Donated by Nvidia, a member of the HPC Advisory Council, the program is designed to help researchers and developers benchmark their software and optimize their code to run on Tesla GPUs. The goal appears to be the finalization of software to make sure that the software will perform exactly as advertised. The HPC Council told us that the cluster will have 16 GPUs with four being available at this time. The GPU time that can be allocated to a user depends on the specific needs.
"Researchers need an easy way to benchmark their models on the growing number of GPU-accelerated applications before making a buying decision," said Sumit Gupta, director of Tesla business at Nvidia. "The new Center provides a valuable resource to help developers optimize their codes for GPUs, and ensure that applications will perform precisely as advertised."
Gilad Shainer, chairman of the HPC Advisory Council noted that HPC systems have been donated by other member companies in the past, including AMD and Intel.