rchitecture
NVIDIA Blackwell Architecture
NVIDIA Tensor Cores
203 (5th Gen)
Process Size
4 nm Custom Process
CUDA Parallel Processing Cores
10,496
NVIDIA RT Cores
82 (4th Gen)
Single-Precision Performance (FP32)
51 TFLOPS
TF32 Tensor core
203 TFLOPS
FP16 | BF16 Tensor core
406 TFLOPS
FP8 Tensor core
811 TFLOPS
FP4 tensor core
1.6 PFLOPS
INT 8 Tensor Core
811 TOPS
Peak RT Core Performance
154 TFLOPS
Memory Bandwidth
800 GB/s
Multi-Instance GPU
Up to 2 MIG @ 16GB
Confidential Compute
Supported
Interconnect
PCI Express 5.0 x 16
Form Factor
4.4” H x 10.5” L, FHFL Single Slot
Product Weight
1055 Grams
Power Connector
1x PCIe CEM5 16-pin