NVIDIA Pascal architecture

E762898
GPU microarchitecture

NVIDIA Pascal architecture is a GPU microarchitecture from NVIDIA designed to deliver high-performance computing and deep learning acceleration through improved efficiency, memory bandwidth, and parallel processing capabilities.

Try in SPARQL Jump to: Surface forms Statements Referenced by

Observed surface forms (4)

Surface form Occurrences
NVIDIA Tesla V100 2
NVIDIA Tesla P4 1
NVIDIA Tesla P40 1

Statements (58)

Predicate Object
instanceOf GPU microarchitecture ⓘ
announcementDate 2015-03-17 ⓘ
architectureGeneration predecessor to NVIDIA Volta ⓘ
successor to NVIDIA Maxwell ⓘ
codename Pascal NERFINISHED ⓘ
developer NVIDIA NERFINISHED ⓘ
firstProductsLaunchYear 2016 ⓘ
flagshipComputeGPU NVIDIA Tesla P100 NERFINISHED ⓘ
flagshipConsumerGPU NVIDIA GeForce GTX 1080 NERFINISHED ⓘ
NVIDIA GeForce GTX 1080 Ti NERFINISHED ⓘ
flagshipDatacenterGPU NVIDIA Tesla P100 NERFINISHED ⓘ
NVIDIA Tesla P4 NERFINISHED ⓘ
NVIDIA Tesla P40 NERFINISHED ⓘ
manufacturingProcess 14 nm FinFET ⓘ
16 nm FinFET ⓘ
marketSegment datacenter accelerators ⓘ
desktop GPUs ⓘ
mobile GPUs ⓘ
workstation GPUs ⓘ
notableUseCase real-time graphics rendering ⓘ
scientific simulations ⓘ
training deep neural networks ⓘ
predecessor NVIDIA Maxwell architecture NERFINISHED ⓘ
successor NVIDIA Volta architecture NERFINISHED ⓘ
supportedMemoryType GDDR5 ⓘ
GDDR5X ⓘ
HBM2 ⓘ
supportsAPI CUDA 8 NERFINISHED ⓘ
CUDA 9 (on some GPUs) NERFINISHED ⓘ
DirectX 12 feature level 12_1 ⓘ
OpenCL 1.2 NERFINISHED ⓘ
OpenGL 4.5 NERFINISHED ⓘ
Vulkan NERFINISHED ⓘ
supportsFeature Ansel NERFINISHED ⓘ
FP64 double-precision compute (higher ratio on Tesla P100) ⓘ
Fast Sync ⓘ
GPU Boost 3.0 NERFINISHED ⓘ
NVIDIA GPU Direct RDMA (on Tesla products) NERFINISHED ⓘ
NVIDIA NVLink NERFINISHED ⓘ
Simultaneous Multi-Projection NERFINISHED ⓘ
enhanced asynchronous compute ⓘ
enhanced power efficiency ⓘ
error-correcting code memory (ECC) on Tesla products ⓘ
improved instruction scheduling ⓘ
improved memory compression ⓘ
mixed-precision compute (FP16 on some GPUs) ⓘ
multi-GPU via SLI HB bridge (on some GeForce cards) ⓘ
preemption at instruction level (on some GPUs) ⓘ
unified memory with page migration ⓘ
targetDomain consumer graphics ⓘ
deep learning ⓘ
high-performance computing ⓘ
professional visualization ⓘ
usedInProductLine NVIDIA GeForce 10 series NERFINISHED ⓘ
NVIDIA Quadro P series NERFINISHED ⓘ
NVIDIA Tesla P series NERFINISHED ⓘ
NVIDIA Titan X (Pascal) NERFINISHED ⓘ
NVIDIA Titan Xp NERFINISHED ⓘ

Referenced by (6)

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

NVIDIA Tesla data center GPUs → includesModel → NVIDIA Pascal architecture ⓘ
this entity surface form: NVIDIA Tesla P4
NVIDIA Tesla data center GPUs → includesModel → NVIDIA Pascal architecture ⓘ
this entity surface form: NVIDIA Tesla P40
NVIDIA Tesla data center GPUs → includesModel → NVIDIA Pascal architecture ⓘ
this entity surface form: NVIDIA Tesla V100
NVIDIA Tesla data center GPUs → architectureBasedOn → NVIDIA Pascal architecture ⓘ
Tensor Cores → introducedWithGPU → NVIDIA Pascal architecture ⓘ
this entity surface form: NVIDIA Tesla V100
Nvidia Maxwell GPU → followedBy → NVIDIA Pascal architecture ⓘ
subject surface form: Nvidia Maxwell
this entity surface form: Nvidia Pascal