13°

Reels · 2 of 50

1:15

Reel · 1:15

NVFP4 without Blackwell: I ran NVIDIA's Qwen-Image on an RTX 4070 Ti SUPER and the NF4 grain was in 8 matrices

Narrated: a loupe goes down from a grainy Qwen-Image NF4 photo. The obvious suspect, the 112 modulation matrices, turns out innocent: switching them to NVFP4 leaves the grain. The culprit is 8 matrices in the edge blocks; keeping them in bf16 lowers noise in 19 of 21 images at the same time per image.

Length
1:15
Published

IABenchmarksInferencia localCuantización

This reel sums up NVFP4 without Blackwell: I ran NVIDIA's Qwen-Image on an RTX 4070 Ti SUPER and the NF4 grain was in 8 matrices, where the method, the tables and what did not work are.

Read the articleOpen in the viewer