<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=145304570664993&amp;ev=PageView&amp;noscript=1">
Lenovo ThinkSystem SR630 V2 Rack Server

May 13, 2022

Lenovo servers now available for Graphcore IPU systems

Written By:

Jason Lu

Join the IPU conversation

Join our Graphcore community for free. Get help and share knowledge, find tutorials and tools that will help you grow.

Join on Slack

Graphcore is delighted to announce that its Elite Partners are now able to offer Lenovo servers as part of the Bow Pod product range.

Lenovo is one of the world’s leading providers of datacentre technology, used and trusted by thousands of customers across a wide range of industries and applications.

The addition of Lenovo servers means even greater choice for AI users looking to make the switch to IPU Pod systems, with the performance and efficiency benefits that Graphcore has become known for.


Choosing the right companion servers for the IPU and configuring them optimally is fundamental to enable efficient deployment of machine intelligence at scale.

To ensure Graphcore customers have the best possible experience, we work with leading systems providers—including Lenovo—to test and validate server systems for a wide range of configurations.

Click here for a full list of Graphcore-approved servers

The first Lenovo server to be validated and supported for use with Bow Pod systems is the ThinkSystem SR630 V2, a dual socket 1U server.

The ThinkSystem SR630 V2 is an ideal companion to Graphcore’s new Bow IPU-based systems, enabling a platform that is designed both for today’s most demanding AI applications and for tomorrow’s innovations.Lenovo ThinkSystem SR630 V2 rack server

Architectural advantage

As with all Graphcore customers, users of Lenovo systems also benefit from the IPU Pod’s disaggregated server architecture, allowing systems to be optimised around the workloads they are most likely to undertake.

For example, customers focusing on NLP applications can choose to specify fewer servers than those working on computer vision-related tasks due to the lower host server requirement.

In the most recent MLPerf 1.1 training round, Graphcore demonstrated the practical advantage of its disaggregated server architecture with an industry-leading single-server time-to-train result for BERT.