RTX 5090 Test: SageAttention Nearly Doubles Video Generation Speed
gabxav · reddit · 2026-08-14
The author compared no acceleration, SageAttention, and ComfyUI-Kitchen using a MiniMax H3 video generation workflow on an RTX 5090.
- Speed: Unaccelerated took 335s, SageAttention took 173s (1.94x speedup), and ComfyUI-Kitchen took 181s (1.85x speedup).
- Conclusion: SageAttention was the fastest, beating ComfyUI-Kitchen by about 4.6%.
- Quality: The author provided side-by-side video comparisons, noting that visual quality (detail preservation, motion consistency) is very close between the two, and asked the community for their visual assessments.
Related event: RTX 5090 Test: SageAttention Significantly Boosts Video Generation(2 posts)→
More from Infra
- New hardware day: local LLM deployment the moment it arrives — lxfater · 2026-10-16
- Together AI's score centering fixes RL training collapse under quantized inference — nrehiew_ · 2026-10-16
- Inference is splitting prefill and decode across specialized chips, not just H100s — AccBalanced · 2026-10-16
- Rebasely launches in beta: Git hosting built for humans and coding agents — handotdev · 2026-10-16
- Drama3 TTS lands in audio.cpp: up to 5x real-time on RTX 5090, <8GB VRAM mode too — Acceptable-Cycle4645 · 2026-10-16
- Three CoreWeave co-founders have sold $3.3B in shares since IPO amid credit worries — SumitGup · 2026-10-16