Back to AIBriefs
AnalysisAI Models

Krea 2 Turbo int8 1.9x faster than fp8 on 3090

Krea 2 Turbo int8 1.9x faster than fp8 on 3090View on Reddit

Reddit benchmark shows int8 precision yields ~1.9× speedup over fp8 for Krea 2 Turbo on an RTX 3090 (8 steps, same sampler and seed). Both use the same prompt and resolution.

·
Jun 24, 10:33 PM
Krea 2 Turbo int8 1.9x faster than fp8 on 3090 — AIBriefs