PORTLAND, Ore. — At the Supercomputer Conference (SC13, Nov. 17-22, Denver) Nvidia and IBM announced that they will collaborate on accelerating IBM's enterprise software applications with Nvidia's graphics processing unit (GPUs) for IBM's Power System servers. The partners also plan to integrate Tesla's GPUs with IBM's Power8 processors.
Although the announcement was made at SC13, Nvidia and IBM emphasized that their collaboration is aimed not just at high-performance computing (HPC) users but also at moving what has until now been almost exclusively a supercomputer coprocessor -- Nvidia's Tesla -- into enterprise-grade servers to help deal with the increasing volume of streaming Big Data in business contexts.
Nvidia coprocessor cards for supercomputers are moving into
IBM Power servers.
IBM already supports Intel's Xeon Phi coprocessor on its System x servers, which use the same x86 programming model as the main Xeon processors, but will now also support the Cuda programming model for Nvidia's Tesla coprocessors on the Power8 processors in its Power System servers.
IBM plans to support Nvidia's Tesla coprocessors and Cuda programming model both for the scientific, engineering, and visualization applications developed for supercomputer applications as well as with its Rational suite of enterprise software development tools.
Nvidia and IBM are also members of the OpenPower Consortium, which is building a Power System open ecosystem with Google, Mellanox, and Tyan.
— R. Colin Johnson, Advanced Technology Editor, EE Times