NVIDIA Tesla P100 GPU accelerator for PCIe servers to tackle AI, HPC needs

Source: NVIDIA. The NVIDIA Tesla P100 GPU accelerator.

NVIDIA today introduced the NVIDIA Tesla P100 GPU accelerator for PCIe servers, which delivers leaps in performance and value compared with CPU-based systems.

The Tesla P100 GPU accelerator for PCIe meets computational demands through the performance and efficiency of the NVIDIA Pascal GPU architecture. It 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*.

“Accelerated computing is the only path forward to keep up with researchers’ insatiable demand for HPC and AI supercomputing,” said Ian Buck, VP of accelerated computing at NVIDIA. “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.”

The Tesla P100 for PCIe is available in a standard PCIe form factor and is compatible with today’s GPU-accelerated servers. It is optimised to power computationally intensive artificial intelligence (AI) and high performance computing (HPC) data centre applications. A single Tesla P100-powered server delivers higher performance than 50 CPU-only server nodes when running the AMBER molecular dynamics code, and is faster than 32 CPU-only nodes when running the VASP material science application**.

Key features include:

 Unmatched application performance for mixed-HPC workloads, delivering 4.7 teraflops and 9.3 teraflops of double-precision and single-precision peak performance, respectively, a single Pascal-based Tesla P100 node provides the equivalent performance of more than 32 commodity CPU-only servers.

 Chip on wafer on substrate (CoWoS) with HBM2*** for unprecedented efficiency – The Tesla P100 unifies processor and data into a single package to deliver unprecedented compute efficiency. An innovative approach to memory design, with HBM2, provides a 3x boost in memory bandwidth performance, or 720GBpsec, compared to the NVIDIA Maxwell architecture.

 Page Migration Engine for simplified parallel programming frees developers to focus on tuning for higher performance and less on managing data movement, and allows applications to scale beyond the GPU physical memory size with support for virtual memory paging.

 Unmatched application support, there are 410 GPU-accelerated applications, including nine of the top 10 HPC applications supported on the Tesla platform.

Interested?

The NVIDIA Tesla P100 GPU accelerator for PCIe-based systems is expected to be available beginning Q416 from NVIDIA reseller partners and server manufacturers, including Cray, Dell, HP and IBM. Two configurations will be offered, 16GB of CoWoS HBM2 stacked memory, delivering 720GBpsec of memory bandwidth and 12GB of CoWoS HBM2 stacked memory, delivering 540GBpsec of memory bandwidth

*CPU server: Dual socket Intel E5-2680v3 12 cores, 128 GB DDR4 per node, FDR IB/GPU server: 8x Tesla P100 for PCIe with Dual Socket Intel E5-2680v3 

**VASP 5.4.1_05Feb16, Si-Huge Dataset. 16, 32 nodes are estimated based on same scaling from four to eight nodes.

***HBM2, or High Bandwidth Memory 2, is a standard that offers more bandwidth and capacity than is available with traditional memory standards.