Graphcore, the U.K.-based startup that launched the Intelligence Processing Unit (IPU) for AI acceleration in 2018, has introduced the IPU-Machine. This second-generation platform has greater processing power, more memory and built-in scalability for handling extremely large parallel processing workloads. The well-funded startup has a blue-ribbon pedigree of engineers, advisers and investors, and enjoys a valuation approaching $2 billion. Its first-generation hardware is now available on the Microsoft Azure cloud as well as in Dell-EMC servers. Both companies are investors. Graphcore is now betting its future on this second-generation platform, a plug-and-play building block for massive scalability that is currently unique in the industry.
You can download the paper by clicking on the logo below:
Table Of Contents:
- Introduction
- The Colossus MK2 IPU (GC200)
- The IPU-Machine (M2000)
- The Second-Get IPU-Fabric
- Updates To The Graphcore Software Stack
- Performance
- Conclusions
- Figure 1: The Graphcore IPU-Machine
- Figure 2: IPU-Machine M2000 Architectural Diagram
- Figure 3: IPU-Machine Stacking Options
- Figure 4: IPU-POD For Super Computing Scale
- Figure 5: IPU-POD64 Reference Architecture
- Figure 6: POPLAR SDK
- Figure 7: IPU-POD Support For Multi-Tenancy
- Figure 8: Colossus MK2 Performance
Companies Cited
- Dell EMC
- Graphcore
- Microsoft Azure
- NVIDIA
Patrick Moorhead is the founder, CEO, and chief analyst of Moor Insights & Strategy. His big-picture view of technology is grounded in more than 20 years as an executive leading strategy, product management, product marketing, and corporate marketing functions at NCR, AT&T, Compaq, and AMD. He has shared his expertise in areas from silicon to infrastructure to enterprise SaaS and everything in-between in thousands of national broadcast appearances (CNBC, Yahoo Finance), articles (Forbes, CIO), research-based analyses, and podcast episodes. Today, he has 100+ CXO-level advisory clients and is often ranked the #1 technology industry analyst by ARInsights.
