We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hacker News·Jun 24·5 min read
LineShine Debuts at No. 1 as the TOP500 Enters a New Global Exascale Era
TL;DR
Supercomputers have been limited in their performance, with few exceeding one exaflop of sustained computing power. LineShine, a new supercomputer in China, achieved 2.198 Exaflop/s using a custom processor and the LingKun platform.
✦ Why It Matters
Engineers can leverage insights from LineShine's architecture to enhance their own high-performance computing designs.
Key Takeaways
- Explore custom processor designs for high-performance computing.
- Consider energy efficiency metrics like Gigaflops/Watt in system design.
- Stay updated on global supercomputer rankings for competitive insights.
- Investigate the implications of exascale computing on research capabilities.
Full Summary
The TOP500 list ranks the world's most powerful supercomputers, and the latest edition reveals a shift in leadership with the debut of LineShine at No. 1.
Built at the National Supercomputing Centre in Shenzhen, LineShine utilizes a custom Chinese processor and the LingKun platform, featuring 13.79 million cores across 304-core LX2 processors. It achieved 2.198 Exaflop/s on the High Performance Linpack (HPL) benchmark, representing about 80 percent of its theoretical peak performance of 2.736 Exaflop/s.
This system also leads the HPCG ranking with 22.00 HPCG-Petaflop/s. The introduction of LineShine increases the number of supercomputers exceeding one exaflop/s on HPL from four to five, showcasing a global exascale capability across Asia, North America, and Europe.
This development highlights the growing competition in high-performance computing and the potential for advancements in computational research.