Sign up for our newsletter and get the latest HPC news and analysis.
Send me information from insideHPC:

Consumer GPUs Power 2 Petaflop Deep Learning Cluster at ASTRON

Today ClusterVision announced the successful deployment of a new high performance computing GPU cluster system for the Netherlands Institute for Radio Astronomy (ASTRON). “By utilizing the deep learning capabilities of the GPU cluster, the telescopes will be able to detect pulsar flashes with much greater accuracy through self-learning. In the past, ASTRON scientists had to manually detect and input pulsar patterns. With deep learning, ARTS does it for them.”

Impact of Meltdown and Spectre Vulnerabilities on HPC

“Performance is important in high performance computing. A performance impact on HPC use can be more than a simple annoyance. Because of the high investment cost and energy usage of HPC systems, a performance impact can have a signi cant effect on price per compute job. As we have discussed in the previous sections, the “page table isolation” workarounds have an overhead on entering and leaving the kernel. Luck happens that in HPC, computers spend most of their time in user space, doing computations. Low-latency networks such as In niband allow HPC systems to perform network communication without involving the kernel. Therefore during HPC computations, the Linux kernel might not be actively involved and the CPUs are running at full power.”

ClusterVision White Paper Looks at HPC Performance Impact of Spectre and Meltdown

While various kernel patches are already out for the Spectre and Meltdown, the performance impact of these patches on HPC performance has been a big question. Now ClusterVision has published a timely white paper on this important topic. “These vulnerabilities have only been discovered recently, so information is still developing. Therefore, this document should not be interpreted as a complete overview of the situation but as an informative view of the potential impact on HPC.”