Penguin Computing Offering Intel® Omni-Path Architecture
Penguin Computing is dedicated to offering the best performance per cost solutions to our customers. This is why we offer Intel® Omni-Path architecture, an element of Intel® Scalable System Framework.
Penguin Computing recently won a contract, called CTS-1, with the National Labs to deliver petaFLOPS-scale computing with our Tundra Extreme Scale HPC solution. In the CTS-1 contract, they selected Intel® Omni-Path architecture for the fabric both for its performance and price.
Network fabric performance and reliability are critical for high-performance computing (HPC) clusters because so many machines have to work together in parallel. Intel® Omni-Path, with both low latency and high performance at 100 gigabits per second, is a valuable addition that accelerates and unleashes the performance of the CPU.
One of the innovations of Omni-Path is increasing the size from 36 to 48 ports. This allows us to build larger fabrics with fewer layers … and fewer layers means lower latency and fewer switches and, thus, lower cost.
Intel® dedication to open sourcing the software that manages the Omni-Path architecture has been a huge benefit in making this technology accessible. Open source technologies help democratize high-performance computing, making it available to more scientists and engineers for innovative and critical research.
Penguin Computing CTO, Phil Pokorny, discusses Intel Omni-Path fabric in this video:
Learn more about Intel’s Omni-Path fabric.