AI Dynamics

Global AI News Aggregator

CUDA Memory Optimization: 511×512 vs 512×512 Performance Analysis

Some people are misreading this — 511×511 was FASTER. It looks like at 512×512 and above it falls to another path that requires internal CudaMalloc/Free calls.

→ View original post on X — @id_aa_carmack,

Commentaires

Leave a Reply

Your email address will not be published. Required fields are marked *