Apple’s M2 Ultra SoC is benchmarked right and left and we can compare these numbers against AMD, Intel and NVIDIA chips to see if the new SoC really delivers the required performance numbers.
Apple M2 Ultra slower than last year’s processors from Intel and AMD, 50% slower than NVIDIA RTX 4080
Apple M2 Ultra SoC is the flagship design offered by the company. Features a total of 134 billion transistors, 24 CPU cores running in a configuration of 16 (performance) and 8 (efficiency), up to 76 GPU cores, 192GB of unified memory and accessible across the entire chip at a maximum bandwidth of 800GB / second. There’s also the new upgraded Neural Engine with 32 cores and the company claims 20% speedup on the CPU and 30% on the GPU side.
The performance benchmarks of the Apple M2 Ultra SoC have been leaked inside Geekbench 6 and the older Geekbench 5. We’re getting a taste of both the CPU and GPU performance since the computing scores were also leaked. So let’s start from the CPU side, first we have clock speeds that range from 3GHz to 3.7GHz. Obviously, we can’t say for sure what the performance and efficiency cores are because the benchmark doesn’t separate cores based on their architecture or states.
The CPU managed a single-core score of 2,809 points in the single-core and 21,531 points in the multi-core tests. For comparison, the Intel Core i9-13900KS scored 3083 points while AMD’s Ryzen 9 7950X scored 2875 points. In multi-threaded benchmarks, the same chips score 21,665 and 19,342 points, respectively. As you can see, the Apple M2 Ultra SoC at the workstation level isn’t any faster than mainstream CPU offerings from Intel and AMD.
If you compare the chip to something like an AMD Threadripper and an Intel Xeon W chip this chip will absolutely crush the M2 Ultra in multi-threaded tests but the single-threaded lead will be over Apple’s M2 Ultra because AMD and Intel Pro CPUs are designed for heavy multi-threaded workloads and that’s where They expand it their multi-core drive to.
Moving on to Compute benchmarks, here’s where I saw comparisons being made to NVIDIA RTX 4080 and AMD GPUs under a Metal API that’s mostly optimized for Apple SoCs. NVIDIA and AMD GPUs are not known for their metal optimizations because the market share for Apple OS based GPUs is very small and non-existent. As such, in OpenCL the M2 Ultra SoC ends up 50% slower than NVIDIA’s RTX 4080 and that’s not even the main GPU.
In its most popular OpenCL API, the Apple M2 Ultra ends up close to the NVIDIA GeForce RTX 4060 Ti which is a mainstream graphics card and yes, that’s the full 76 GPU core configuration of the M2 Ultra we’re comparing.
Once again, these benchmarks prove that it’s wise to take the numbers and performance provided by Apple (and all other companies for that matter) with a grain of salt and look at the independent results to see if the numbers match their claims. We only used mainstream parts for comparison in these benchmarks, and using workstation-oriented solutions will tilt the results more in favor of Intel, NVIDIA, and AMD who have more powerful CPU and GPU designs being offered to the Pro and Workstation users out there. .
Moreover, Apple’s M2 Ultra SOC-based Mac Pro offerings will come with a price The base price is 7,000 USD And that doesn’t even use the full SoC specification or the unified 192GB memory.
“Certified food guru. Internet maven. Bacon junkie. Tv enthusiast. Avid writer. Gamer. Beeraholic.”