NVIDIA Ampere Architecture

Creator
Creator
Seonglae Cho
Created
Created
2023 May 31 4:49
Editor
Edited
Edited
2025 Jul 3 9:57
Refs
Refs
  • Ampere (3rd Gen): Expanded MMA to full 32-thread warps, added BF16 support, reduced register burden with cp.async asynchronous global→shared memory copies and ldmatrix vectorized loads.
NVIDIA Ampere Architecture Products
 
 
 
 
 
 

Recommendations