Posted on 16 February, 2021
Industry: Theoretical science research
Challenges: Delivering cost-effective HPC modeling and analysis capability to nearly 50 different research groups
Solution: Deploy servers with AMD EPYC CPU and Radeon Instinct MI50 GPU
Theoretical science pushes boundaries, furthering our understanding of fundamental natural phenomena. The computing platforms used to model and analyse such complexity in detail must be capable of handling everything imaginable, and more. For the Center for Scientific Computing (CSC) at Frankfurt Institute for Advanced Studies (FIAS), located at Goethe University, that demand is amplified by the challenges of supporting HPC needs across nearly 50 independent research groups.
To overcome that limitation, Professor Volker Lindenstruth, chairman of the Board of Director, FIAS, and his team selected the combination of 2nd Gen AMD EPYC processors and AMD Radeon Instinct MI50 GPU computing accelerators to obtain the combination of performance, advanced I/O capabilities, and operating efficiency needed to accomodate their clients' many varied applications.
"More than 98% of processing for the 600 gigabytes per second data stream runs entirely on servers with eight AMD MI50 GPUs that deliver 90% of the total compute performance we need using just the GPUs." says Professor Lindenstruth.
To read more about achieving cost-performance objectives to grow capacity to support more research, click here to read the full case study.