{"slug":"faster-scientific-simulations-with-new-fp64-algo-d12bda83","url":"https://silinex.xyz/wire/faster-scientific-simulations-with-new-fp64-algo-d12bda83","json_url":"https://silinex.xyz/wire/faster-scientific-simulations-with-new-fp64-algo-d12bda83.json","md_url":"https://silinex.xyz/wire/faster-scientific-simulations-with-new-fp64-algo-d12bda83.md","headline":"Faster Scientific Simulations with New FP64 Algorithms in NVIDIA cuBLAS","headline_zh":"NVIDIA cuBLAS 新增 FP64 算法，加速科学模拟","title":"Faster Scientific Simulations with New FP64 Algorithms in NVIDIA cuBLAS","summary":"According to public reporting via NVIDIA Developer Blog, many scientific simulations, from molecular chemistry to materials science, spend the majority of their execution time computing high-precision dense general...","summary_zh":"据NVIDIA开发者博客的公开报道，许多科学模拟——从分子化学到材料科学——大部分执行时间都花在计算高精度稠密通用矩阵乘法上。","date_published":"2026-11-13T22:13:51.000+00:00","date_modified":"2026-11-13T22:21:42.023+00:00","author":{"type":"Organization","name":"NVIDIA Developer Blog","url":"https://developer.nvidia.com/blog/faster-scientific-simulations-with-new-fp64-algorithms-in-nvidia-cublas/"},"publisher":{"name":"Silicon Nexus","url":"https://silinex.xyz/"},"category":"chips","language":"en","source_url":"https://developer.nvidia.com/blog/faster-scientific-simulations-with-new-fp64-algorithms-in-nvidia-cublas/","source_name":"NVIDIA Developer Blog","key_facts":["Many scientific simulations, from molecular chemistry to materials science, spend the majority of their execution time computing high-pre…","These operations are called DGEMM for real 64-bit floating-point numbers and ZGEMM for complex 64-bit floating-point numbers. cuBLAS is w…","Rapid advances in lower-precision hardware have motivated exploration of a new class of algorithms that deliver full double-precision res…","With CUDA Toolkit 13.4.2, cuBLAS no longer has to choose between accuracy and speed. cuBLAS breaks each double-precision number into inte…"],"key_facts_zh":["从分子化学到材料科学的许多科学模拟，大部分执行时间都花在计算高精度……上。","这些运算针对实数64位浮点数称为DGEMM，针对复数64位浮点数称为ZGEMM。cuBLAS是……","低精度硬件的快速进步促使人们探索一类新算法，这类算法能够提供完整的双精度结果……","借助CUDA Toolkit 13.4.2，cuBLAS不再需要在精度与速度之间做出取舍。cuBLAS将每个双精度数拆分为整……"],"body_md":"## Wire brief\n\nAccording to public reporting via NVIDIA Developer Blog, many scientific simulations, from molecular chemistry to materials science, spend the majority of their execution time computing high-precision dense general...\n\nIn related context, these operations are called DGEMM for real 64-bit floating-point numbers and ZGEMM for complex 64-bit floating-point numbers. cuBLAS is widely used by HPC and AI applications to accelerate GEMM and other Basic Linear Algebra Subprograms (BLAS) on NVIDIA GPUs. Rapid advances in lower-precision hardware have motivated exploration of a new class of algorithms that deliver full double-precision results on the faster integer Tensor Cores. With CUDA Toolkit 13.4.2, cuBLAS no longer has to choose between accuracy and speed. cuBLAS breaks each double-precision number into integer pieces, multiplies them on the GPU’s Tensor Cores, and reassembles the exact result.\n\nReporting also notes that the answers keep full double-precision accuracy, but the work runs on faster hardware. Two new algorithms make this more efficient: Ozaki-II needs fewer integer multiplications than the earlier Ozaki-I approach, and the 2M method cuts a complex matrix multiplication from four real multiplications to two. This builds on the Automatic Dynamic Precision (ADP) framework introduced in Unlocking Tensor Core Performance with Floating Point Emulation in cuBLAS , which delivers higher double-precision (FP64) matrix-multiplication performance by mapping FP64 arithmetic onto integer Tensor Cores while preserving native FP64 accuracy. cuBLAS delivers automatic acceleration of DGEMM and ZGEMM with no application code changes. Because speed matters only when accuracy is not compromised, we validate both on real-world applications: coupled cluster singles and doubles (CCSD) in quantum chemistry, and BerkeleyGW in materials science.\n\nPoints still worth watching: This post explains how the algorithms work, shows benchmark and application results, and covers how to get started. This Silicon News Wire item is a paraphrased brief; full wording remains on the origin page.\n\n## Source & notice\n\nThis Silicon News Wire brief paraphrases public reporting from **NVIDIA Developer Blog** for carbon readers and silicon citation. **It is not a verbatim full-text reprint.**\n\nRead the original: [Faster Scientific Simulations with New FP64 Algorithms in NVIDIA cuBLAS](https://developer.nvidia.com/blog/faster-scientific-simulations-with-new-fp64-algorithms-in-nvidia-cublas/)\n","body_md_zh":"## 电讯正文\n\n据NVIDIA开发者博客的公开报道，许多科学模拟——从分子化学到材料科学——大部分执行时间都花在计算高精度稠密通用矩阵乘法上。\n\n在相关语境中，这些运算针对实数64位浮点数称为DGEMM，针对复数64位浮点数称为ZGEMM。HPC和AI应用广泛使用cuBLAS在NVIDIA GPU上加速GEMM及其他基础线性代数子程序（BLAS）。低精度硬件的快速进步，促使人们探索一类新算法，在更快的整数Tensor Core上给出完整的双精度结果。借助CUDA Toolkit 13.4.2，cuBLAS不再需要在精度与速度之间做取舍。cuBLAS将每个双精度数拆分为若干整数片段，在GPU的Tensor Core上相乘，再重组出精确结果。\n\n报道还指出，结果保持完整的双精度精度，但计算运行在更快的硬件上。两种新算法让这一过程更高效：与早期的Ozaki-I方法相比，Ozaki-II所需的整数乘法更少；2M方法将一次复数矩阵乘法从四次实数乘法减少到两次。这建立在《Unlocking Tensor Core Performance with Floating Point Emulation in cuBLAS》中引入的自动动态精度（ADP）框架之上，该框架通过将FP64算术映射到整数Tensor Core，同时保持原生FP64精度，从而提供更高的双精度（FP64）矩阵乘法性能。cuBLAS可自动加速DGEMM和ZGEMM，无需修改应用代码。由于只有在不牺牲精度的前提下速度才有意义，我们在真实应用中验证了二者：量子化学中的耦合簇单双激发（CCSD），以及材料科学中的BerkeleyGW。\n\n仍值得关注的点：本文解释了这些算法的工作原理，展示了基准测试和应用结果，并介绍了如何上手。本Silicon News Wire条目为改写摘要；完整措辞仍保留在原始页面。\n\n## 来源与声明\n\n本电讯由 Silicon News Wire 依据 **NVIDIA Developer Blog** 公开报道做要点改写与润色，便于碳基阅读与硅基引用；**不是原文全文转载**。\n\n阅读原文：[NVIDIA cuBLAS 新增 FP64 算法，加速科学模拟](https://developer.nvidia.com/blog/faster-scientific-simulations-with-new-fp64-algorithms-in-nvidia-cublas/)\n","observations":{"human":0,"bot":2,"crawlers":[{"crawlerId":"ClaudeBot","hits":2,"lastSeenAt":"2026-11-13T23:45:55.864Z"}],"last_seen_at":"2026-11-13T23:45:55.864Z","note":"Counts are request logs classified by User-Agent. User-Agents can be spoofed. This is not proof of search-engine or LLM inclusion."}}