[GH-ISSUE #184] performance optimizations on CUDA 12.9? #104

Closed
opened 2026-10-26 21:12:40 +00:00 by bsakaguchi · 1 comment
Owner

Originally created by @crischeng on GitHub (Sep 4, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepGEMM/issues/184

What performance optimizations were added in CUDA 12.9?
I run test_fp8.py in the H20-96G using CUDA 12.8 and CUDA 12.9, the performance was consistent.
DeepGemm commit f85ec649d7
cuda 12.8:

Image

cuda12.9:

Image
Originally created by @crischeng on GitHub (Sep 4, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepGEMM/issues/184 What performance optimizations were added in CUDA 12.9? I run test_fp8.py in the H20-96G using CUDA 12.8 and CUDA 12.9, the performance was consistent. DeepGemm commit f85ec649d76846552cfd637b6c99fb9c985fd9eb cuda 12.8: <img width="800" height="515" alt="Image" src="https://github.com/user-attachments/assets/19bee47d-935a-44ed-9350-a19885482e2b" /> cuda12.9: <img width="897" height="488" alt="Image" src="https://github.com/user-attachments/assets/a1166cb0-da5a-407e-9d42-0cc4664f09ad" />
Author
Owner

@LyricZhao commented on GitHub (Sep 10, 2025):

NVCC in CUDA 12.9 has an optimization for better tensor core/CUDA core overlapping. This will affect H800 (1979 peak TFLOPS) devices more. I guess as H20 has much lower peak TFLOPS, so it has no effect.

<!-- gh-comment-id:3274213868 --> @LyricZhao commented on GitHub (Sep 10, 2025): NVCC in CUDA 12.9 has an optimization for better tensor core/CUDA core overlapping. This will affect H800 (1979 peak TFLOPS) devices more. I guess as H20 has much lower peak TFLOPS, so it has no effect.
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepGEMM#104
No description provided.