DLSS 5 neural rendering on a Tesla V100: pure PyTorch path skips NGX entirely
Ancient-Tomorrow-871 · reddit · 2026-09-26
A developer got DLSS 5 neural rendering running in ComfyUI on a Volta V100 by bypassing NVIDIA's NGX/D3D12 runtime entirely — the network runs as plain PyTorch. Measured 0.700s/frame at 1080p with 1732MB VRAM; FP8 rounding works on SM 7.0 since it's rounding, not matmul. Open-sourced with caveats: it's a reimplementation, quality is DLSS-5-like not equal.
More from Infra
- Running Qwen2.5-Coder-3B on Android: Jupyter + Pydantic Evals on-device eval pipeline — alex_b3b · 2026-11-11
- 22% of Nvidia revenue now books to Singapore, but only ~1% is local buyers — Gregory_C_Allen · 2026-11-11
- Huawei's Kirin 9050 NPU claims offline 30B MoE on smartphones, +140% perf — AIFlow_ML · 2026-11-11
- Stratechery: Intel vs AMD's diverging earnings and Intel's need for an old-new path — Stratechery · 2026-11-11
- a16z Podcast: Mike Volpi on What the 1990s Internet Boom Teaches Us About the AI Buildout — a16z Podcast · 2026-11-11
- Alibaba files to list chip-design unit T-Head in Hong Kong to fund AI push — pstAsiatech · 2026-11-11