IBM, NVIDIA and Wistron have just introduced their second generation server aimed at powering high performance computing (HPC) applications at this year's OpenPOWER summit.
The new server will use Power8 microprocessors from IBM and Nvidia's new Tesla P100 GPU, along with its NVLink interconnection technology. Software developers will need to port their apps to this new HPC platform and IBM and NVIDIA are prepared to assist companies in doing so.
Although IBM, NVIDIA and Wistron have not revealed the full specifications of their HPC platform, it is likely to include two IBM Power8 CPUs and either 4 or 8 Tesla P100's from NVIDIA.
The POWER8 CPU from IBM contains 12 cores that are able to independently handle eight hardware threads simultaneously. It can operate at high clock speeds between 3-3.5 GHz and has a total of 6MB of L2 cache as well as 96 MB of L3 cache. Each POWER8 processor will be able to support up to 1 TB of DDR3 or DDR4 memory with a total of up to 230 GB per second sustained bandwidth.
The Tesla P100 will also bring a lot of power to the system and each unit has 3584 stream processors, 4 MB of L2 cache and 16 GB of HBM2 memory. Each card can reach 10.6 Teraflops of single precision performance as well.
IBM, NVIDIA and Wistron plan on bringing the second generation of their HPC platform to market in the 4th quarter of 2016.