4 ms·
What if they are using the GPU for inference.
by wmf 2y ago
What if they are using the GPU for inference.
- forrestthewoods 2y agoStill a waste of CPU. Seems hard to believe that workloads would saturate both GPU and NPU. It’s got to be sub-optimal. I’m just not sure by how much. My ill informed spidey sense says by quite a bit. But I don’t know which is why I asked!