NVlink will be the backbone of 150 Petaflop Coral supercomputers coming to ORNL and LLNL. As the world’s first high-speed GPU interconnect, NVLink will let data move between GPUs and CPUs 5-12 times faster than possible today.
Today the Oak Ridge Leadership Computing Facility (OLCF) announced a significant step forward in the delivery of Summit, the next-generation supercomputer IBM will deliver in 2017 to greatly advance the nation’s energy and science agenda. Following a formal call for proposals and a review process, ORNL revealed 13 science application readiness projects selected as part of the Center for Accelerated Application Readiness (CAAR) program.
“The use of GPUs to accelerate applications is mainstream nowadays, but their adoption in cur- rent clusters presents several drawbacks. In this talk we present the last developments of the rCUDA remote GPU virtualization framework, which is the only one supporting the most recent CUDA version, in addition to leverage the InfiniBand fabric for the sake of performance.”
In this episode, the Radio Free HPC team wraps up the GPU Technology Conference. The theme of the show this year was Deep Learning, a topic that is heating up the market for GPUs with challenges like image recognition and self-driving cars. As a sister conference, the OpenPOWER Summit this week in San Jose showcased the first OpenPower hardware, including a prototype HPC server from IBM that will pave the way to the two IBM/Nvidia/Mellanox Coral supercomputers expected in 2017.
This week insideHPC will be streaming live keynotes from the GPU Technology Conference in San Jose. Today’s keynote will feature Google Senior Fellow Jeff Dean. “Google has built large-scale computer systems for training neural networks, and then applied these systems to a wide variety of problems that have traditionally been very difficult for computers.”