NVIDIA Unveils the Tesla P100 HPC Board based on Pascal GPU Architecture

NVIDIA unveiled the Tesla P100, the first product based on the company’s “Pascal” GPU architecture. Tesla P100 powered by new multi-chip module, similar type layout to the AMD “Fiji.” Tesla P100 has 15 billion-transistor with a 4096-bit wide HBM2 memory interface wires it to four 3D HBM2 stacks; sitting on top a fiberglass substrate that’s rooted into the PCB. The GPU die is built on the 16 nm FinFET process, and is 600 mm² in area.

Tesla P100 Tesla P100 1
Tesla P100 2 Tesla P100 3

The P100 offers a double-precision (FP64) compute performance of 5.3 TFLOP/s, FP32 performance of 10.6 TFLOP/s, and FP16 performance of a whopping 21.2 TFLOP/s. The chip has registers as big as 14.2 MB, and an L2 cache of 4 MB. In addition to PCI-Express, each P100 chip will be equipped with NVLink, and in-house developed high-bandwidth interconnect by NVIDIA, with bandwidths as high as 80 GB/s per direction, 160 GB/s both directions. This allows extremely high-bandwidth paths between GPUs, so they could share memory and work more like single-GPUs. The P100 is already in volume production, with its target customers already having bought it all the way up to its OEM channel availability some time in Q1-2017.