiming almost double the performance at 60W. It’s likely that the M1 Ultra’s 16 high-performance cores. And four power-efficient cores are capable of demonstrating Intel’s edge in some performance. And benchmarks, though further research is needed to really see how the two chips lose in performance.

M1

The M1 Ultra is a chip that no doubt looks great on paper and is likely to work great for the kind of workloads Apple designed it for – the workstation’s creative space. However, we need to see how it behaves in real tests (where tests and test conditions are actually specified).

However, I think you can look at what Apple has been able to do with its own multi-chip SoC as a promising sign of what’s to come for PC gaming. Intel has been working on tessellated SoC designs that combine interconnected Arc graphics chiplets and next-generation CPU architectures, starting with Meteor Lake in 2023. Whereas AMD is apparently not far off stacking CPU VRAM and multi-core GPUs. Nvidia is also said to be gearing up for a massive increase in transistor count (and power) with its Lovelace and Hopper architectures.

GPU development

We’re on the cusp of a very exciting time in GPU development, and Apple’s M1 Ultra is a glimpse of what’s to come from a bunch of companies now vying for performance excellence through complex designs and cutting-edge technology nodes.

And it would be remiss of me not to discuss the price of the Apple M1 Ultra chip. The Apple chip is part of Mac Studio, Apple’s new fancy desktop box. With a full size M1 Ultra inside and 128GB of storage, you’re looking at Package for $5799. And that’s just with a 1TB SSD. That’s $7,999 for the 8TB model. You can drop that price down to $3,999 if you ditch the top-tier M1 Ultra in favor of a 48-core GPU model and go for Only 64 GB of memory.

So consider the M1 Ultra a high-end processor. Apple also recently added height adjustment to its monitor stand and added another $400 to the price for the perk. Some things never change.

 

You may also like...

Leave a Reply

Your email address will not be published. Required fields are marked *