Instruction throughput/latency documentation updates for Blackwell (CC 10.0 and CC12.0)

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m unable to find any updates on Compute Capabilities 10.0 and 12.0 for instruction throughput and latencies for Blackwell GPUs despite the official release of the hardware and CUDA 12.8. E.g. the tables in ‘CUDA C++ Programming Guide - 5.4. Maximize Instruction Throughput’ only go up to CC 9.0.

Is there somewhere to find this information or perhaps CUDA docs team have accidentally omitted an update?

2

you may wish to file a bug requesting a doc update.

1 Like