Krea 3 in 3.3 seconds on AMD.
@hedra_labs made our diffusion transformer 6.9x faster on MI355X at 98.3% of the original quality.
if you're running image editing in production - talk to us below 👇
Krea 3 runs 6.9x faster on AMD MI355X at 98.3% of original quality
AIKrea reports that Hedra Labs made the diffusion transformer behind Krea 3 6.9 times faster on AMD Instinct MI355X, cutting generation time from 22.9 seconds to 3.3 seconds. The post says MXFP6 GEMM runs 2.8 times faster than BF16, GPU transfer waiting fell from 18% to 2% of time, and output keeps 98.3% of BF16 quality by SSIM. The work was done with Hedra's autonomous research system.
@krea_ai
We made @krea_ai Krea 3's diffusion transformer 6.9× faster on @AMD Instinct MI355X: 22.9s → 3.3s. Key stats: → 2.8× faster MXFP6 GEMM vs BF16 → 18% → 2% of time waiting on GPU transfers → 98.3% of BF16 quality (SSIM) Here's the work done by the team using our autonomous research system ↓