2020 Mac Mini – Putting Apple Silicon M1 To The Test

Post by egsmi » Tue, Nov 17, 2020, 4:37 PM UTC

kllrnohj wrote: post #192

314 wrote: post #109

Earlier quoted context omitted.

The Zen3 processor that they are comparing it to is the 5950x - the fastest desktop processor with a TDP of 105W. The entire system power of the M1 mini under load was 28W. What the article is pointing out is that the mobile low-power version of the M1 (as the mini is really just a laptop in a SFF box) is competitive with the top-end Zen3 chip; the benchmark gap is smaller than 2x. We don't know yet how far the M1 sc…

> The Zen3 processor that they are comparing it to is the 5950x - the fastest desktop processor with a TDP of 105W. The entire system power of the M1 mini under load was 28W. This is a very misleading statement. They primarily only used the 5950X in single-core tests, and in those tests it doesn't come remotely close to 105W. In fact per Anandtech's own results[1] the 5950X CPU core in a single-core load draws around…

Post by jonplackett » Tue, Nov 17, 2020, 4:38 PM UTC

Post by diimdeep » Tue, Nov 17, 2020, 4:42 PM UTC

phire wrote: post #233

formerly_proven wrote: post #203

The Cinebench R23 results seem kinda weird to me. The 5950X would have almost a 40 % clock speed advantage over the M1 (~5 GHz 1C vs 3.2 GHz 1C), yet the M1 is only about 8 % slower in an entirely ALU-limited SIMD benchmark? This suggests the M1 core has like 50 % more EUs and achieves much higher throughput than Zen 3. The SPEC results are... decisive to say the least. Without Zen 3, x86 CPUs would look, well... lik…

You overestimate how ALU limited Cinebench is. I managed to find an AVX vs AVX off benchmark run for Cinebench r20 [1]. Going from 128bit SSE to 256bit AVX and doubling the ALUs only results in a 10-12% increase in performance. I assume this has to do with how each SIMD lane of calculation might need to branch independently, limiting the performance speedup from just throwing wider SIMD ALUs at it. [1] https://www.te…

Post by ric2b » Tue, Nov 17, 2020, 4:42 PM UTC

kllrnohj wrote: post #214

foldr wrote: post #116

Earlier quoted context omitted.

From the article: "While AMD’s Zen3 still holds the leads in several workloads, we need to remind ourselves that this comes at a great cost in power consumption in the +49W range while the Apple M1 here is using 7-8W total device active power." Looking through the benchmarks, the zen 2 parts generally seem to have lower performance than the M1. The cinebench multithreaded benchmark is one exception. It's not that sur…

The Zen2 4800HS also outperformed the M1 in the Specint2017 multi-threaded results, too. The M1's float results are weirdly good relative to the int results, though. Not sure why Apple seems to have prioritized that so much in this category of CPU.

Post by Shivetya » Tue, Nov 17, 2020, 4:43 PM UTC