Hacker News Reader

According to this paper[0] ~1600GB/s raw memory bandwidth, ~590GB/s in practice. I haven't seen any actual benchmarks of M2 Ultra CPU bandwidth, but the M1 Ultra has been benchmarked at ~240GB/s[1].

[0] https://www.ixpug.org/images/docs/ISC23/McCalpin_SPR_BW_limi...

[1] https://macperformanceguide.com/blog/2022/20220618_1934-Appl...

View on HN

> The cores used in the Xeon Max processors do not support enough concurrency to reach full HBM bandwidth with the current core counts. 2x more concurrency or 2x more cores.

is a good punchline from the report. The Apple cores have that problem but a lot worse. They are slow. It goes back to this flawed idea in the community that you can "just" "add" "more memory," when parts like the H100 have their memory size matched to the architecture (physical and software) of the dozens of CPUs on them.

I'm not sure why the conception persists that a 60W laptop part would be comparable to a 300W server part of the same process generations, let alone this particular part.

That's insane.

Reply on news.ycombinator.com