Benchmarking Minimax H3 Video Acceleration: 10s Video in 60s on a Single RTX 5090
nik_amaze · reddit · 2026-08-10
A developer on Reddit shared an aggressive local acceleration workflow for the Minimax H3 video generation model. Renting an RTX 5090 (32GB VRAM) with 56GB RAM from Vast.ai, the setup utilizes CUDA 13, SageAttn 2.2.0, and Triton 3.6.0 (installing Sol-Attn via git).
With all acceleration toggles enabled (excluding easyCache due to quality degradation), the int8 unpruned model with two image references generates a 10-second video at 1mp in 180s, and at 0.5mp in 60s. The creator shared the ComfyUI workflow and sought further optimization tips from the community.
Related event: MiniMax H3 on RTX 5090: Efficient Video Generation(6 posts)→
More from Infra
- MLCommons launches MLPerf Endpoints, a benchmark that measures inference endpoints the way buyers actually purchase them — TheKanter · 2026-10-16
- Neocloud M&A decoded: CoreWeave, Nebius, Crusoe, Nscale all bet margins vanish beyond GPU scarcity — AccBalanced · 2026-10-16
- AI-Driven Timing-Aware Power Optimization Uses Learned Ranking With Baseline-Preserving Repair — Amit Sharma · 2026-10-16
- Nscale AI datacenter plagued by coolant leaks, transformer fire, 10 safety violations — saranormous · 2026-10-16
- WSL 3.0 ships with native containers and Linux 6.18 kernel, syscall overhead down ~16% — solyarisoftware · 2026-10-16
- Are 22 GB modded RTX 2080 Ti cards a viable budget local LLM option? — Dave86ch · 2026-10-16