3 ms·

A quick read suggests a problem multi-core on the m2. Does anybody have any benchmarks for multi threaded rather than multi process calculations / operations. I

A quick read suggests a problem multi-core on the m2. Does anybody have any benchmarks for multi threaded rather than multi process calculations / operations. I am curious whether this is a hardware bottleneck or a young compiler or a lack of support libraries…

catchnear4321 3y ago
a quick read suggests the author is for a given metric, relying on picking the winner between amd and intel, and then declaring apple inferior. this isn’t an article.
- jleyank 3y ago
  Can you suggest alternative articles that address multi process and multi threaded performance metrics? That would be be very helpful and appreciated.
  - acdha 3y ago
    I’d wait for real experiences to show up - GeekBench isn’t a great benchmark for pro work but it’s very accessible and people who get paid by ad impressions know there’s a big advantage to getting the story out first so we tend to see these things pretty reliably after new hardware comes out. Later, you’ll see stories from people who do real work but by then the attention will have dissipated.
olliej 3y ago
Geekbench is multithreaded. Intel's semi-multithreading on a single core also helps it out in multithreading tests because MT benchmarks are generally ideal for the scenario. I think the single core performance matters for the majority of the apple product line: varying degrees of single core perf bounds, and performance/watt are important to anything with a battery, and that's clearly what they've optimized for. That breaks down for actual wall connected "I am doing numerics" performance that you expect/want from a Mac Pro or Studio (I feel an iMac is still intended as a low power/quiet machine). Personally I was hoping for some improvement on the desktop GPU front - afaict the Mx GPUs do well ... for integrated GPUs. They don't seem to hold a candle to desktop GPUs in PCs.

catchnear4321 3y ago

a quick read suggests the author is for a given metric, relying on picking the winner between amd and intel, and then declaring apple inferior. this isn’t an article.

jleyank 3y ago
Can you suggest alternative articles that address multi process and multi threaded performance metrics? That would be be very helpful and appreciated.
- acdha 3y ago
  I’d wait for real experiences to show up - GeekBench isn’t a great benchmark for pro work but it’s very accessible and people who get paid by ad impressions know there’s a big advantage to getting the story out first so we tend to see these things pretty reliably after new hardware comes out. Later, you’ll see stories from people who do real work but by then the attention will have dissipated.

olliej 3y ago

Geekbench is multithreaded. Intel's semi-multithreading on a single core also helps it out in multithreading tests because MT benchmarks are generally ideal for the scenario. I think the single core performance matters for the majority of the apple product line: varying degrees of single core perf bounds, and performance/watt are important to anything with a battery, and that's clearly what they've optimized for. That breaks down for actual wall connected "I am doing numerics" performance that you expect/want from a Mac Pro or Studio (I feel an iMac is still intended as a low power/quiet machine). Personally I was hoping for some improvement on the desktop GPU front - afaict the Mx GPUs do well ... for integrated GPUs. They don't seem to hold a candle to desktop GPUs in PCs.