Volta is the codename, but not the trademark,^[2] for aGPUmicroarchitecture developed byNvidia, succeedingPascal. It was first announced on a roadmap in March 2013,^[3] although the first product was not announced until May 2017.^[4] The architecture is named after 18th–19th century Italian chemist and physicistAlessandro Volta. It was Nvidia's first chip to featureTensor Cores, specially designed cores that have superior deep learning performance over regular CUDA cores.^[5] The architecture is produced withTSMC's12 nmFinFET process. TheAmpere microarchitecture is the successor to Volta.

The first graphics card to use it was the datacenterTesla V100, e.g. as part of theNvidia DGX-1 system.^[4] It has also been used in the Quadro GV100 and Titan V. There were no mainstream GeForce graphics cards based on Volta.

After two USPTO proceedings,^[6][7] on July 3, 2023 Nvidia lost the Volta trademark application in the field of artificial intelligence. The Volta trademark^[8] owner remains Volta Robots, a company specialized in AI and vision algorithms for robots and unmanned vehicles.

Architectural improvements of the Volta architecture include the following:

• CUDA Compute Capability 7.0
  • concurrent execution of integer and floating point operations
• TSMC's12 nmFinFET process,^[9] allowing 21.1 billiontransistors.^[10]
• High Bandwidth Memory 2 (HBM2),^[9][11]
• 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 usingPCI Express; estimated to provide 25 Gbit/s per lane.^[12] (Disabled for Titan V)
• Tensor cores: A tensor core is a unit that multiplies two 4×4FP16 matrices, and then adds a third FP16 or FP32 matrix to the result by usingfused multiply–add operations, and obtains an FP32 result that could be optionally demoted to an FP16 result.^[13] Tensor cores are intended to speed up the training of neural networks.^[13] Volta's Tensor cores are first generation while Ampere has third generation Tensor cores.^[14][15]
• PureVideo Feature Set I hardware video decoding

Comparison of Compute Capability: GP100 vs GV100 vs GA100^[16]

Comparison of Precision Support Matrix^[17][18]

Legend:

• FPnn: floating point with nn bits
• INTn: integer with n bits
• INT1: binary
• TF32:TensorFloat32
• BF16:bfloat16

Comparison of Decode Performance

Volta has been announced as the GPU microarchitecture within theXavier generation ofTegraSoC focusing onself-driving cars.^[19][20]

At Nvidia's annual GPU Technology Conference keynote on May 10, 2017, Nvidia officially announced the Volta microarchitecture along with theTesla V100.^[4] The Volta GV100 GPU is built on a 12 nm process size using HBM2 memory with 900 GB/s of bandwidth.^[21]

Nvidia officially announced the Nvidia TITAN V on December 7, 2017.^[22][23]

Nvidia officially announced the Quadro GV100 on March 27, 2018.^[24]

Volta is also reported to be included in theSummit andSierra supercomputers, used for GPGPU compute.^[29][30] The Volta GPUs will connect to thePOWER9 CPUs viaNVLink 2.0, which is expected to supportcache coherency and therefore improve GPGPU performance.^[31][12][32]

Comparison of accelerators used in DGX:^[33][34][35]

• List of eponyms of Nvidia GPU microarchitectures
• List of Nvidia graphics processing units

• Graphics microarchitectures
• Nvidia microarchitectures

• Articles with short description
• Short description is different from Wikidata