3 ms·

> M1 is ARM. AWS Graviton is ARM Graviton 2 has an absolutely huge number of cores (64), but lacks the memory bandwidth to keep them all fed at once. >the L3

> M1 is ARM. AWS Graviton is ARM Graviton 2 has an absolutely huge number of cores (64), but lacks the memory bandwidth to keep them all fed at once. >the L3 cache of the Graviton2 was shared amongst all its cores, and we also discovered how only 8-16 cores were able to saturate the memory controllers of the system. Scaling linearly across cores might be easy for some workloads, but for anything that even remotely has some kind of memory pressure should see greater slowdowns given that all the threads are competing for the shared L3 and DRAM resources. https://www.anandtech.com/show/15578/cloud-clash-amazon-graviton2-arm-against-intel-and-amd/6 https://www.anandtech.com/show/15578/cloud-clash-amazon-grav... There's more going on than "ARM is ARM".

pmarreck 5y ago
This is an irrelevant criticism. The claim I was originally responding to was that "Neither cloud computing nor 80% of the desktop world run on M1, regardless how great they might be" I think it's pretty clear I shot that all to hell by pointing out that M1 is ARM, and ARM is going everywhere, now and in the future. Now, it's quite possible that the architectural details you point out may be design foibles for most workloads in just the Graviton2 series, but the writing is on the wall, the other 3 walls, the floor, and the ceiling that the near future is not x64 but some combination of ARM and RISC-V