CUDA Parallel-Processing Cores 5120
NVIDIA Tensor Cores 640
FP64 Performance 7.4 TFLO
FP32 Performance 14.8 TFLOPS
FP16 Performance 29.6 TFLOPS
Tensor Performance 118.5 TFLOPS
Max Power Consumption 250 W
Graphics Bus PCI Express 3.0 x 16
Display Connectors DP 1.4 (4)
Form Factor 4.4 H x 10.5 L Dual Slot
Four NVIDIA Maxwell GPUs
Up to 32 users per board
2560 NVIDIA CUDA cores (640 per GPU)
32 GB of GDDR5 memory (8 per GPU)
28 H.264 1080p30 streams
ECC protection for increased reliability
Server optimization to deliver the best throughput in the data center
140X HIGHER THROUGHPUT TO KEEP UP WITH EXPLODING DATA
SIMPLIFIED OPERATIONS WITH A SINGLE TRAINING AND INFERENCE PLATFORM
REAL-TIME INFERENCE
FASTER DEPLOYMENT WITH NVIDIA DEEP LEARNING SDK