GPU vendor Nvidia revealed its Tesla GPU earlier this year. The first product using the chip was the DGX-1 "supercomputer in a box".
The company has now announced the Tesla P100 GPU accelerator for PCIe servers, aimed at HPC and AI applications.
The new card is rated at 4.7 double-precision teraflops or 9.3 single-precision teraflops. Combined with up to 720GBps of memory bandwidth, this enables the creation of "super nodes" that provide the throughput of more than 32 commodity CPU-based nodes and deliver up to 70% lower capital and operational costs, the company claimed.
|
|
Nvidia's Page Migration Engine simplifies parallel programming by enabling developers to see a single memory space for the entire node and by supporting virtual memory paging.
Tesla P100 GPU accelerator for PCIe servers uses the standard PCIe form factor and will work with existing GPU-accelerated servers.
It will be available in two configurations: 16GB memory (720GBps of memory bandwidth) or 12GB (540GBps).
The accelerator should ship in the fourth quarter. It will be available from resellers, as well as server manufacturers including Cray, Dell, HP and IBM.
"Accelerated computing is the only path forward to keep up with researchers' insatiable demand for HPC and AI supercomputing," said Nvidia vice-president of accelerated computing Ian Buck.
"Deploying CPU-only systems to meet this demand would require large numbers of commodity compute nodes, leading to substantially increased costs without proportional performance gains. Dramatically scaling performance with fewer, more powerful Tesla P100-powered nodes puts more dollars into computing instead of vast infrastructure overhead."
