24 GB
HBM21440 MHz
Boost Clock165 W
TDP7 nm
ProcessShare on Social Media
Share PG506-217 specifications with othersTechnical Specifications
Core Specifications
Architecture:
Ampere
Process Size:
7 nm
Base Clock:
930 MHz
Boost Clock:
1440 MHz
Shader Units:
3584
SMs:
56
Tensor Cores:
224
TMUs:
224
ROPs:
96
Memory & Performance
Memory Size:
24 GB
Memory Type:
HBM2
Memory Bandwidth:
933.1 GB/s
Bus Interface:
PCIe 4.0 x16
TDP:
165 W
Suggested PSU:
450 W
Cache & Die Information
L1 Cache:
192 KB
L2 Cache:
24.0 MB
Die Size:
826 mm²
Transistors:
54.2 billion
Transistor Density:
65.6 million/mm²
API Support
Cuda Compute
8.0 sm_80Physical Specifications
Board Length:
267 mm
Board Width:
112 mm
Power Connectors:
8-pin EPS
Quick Facts
Performance
Pixel Fill Rate
138.2 GPixel/s
Texture Fill Rate
322.6 GTexel/s
FP32 Performance
10.3 TFLOPS
CUDA Toolkit Compatibility
Cuda Compute
8.0
sm_80
Architecture
Ampere
Supported CUDA Toolkits
11.0
11.1
11.2
11.3
11.4
11.5
11.6
11.7
11.8
12.0
12.1
12.2
12.3
12.4
12.5
12.6
12.7
12.8
High-end Ampere architecture with Tensor Cores and Multi-Instance GPU