4 ms·
Sure, but less than half the memory bandwidth and inference is largely bandwidth bound.
by sliken 2y ago
Sure, but less than half the memory bandwidth and inference is largely bandwidth bound.
- kllrnohj 2y agoThe context was people that don't care about AI or Nvidia GPUs and just want an AArch64 system. So inference performance is irrelevant here > lots of people will buy these machines to get an AArch64 Linux workstation—even if they are not interested in AI or Nvidia GPUs.
- sliken 2y agoAh, sure the Oryon cores are decent, GPU is a bit weak, even compared to the normal cheap/low end APUs.