6 ms·
I believe they mean GPU threads. Plenty of cuda files in their repository.
by Calavar 4mo ago
I believe they mean GPU threads. Plenty of cuda files in their repository.
- pixelesque 4mo agoFair enough, but that's then only absolutely max 1024 threads per SM, which wouldn't get anywhere near 1 million, given 5090 only has 192 SMs... Future proofing I guess...
- cyber_kinetist 4mo agoYou can launch much more logical threads than the available physical threads. The GPU scheduler will automatically dispatch the work to the SMs.
- zipy124 4mo agoI guess they never say that they execute at the same time technically haha
- ks6g10 3mo agoJust like 2 threads can execute on the same core at the "same" time, i.e. no synchronization, the same is true for GPU threads/ thread groups.