New AI servers can boost performance into supercomputing territory with a smaller footprint. Credit: dny59 / kentoh / Getty Images Graphcore, the British semiconductor company that develops accelerators for AI and machine learning, has greatly increased the performance of its massively parallel Intelligence Processing Unit (IPU) servers. Graphcore sells its AI-oriented chips in rack-mounted designs called IPU-PODs. Up until now, the maximum per cabinet has been 64 units (in the IPU-POD64). The newest racks are twice and four times as large: the IPU-POD128 and IPU-POD256. With 32 petaFLOPS of AI compute in the IPU-POD 128 and 64 petaFLOPS in the IPU-POD 256, Graphcore says it can now extend its reach into AI supercomputer territory, and with a much smaller footprint than the typical supercomputer, which could fill a basketball court. The IPU-POD systems disaggregate the AI compute from the servers, which means different types of AI workloads requiring different levels of performance can all be run on the same POD. For example, a POD can be allocated for faster training of large Transformer-based language models across an entire system, or the system can be divided into smaller, flexible vPODs to give more developers IPU access. This is done through Graphcore’s Poplar software stack, which includes an SDK, the Graphcore Communication Library (GCL) for managing communication and synchronization between IPUs, and PopRun and PopDist, which allow developers to run their applications across multiple IPU-POD systems. For intra-rack IPU communication, earlier PODs used 64Gb/s IPU-Links. Now, the IPU-POD128 and IPU-POD256 use new Gateway Links, a horizontal, rack-to-rack connection that extends IPU-Links using tunneling over regular 100Gb Ethernet. Both systems have been developed for cloud hyperscalers, national scientific computing labs and enterprise companies with large AI teams in markets like financial services or pharmaceutical. Graphcore’s initial customers include the U.S. Department of Energy’s Lawrence Berkeley National Laboratory and J.P. Morgan. IPU-POD16, IPU-POD64, IPU-POD128 and IPU-POD256 are shipping to customers today from French IT giant ATOS and other systems integrator partners around the world and are available to buy in the cloud from Cirrascale. Related content news High-bandwidth memory nearly sold out until 2026 While it might be tempting to blame Nvidia for the shortage of HBM, it’s not alone in driving high-performance computing and demand for the memory HPC requires. By Andy Patrizio May 13, 2024 3 mins CPUs and Processors High-Performance Computing Data Center news CHIPS Act to fund $285 million for semiconductor digital twins Plans call for building an institute to develop digital twins for semiconductor manufacturing and share resources among chip developers. By Andy Patrizio May 10, 2024 3 mins CPUs and Processors Data Center news HPE launches storage system for HPC and AI clusters The HPE Cray Storage Systems C500 is tuned to avoid I/O bottlenecks and offers a lower entry price than Cray systems designed for top supercomputers. By Andy Patrizio May 07, 2024 3 mins Supercomputers Enterprise Storage Data Center news Lenovo ships all-AMD AI systems New systems are designed to support generative AI and on-prem Azure. By Andy Patrizio Apr 30, 2024 3 mins CPUs and Processors Data Center PODCASTS VIDEOS RESOURCES EVENTS NEWSLETTERS Newsletter Promo Module Test Description for newsletter promo module. Please enter a valid email address Subscribe