NVIDIA A100 40GB SXM 
Specification A100 40GB SXM
GPU Architecture NVIDIA Ampere
GPU NVIDIA A100 Tensor Core GPU
GPU Memory 40GB HBM2
Memory Interface 5120-bit
Memory Bandwidth 1.6 TB/s
CUDA Cores 6,912
Tensor Cores 432, 3rd Generation
RT Cores N/A
FP64 Performance 9.7 TFLOPS
FP32 Performance 19.5 TFLOPS
TF32 Tensor Performance 156 TFLOPS
BF16 Tensor Performance 312 TFLOPS
FP16 Tensor Performance 312 TFLOPS
INT8 Tensor Performance 624 TOPS
INT4 Tensor Performance 1,248 TOPS
GPU Interconnect NVLink
NVLink Bandwidth 600 GB/s
Form Factor SXM4
TDP 400W
Cooling Passive / Server Cooling
Display Outputs None
ECC Yes
Memory Type HBM2
MIG Up to 7 GPU Instances
Target Market Data Center / AI / HPC
Main Applications AI Training, AI Inference, HPC, Scientific Computing