Shibaura Institute of Technology's "Ozaki Scheme II" has been integrated into NVIDIA's CUDA Toolkit 13.4, enabling AI-focused GPUs to perform high-precision scientific calculations. This software-based method allows existing GPU hardware, originally optimized for lower-precision AI tasks, to be utilized for essential double-precision (FP64) matrix computations. The Ozaki Scheme II, developed by Professor Katsuhisa Ozaki and colleagues, builds upon the earlier Ozaki Scheme I and uses mathematical techniques to achieve FP64 accuracy without requiring new hardware, thereby improving both performance and power efficiency for scientific simulations. AI
IMPACT Enhances scientific simulation capabilities on AI-optimized hardware, potentially accelerating research in fields like weather forecasting and material science.
RANK_REASON Integration of a novel computational method into a major GPU development platform. [lever_c_demoted from significant: ic=1 ai=0.7]
Read on Mastodon — mastodon.social →
- Ampere
- CUDA
- CUDA Toolkit 13.4
- Fugaku-NEXT
- Fugaku Supercomputer
- Katsuhisa Ozaki
- Nvidia
- Nvidia B200
- Nvidia Rubin Gpu
- Ozaki Scheme I
- Ozaki Scheme II
- RIKEN
- Satoshi Matsuoka
- Shibaura Institute of Technology
- The International Journal of High Performance Computing Applications
- Toshiyuki Imamura
- Yuuki Uchino
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
