Thread

Apple M6 Pro achieves the highest single-core CPU score in Geekbench 7

118 points · 145 comments · gainsurier

GeekyBear · 1789810607 · focus · HN ↗
There are also leaks for the M5 Ultra & Base M6.
M5 Ultra CPU:
<a href="https://browser.geekbench.com/search?k=parkdale_cpu&q=mac17%2C15" rel="nofollow">https://browser.geekbench.com/search?k=parkdale_cpu&q=mac17%...
M5 Ultra GPU:
<a href="https://browser.geekbench.com/search?k=grand_gpu&q=mac17%2C15" rel="nofollow">https://browser.geekbench.com/search?k=grand_gpu&q=mac17%2C1...
Base M6 CPU:
<a href="https://browser.geekbench.com/search?k=parkdale_cpu&q=mac18%2C5" rel="nofollow">https://browser.geekbench.com/search?k=parkdale_cpu&q=mac18%...
Base M6 GPU:
<a href="https://browser.geekbench.com/search?k=grand_gpu&q=mac18%2C5" rel="nofollow">https://browser.geekbench.com/search?k=grand_gpu&q=mac18%2C5
The Base M6 CPU single core is averaging a bit over 4000 on Geekbench 7.
For comparison, the AMD Ryzen 9 9950X3D2 averages 3161 on the same test.
<a href="https://browser.geekbench.com/processors/amd-ryzen-9-9950x3d2" rel="nofollow">https://browser.geekbench.com/processors/amd-ryzen-9-9950x3d...
The Intel Core Ultra 7 270K Plus averages 2940 on the same test.
<a href="https://browser.geekbench.com/processors/intel-core-ultra-7-270k-plus" rel="nofollow">https://browser.geekbench.com/processors/intel-core-ultra-7-...
1. revolvingthrow · 1789816318 · focus · HN ↗
  I don’t understand those benchmarks.
  According to geekbench 7800x3d is 2400 single core, 15500 multicore. M4 pro is 3350 single core and 24750 multicore. Yet when I convert video using libsvtav1 with ffmpeg I’m getting noticeably faster performance on the desktop. And that’s with mbp, which doesn’t thermally throttle within 15 seconds.
  Is it the 96mb cache? Avx-512? Are benchmarks bullshit when comparing different architectures?
  1. BoingBoomTschak · 1789818912 · focus · HN ↗
    I think software with a strong focus on SIMD really favours x86 over Apple, since AVX2 support was pretty uniform and AVX-512 (except Intel's fiasco) has been supported for a while, while the post-NEON landscape is strange in ARM country.
    From what I've been able to gather, Apple only supports SME and the small "streaming" part of SVE2 required by SME since the M4. Which seems to be mostly useless for video encoding (I only see NEON/SVE in <a href="https://gitlab.com/AOMediaCodec/SVT-AV1/-/tree/master/Source/Lib" rel="nofollow">https://gitlab.com/AOMediaCodec/SVT-AV1/-/tree/master/Source... or <a href="https://github.com/Multicorewareinc/x265/tree/master/source/common" rel="nofollow">https://github.com/Multicorewareinc/x265/tree/master/source/...) it's basically a GEMM engine.
    So basically, video encoding is a bad show for Apple who seem to be saying "use crappy hardware encoding and buy amd64 if you need more".
    Very hard to find benchmark data. Found <a href="https://openbenchmarking.org/vs/Processor/Apple+M4+Pro,AMD+Ryzen+9+9900X+12-Core" rel="nofollow">https://openbenchmarking.org/vs/Processor/Apple+M4+Pro,AMD+R... that shows a good x265 performance at 1080p but not 4K. Guess the wider SIMD units matter more there.
    api · 1789821184 · focus · HN ↗
    X64 does have a stronger SIMD story, but this is partly compensated for by the fact that ARM64 is much easier to decode wide. M series has a very wide decoder and a lot of instruction level parallelism. It can turn chunks of those 128-bit SIMD operations into what amounts to wider operations.
    AVX still wins though.
    M series still wins on performance per watt and now apparently leads on general purpose code.
    All these leading edge chips are very good. We have an embarrassment of riches when it comes to blistering fast chips here.
    Archit3ch · 1789831361 · focus · HN ↗
    > AVX still wins though.
    On what, microbenchmarks? I have realtime audio workloads where the hot loop is essentially linear algebra. Exactly the kind of work that suits AVX2/AVX512. Guess what, Apple Silicon still pulls ahead because real workloads are branchy, cache-hungry, full of dependencies and do not line up in 8 neat f64 operations per cycle.

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.