The use of machine and deep learning in society continues to grow. Currently the hardware used to accelerate the associated heavy computational workloads is dominated by graphics processors from Nvidia (“GPUs”). These chips have been repurposed from their intended original focus to run AI models.
Graphcore is a leading machine intelligence semiconductor business, which has developed the Intelligence Processing Unit (“IPU”) designed from first principals to address the AI market’s likely future requirements. The IPU is completely different from today’s Central Processing Unit (“CPU”) and GPU processors. It is a highly flexible, easy to use, parallel processor that has been designed from the ground up to deliver state of the art performance on today’s machine intelligence models for both training and inference. In July 2020, Graphcore revealed its second generation IPU with an 8x step up in performance.
Graphcore has continued to invest heavily in its technology, most visibly in its software development kit, where it has released several updates, giving customers further functionality and ease-of-use, and resulting in improved performance.