Six CUDA kernel optimizations squeeze another 1.78× speedup out of 3DGS training
gkopanas · x · 2026-10-29
gkopanas published a blog showing that, even after combining all community-proposed 3D Gaussian Splatting optimizations, the hardware ceiling hasn't been reached: another 1.78× speedup on the full forward-backward training step is available.
Six self-contained optimizations target how CUDA kernels map onto the GPU's warp execution model, register file, L2 cache sectors, and arithmetic execution pipes. Supported by Gemini 4 Argon.
Related event: DeepMind Engineer Speeds Up 3DGS Rendering 10.5x with CUDA Optimizations(3 posts)→
More from Infra
- Reve publishes technical report on training infra that scaled to 10K GPUs with 10x fewer chips — cantrell · 2026-10-29
- DeepSeek Flash fingerprint change hints at migration to a new GPU cluster — teortaxesTex · 2026-10-29
- Coding agents burn 95-99% cached tokens per benchmark task, token usage chart shows — AccBalanced · 2026-10-29
- OpenRouter's 10T Daily DeepSeek Tokens Are 95% Cache Hits; One GB300 Rack Could Serve It All — AccBalanced · 2026-10-29
- Kling 4.0 video model lands on Cloudflare AI Gateway at $0.157/sec for 720p — ritakozlov · 2026-10-29
- Google dodges TPU margin question, admits AI Mode queries are monetizing poorly — firstadopter · 2026-10-29