NVIDIA Unveils PCIe Tesla V100

NVIDIA has formally announced the PCIe version of its Telsa V100 GPU accelerator. The company made the announcement at this year’s International Supercomputing Conference (ISC), which began yesterday. NVIDIA has already unveiled the Tesla V100 with an SXM2 form factor, but this new PCIe version makes the card compatible with more traditional motherboards. Of course, the downside of the PCIe connection, compared to the SXM2 form factor, is the lack of NVLINK support.

Tesla V100 Specifications

• Volta Architecture
• TSMC 12nm FFN Process
• 5120 CUDA Cores
• 640 Tensor Cores
• ~1370MHz Boost Clock
• 1.75Gbps HBM2 Memory Clock
• 4096-bit Memory Bus Width
• 900GB/sec Memory Bandwidth
• 16GB VRAM
• 6MB L2 Cache
• 28 TFLOPS Half Precision
• 14 TFLOPS Single Precision
• 7 TFLOPS (half rate) Double Precision
• 112 TFLOPS Tensor Performance (Deep Learning)
• GV100 (815mm²) GPU
• 21 billion Transistor Count
• 250W TDP
• PCIe Form Factor
• Passive Cooling

NVIDIA Says

“NVIDIA also announced it is making new Tesla V100 GPU accelerators available in a PCIe form factor for standard servers. With PCIe systems, as well as previously announced systems using NVIDIA NVLink™ interconnect technology, coming to market, Volta promises to revolutionize HPC and bring groundbreaking AI technology to supercomputers, enterprises and clouds.

NVIDIA Tesla V100 GPU accelerators for PCIe-based systems are expected to be available later this year from NVIDIA reseller partner and manufacturers, including Hewlett Packard Enterprise (HPE).”

Pricing and Availability

The NVIDIA Tesla V100 PCIe is out later this year. Price is yet to be announced, though expect it to cost around £10,000.