This page was machine-translated and may differ from the original. View original
Graphcore's AI semiconductor 'Bow IPU', the first application of TSMC's 3D packaging technology WoW
40% improved performance and 16% superior power efficiency compared to the previous model
Artificial intelligence (AI) semiconductor company Graphcore has unveiled the first AI processor featuring TSMC's 3D packaging technology, 'Wafer-on-Wafer (WoW),' offering 40% improved performance and 16% better power efficiency compared to existing models.
Graphcore announced on the 7th that it recently launched the world's first 3D wafer-on-wafer (WoW) semiconductor, the 'Bow IPU'.
The Bow IPU is the core of the next-generation Bow POD AI computer system, providing 40% improved performance and 16% better power efficiency compared to existing processors across major AI applications.
In addition, companies can enjoy all these performance improvements at the same price as before without changing the software.
The flagship product Bow POD256 and the extra-large Bow POD1024 support 89 petaflops and 350 petaflops of AI computing, respectively. Through this groundbreaking system, Graphcore plans to address the trend of exponential growth in AI model sizes while simultaneously achieving new innovations in the field of machine intelligence.
The first adoption of the Graphcore Bow system is by the Pacific Northwest National Laboratory (PNNL) under the U.S. Department of Energy, which applies Bow to cybersecurity and computational chemistry applications.
PNNL, which is challenging the limits of machine learning and GNNs (graph neural networks) to solve scientific problems that are difficult to handle with existing technology, has reduced the time required for learning and inference from several days to just a few hours through the GraphCore system.
Cloud service provider (CSP) Cirrascale also offers Bow POD systems as part of its Graphcloud IPU bare metal service.
In addition to this, European CSP G-Core Labs also announced plans to launch Bow IPU cloud instances within the second quarter of 2022.
본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.














