 |
|
 |
       |
 |
|
High Performance Computing (HPC)
High performance computing encompasses advanced computation over parallel processing, enabling faster execution of highly compute intensive tasks such as climate research, molecular modeling, physical simulations, cryptanalysis, geophysical modeling, automotive and aerospace design, financial modeling, data mining and more.
Commodity server clusters and grids, linked with the high-speed, industry standard InfiniBand interconnect, create leading price-performance solutions that deliver intensive compute power.
HPC Server clusters requires a high performance interconnect technology providing high bandwidth with low latency and CPU offload (direct data transfer between nodes). The InfiniBand architecture is the only industry standard technology that advances I/O connectivity for high performance computing clusters.
Node-to-node latency of less than 3us and almost 3GByte/sec has been demonstrated on Mellanox InfiniBand clusters with the popular Message Passing Interface (MPI) protocol.
With InfiniBand's proven scalability and efficiency, small and large clusters easily scale up to thousands of nodes. With 20Gb/s node-to-node and 60Gb/s switch-to-switch solutions available, and a roadmap to 120Gb/s, InfiniBand has proven its performance in personal supercomputing, workgroup, departmental, divisional, and enterprise supercomputers solutions
 |
|