Tesla K20X

General
ManufacturerNVIDIA
SeriesTesla Kepler(Kxx)
GPU NameGK110
ArchitectureKepler
GenerationTesla Kepler(Kxx)
Release DateNov 2012
Bus InterfacePCIe 3.0 x16
Process & Die
FoundryTSMC
Process Size28 nm
Transistors7,080 M
Transistor Density12,600 K/mm²
Die Size561 mm²
Chip PackageBGA-2152
Clock Speeds
Base Clock732 MHz
Boost Clock732 MHz
Memory
VRAM6 GB
Memory TypeGDDR5
Memory Clock1,300 MHz
Memory Bus384-bit
Bandwidth249.6 GB/s
Processing Units
Shading Units2,688
TMUs224
ROPs48
SMs0
Tensor Cores0
RT Cores0
Cache
L1 Cache16 KB
L2 Cache1.5 MB
Power & Thermals
TDP235 W
Suggested PSU550 W
Power Connectors1x 6-pin + 1x 8-pin
Dimensions
Board Length267 mm
Outputs
API Support
DirectX11.0
OpenGL4.6
Vulkan1.2
OpenCL3.0
CUDA3.5
Shader Model5.1
Theoretical Performance
Pixel Rate40.99 GP/s
Texture Rate164 GT/s
FP32 (Single)3,935 GFLOPS
FP64 (Double)1,311.7 GFLOPS
External
TechPowerUpView →