Nvidia has released the latest in its range of Tesla GPUs, which are general purpose graphical processors designed to handle supercomputing calculations.
The Tesla M2090 GPU, announced by the company on Tuesday, contains 512 Cuda cores that are designed to work in parallel on typical high-performance computing (HPC) tasks, such as the simulation of microscopic biological interactions.
HP plans to sell a ProLiant SL390 G7 four rack-unit server that will incorporate up to eight M2090s and two CPUs. The server is meant to handle tasks such as quantum chemistry, seismic processing and data analytics.
The M2090s have 64 cores more than their M2070 predecessors, and will provide 177 gigabytes-per-second of bandwidth compared with 150GBps for the M2070s.
More than 7,000 Tesla M2050s — two generations prior to the M2090 — underpinned the world's fastest supercomputer, the Chinese Tianhe-1A .
The M-series of Tesla GPUs are not a standalone product and are sold within servers that integrate them with CPUs, such as HP's SL390 G7. They are supported by 64-bit Windows Server 2008 and 2008 R2; 32- and 64-bit versions of Red Hat Enterprise Linux Server 5.4 and 4.8; Ubuntu 9.10; and Suse Linux Enterprise Server 11.
Pricing has not been announced and will vary on a vendor-by-vendor basis.