Google Cloud has
announced a technology partnership with AMD, which will move some of its electronic design automation (EDA) workloads for chip design to Google Cloud to expand the capabilities of its own data centers. Instances AMD plans to use include C2D VM instances powered by its own EPYC processors, but it appears that other non-EPYC processor instances will be used as well.
AMD also talked about leveraging Google Cloud’s global networking, storage, artificial intelligence, and machine learning capabilities to further improve its hybrid and multi-cloud strategy for related EDA workloads. Simply put, AMD would rent its own chips from Google Cloud to design new chips.
“In today’s semiconductor environment, the speed, scale, and security of the cloud unlock much needed flexibility,” said Sachin Gupta, GM and VP, Infrastructure, at Google Cloud. “We are pleased to provide the infrastructure required to meet AMD’s compute performance needs and equip the company with our AI solutions to continue designing innovative chips.“
“Leveraging the Google Cloud C2D instances powered by 3rd Gen EPYC processors for our complex EDA workloads has helped our engineering and IT teams tremendously. C2D has allowed us to be more flexible and provided a new avenue of high-performance resources that allows us to mix and match the right compute solution for our complex EDA workflows,” said Mydung Pham, corporate vice president, Silicon Design Engineering, at AMD. “We’re happy to work with Google Cloud to take advantage of their wealth of cloud features and the capabilities of 3rd Gen EPYC.”
Over the past two years, Google Cloud and AMD have worked closely together. Last year, Google Cloud
announced that its Compute Engine virtual machine product line plans to introduce a new Tau virtual machine, or T2D, whose hardware is based on AMD’s third-generation EPYC processors codenamed Milan. Google Cloud said that compared with the previous virtual machine, the performance of the T2D virtual machine using AMD technology has been greatly improved.