Volta architecture

E790550

Volta architecture is NVIDIA’s GPU microarchitecture that debuted specialized Tensor Cores to accelerate deep learning and high-performance computing workloads.

Try in SPARQL Jump to: Statements Referenced by

Statements (48)

Predicate Object
instanceOf GPU microarchitecture
architectureGeneration NVIDIA 7th-generation GPU architecture
codename Volta NERFINISHED
designedFor deep learning workloads
high-performance computing workloads
developer NVIDIA NERFINISHED
feature enhanced SM (Streaming Multiprocessor) design
hardware-accelerated matrix multiply-accumulate
improved instruction scheduling
larger register file per SM
unified L1 data cache and shared memory
firstLaunchYear 2017
firstProduct NVIDIA Tesla V100 NERFINISHED
foundry TSMC NERFINISHED
introducedFeature Tensor Cores NERFINISHED
manufacturer NVIDIA NERFINISHED
marketSegment enterprise
professional
memoryType HBM2 NERFINISHED
namedAfter Alessandro Volta NERFINISHED
notableGPU Quadro GV100 NERFINISHED
Tesla V100 NERFINISHED
Titan V NERFINISHED
optimizationFor large-scale parallel workloads
mixed-precision training
predecessor Pascal architecture NERFINISHED
processNode 12 nm FinFET
successor Ampere architecture NERFINISHED
Turing architecture NERFINISHED
supports CUDA NERFINISHED
CUDA Compute Capability 7.0
ECC memory
FP16 compute
FP32 compute
FP64 compute
HBM2 memory
INT8 operations
NVIDIA NVLink high-bandwidth interconnect
NVLink 2.0 NERFINISHED
Tensor Cores mixed-precision operations
Unified Memory
targetMarket AI training
data center
scientific computing
useCase GPU-accelerated supercomputing
HPC simulations
deep learning inference
neural network training

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Tensor Cores introducedIn Volta architecture