IBM and Nvidia Develop Next-Generation Supercomputers

    In 2011, IBM showed a technological breakthrough when the Watson supercomputer beat the Jeopardy game show champions. And now the company wants to reboot its high-performance Power servers, using the GPU for the first time.



    Starting next year, IBM will begin to use Nvidia's Tesla GPUs in servers with Power chips, thanks to which there are such supercomputers as Watson, Seqoia and others.

    An agreement between the corporations will unite the new Tesla K40 GPU and the upcoming IBM Power8 processor.

    In terms of features, the new K40It has 2880 CUDA cores and is based on GDD5 graphics memory with a clock frequency of 3 GHz. The basic version of the Tesla K40 has 12 GB of memory. And the new Power8 processor is manufactured using the 22-nm SOI process technology, contains 12 cores operating at frequencies up to 4 GHz, 12 x 0.5 MB of the second level cache and 96 MB of the shared third-level cache (eDRAM).

    The alliance between IBM and Nvidia will not only increase our chances of hearing about new amazing records in the world of supercomputers, but also help business partners process and analyze data faster (this is especially important for such critical tasks as processing financial transactions).

    Also popular now: