{"slug":"tackling-gpu-bottlenecks-in-plane-wave-dft-with-ee8dff57","url":"https://silinex.xyz/wire/tackling-gpu-bottlenecks-in-plane-wave-dft-with-ee8dff57","json_url":"https://silinex.xyz/wire/tackling-gpu-bottlenecks-in-plane-wave-dft-with-ee8dff57.json","md_url":"https://silinex.xyz/wire/tackling-gpu-bottlenecks-in-plane-wave-dft-with-ee8dff57.md","headline":"Tackling GPU Bottlenecks in Plane-Wave DFT with K-Point Batching and Mixed Precision","headline_zh":"应对平面波DFT中的GPU瓶颈：K点批处理与混合精度","title":"Tackling GPU Bottlenecks in Plane-Wave DFT with K-Point Batching and Mixed Precision","summary":"According to public reporting via NVIDIA Developer Blog, training a next-generation machine learning interatomic potential (MLIP) requires ground truth: tens to hundreds of millions of density functional theory (DFT)...","summary_zh":"据NVIDIA开发者博客的公开报道，训练下一代机器学习原子间势（MLIP）需要基准真值：数千万到数亿次密度泛函理论（DFT）计算……","date_published":"2026-11-18T15:00:00.000+00:00","date_modified":"2026-11-18T16:21:54.792+00:00","author":{"type":"Organization","name":"NVIDIA Developer Blog","url":"https://developer.nvidia.com/blog/tackling-gpu-bottlenecks-in-plane-wave-dft-with-k-point-batching-and-mixed-precision/"},"publisher":{"name":"Silicon Nexus","url":"https://silinex.xyz/"},"category":"chips","language":"en","source_url":"https://developer.nvidia.com/blog/tackling-gpu-bottlenecks-in-plane-wave-dft-with-k-point-batching-and-mixed-precision/","source_name":"NVIDIA Developer Blog","key_facts":["Training a next-generation machine learning interatomic potential (MLIP) requires ground truth: tens to hundreds of millions of density f…","Plane-wave DFT, with VASP as its dominant implementation, is the gold standard for generating that data.","It is accurate, well-validated across materials classes, and ubiquitous at the national labs and research institutions running these calc…"],"key_facts_zh":["训练下一代机器学习原子间势（MLIP）需要真值：数千万到数亿个密度泛函……","平面波DFT（密度泛函理论），以VASP为其主流实现，是生成此类数据的黄金标准。","它准确无误，在多种材料类别中经过充分验证，并且在运行这些计算的国家实验室和研究机构中无处不在……"],"tldr":["Training a next-generation machine learning interatomic potential (MLIP) requires ground truth: tens to hundreds of millions of density f…","Plane-wave DFT, with VASP as its dominant implementation, is the gold standard for generating that data.","It is accurate, well-validated across materials classes, and ubiquitous at the national labs and research institutions running these calc…"],"tldr_zh":["训练下一代机器学习原子间势（MLIP）需要真值：数千万到数亿个密度泛函……","平面波DFT（密度泛函理论），以VASP为其主流实现，是生成此类数据的黄金标准。","它准确无误，在多种材料类别中经过充分验证，并且在运行这些计算的国家实验室和研究机构中无处不在……"],"entities":[],"signal_type":"other","impact_score":0,"key_metrics":[],"body_md":"## Wire brief\n\nAccording to public reporting via NVIDIA Developer Blog, training a next-generation machine learning interatomic potential (MLIP) requires ground truth: tens to hundreds of millions of density functional theory (DFT)...\n\nIn related context, plane-wave DFT, with VASP as its dominant implementation, is the gold standard for generating that data. It is accurate, well-validated across materials classes, and ubiquitous at the national labs and research institutions running these calculations at scale. As MLIP training datasets have grown, so has the pressure to generate DFT data faster and cheaper.\n\nReporting also notes that but the small unit cells and dense k-point grids common in MLIP data generation can leave significant GPU performance unused. VASP’s OpenACC implementation has delivered meaningful speedups, especially for large-system calculations. However, for smaller systems, sequential k-point processing limits GPU utilization. At the same time, AI workloads are reshaping GPU hardware development toward lower-precision data types, creating a growing gap between what modern NVIDIA GPUs can deliver and what FP64-only HPC codes are optimized for.\n\nPoints still worth watching: This post introduces two new NVIDIA capabilities that address these bottlenecks.\n\n---\n\nSource desk: **NVIDIA Developer Blog** · [origin link](https://developer.nvidia.com/blog/tackling-gpu-bottlenecks-in-plane-wave-dft-with-k-point-batching-and-mixed-precision/)\n","body_md_zh":"## 电讯正文\n\n据NVIDIA开发者博客的公开报道，训练下一代机器学习原子间势（MLIP）需要基准真值：数千万到数亿次密度泛函理论（DFT）计算……\n\n在相关背景下，平面波DFT以VASP为其主要实现，是生成此类数据的黄金标准。它精度高、在各类材料体系中经过充分验证，并且在规模化运行这些计算的国家实验室和研究机构中无处不在。随着MLIP训练数据集不断增大，以更快、更低成本生成DFT数据的压力也随之增加。\n\n报道还指出，但MLIP数据生成中常见的小晶胞和密集k点网格可能导致大量GPU性能未被利用。VASP的OpenACC实现已带来显著加速，尤其是对大规模体系计算。然而，对于较小体系，顺序k点处理限制了GPU利用率。与此同时，AI工作负载正推动GPU硬件开发转向更低精度的数据类型，导致现代NVIDIA GPU所能提供的性能与仅针对FP64优化的HPC代码之间存在日益扩大的差距。\n\n仍值得关注的要点：本文介绍了两项应对这些瓶颈的新NVIDIA能力。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---\n\n来源台：**NVIDIA Developer Blog** · [原文链接](https://developer.nvidia.com/blog/tackling-gpu-bottlenecks-in-plane-wave-dft-with-k-point-batching-and-mixed-precision/)\n","observations":{"human":27,"bot":7,"crawlers":[{"crawlerId":"AhrefsBot","hits":3,"lastSeenAt":"2026-11-19T13:59:18.576Z"},{"crawlerId":"ClaudeBot","hits":2,"lastSeenAt":"2026-11-20T09:57:29.989Z"},{"crawlerId":"Googlebot","hits":2,"lastSeenAt":"2026-11-19T16:17:51.447Z"}],"last_seen_at":"2026-11-20T09:57:29.989Z","note":"Counts are request logs classified by User-Agent. User-Agents can be spoofed. This is not proof of search-engine or LLM inclusion."}}