Shibaura Tech's Ozaki Scheme II lands in CUDA 13.4, squeezing FP64 from AI-focused GPUs
udmrzn · x · 2026-10-01
Ozaki Scheme II, a numerical method by Prof. Katsuhisa Ozaki at Shibaura Institute of Technology, has been adopted into NVIDIA's CUDA Toolkit 13.4 (Sept 2026), following the original scheme's 2025 inclusion.
Background: GPU double-precision (FP64) performance has collapsed as chips optimize for AI low-precision math — B200 offers 40 TFLOPS FP64 while Blackwell Ultra (B300) drops to 1.2–1.3 TFLOPS, a 30x decline.
How it works: The scheme mathematically combines multiple low-precision operations (error-free transformations, modular arithmetic via the Chinese Remainder Theorem) to reproduce double-precision accuracy in software, running on Ampere-and-later GPUs with no extra hardware — repurposing idle low-precision units for scientific computing.
Impact: Useful for weather forecasting, seismic design, quantum chemistry and materials science. RIKEN lists it among methods for the Fugaku NEXT supercomputer, and it's reportedly used in US DOE science programs.
More from Infra
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16
- vLLM serves DeepSeek V4.1-Flash at 954K tokens/s per GPU, $38.96/GPU-hour on GB300 NVL72 — bookwormengr · 2026-11-16
- Colossus 2's 1.2M B200s could generate ~$400B/year in tokens at DeepSeek Flash rates — bookwormengr · 2026-11-16
- BIS issues rule blocking remote access to US AI chips via third-country data centers — teortaxesTex · 2026-11-16
- Qdrant 1.20.1: 50% fewer cold-storage reads, ~2x faster quantization, agent-ready setup — qdrant_engine · 2026-11-16
- One 16GB GPU runs songs, voice, images, 720p video and local RAG — all offline and open source — llama-of-death · 2026-11-16