User:Victorbusque/sandbox

Volta is the codename for a GPU microarchitecture developed by Nvidia as the successor to the Pascal microarchitecture and announced as a future roadmap ambition in March 2013. , the first products featuring this architecture are anticipated in Q4 2017. The architecture is named after Alessandro Volta, the physicist, chemist, and inventor of the electrical battery.

Details
In March 2013, Nvidia announced that the successor of the Pascal microarchitecture would be named Volta and include on-chip, stacked DRAM with 1 TB/s of bandwidth.

It was reported in March 2017 that TSMC would be fabricating Volta using a 12 nm process and that the new microarchitecture is expected in 2018.

Architectural improvements of the Volta architecture include the following:
 * CUDA Compute Capability 7.0 (GV100 only)
 * High Bandwidth Memory 2
 * NVLink 2.0: a high-bandwidth bus between the CPU and GPU, and between multiple GPUs. Allows much higher transfer speeds than those achievable by using PCI Express; estimated to provide 25 Gbit/s per lane.
 * Tensor cores: A tensor core is a unit that multiplies two 4×4 FP16 matrices, and then adds a third FP16 or FP32 matrix to the result by using fused multiply–add operations, and obtains an FP32 result that could be optionally demoted to an FP16 result. Tensor cores are intended to speed up the training of neural networks.

Products
Volta has been announced as the GPU microarchitecture within the Xavier generation of Tegra SoC focusing on self-driving cars.

Nvidia officially announced the Volta microarchitecture as part of the Tesla V100 product announcement on May 10, 2017. The Volta GV100 GPU is built on a 12 nm process size using HBM2 memory with 900 GB/s of bandwidth.

Application
Volta is also reported to be included in the Summit and Sierra supercomputers, used for GPGPU compute. The Volta GPUs will connect to the POWER9 CPUs via NVLink 2.0, which is expected to support cache coherency and therefore improve GPGPU performance.