MiniMax H3 Local Test: RTX PRO 6000 Outperforms B300 by Almost Half in Inference Time
Resident_Sympathy_60 · reddit · 2026-08-05
A developer benchmarked the MiniMax H3 video generation model across different hardware. Using Easy Cache for acceleration and upscaling from 352x640 to 704x1280, the RTX PRO 6000 took 55 seconds, while the B300 took 85 seconds.
The tester tried enabling Sage Attention on the B300, but incompatibility issues increased the time to 101 seconds. Testing Sol Attention yielded no speed improvements. The generated video quality appears consistent across both cards.
Related event: MiniMax H3 Local Tests: Performance and Optimization Across GPUs(13 posts)→
More from Infra
- Trading 2 RTX 5090s for 4 R9700s: does doubled VRAM pay off? — poofph · 2026-10-16
- Running PolyStrata + Cline coding agent on a $0-fancy laptop at ~90 tok/s — rpwoerk · 2026-10-16
- TensorFold 1.0.12 runs 120B vision model on 1-2 DGX Sparks, adds Mac image input — HankYeomans · 2026-10-16
- vLLM brings frontier model local deployment to NVIDIA RTX Spark PCs — vllm_project · 2026-10-16
- NVIDIA signals $2B in new European AI deals at GTC Berlin, totaling ~$10B in a year — CurieuxExplorer · 2026-10-16
- Making Your Data Ready for Agentic AI: trust first, then semantics, then writes — bibryam · 2026-10-16