Share on:
GPU Architecture | NVIDIA Ampere Architecture | |
CUDA Parallel Processing cores | 3,328 | |
NVIDIA Tensor Cores | 104 | |
NVIDIA RT Cores | 26 | |
GPU Memory | 6 GB GDDR6 with ECC | 12 GB GDDR with ECC | |
Single-Precision Performance1 | 8 TFLOPS | |
RT Core Performance1 | 15.6 TFLOPS | |
Tensor Performance1 | 63.9 TFLOPS2 | |
Memory Interface | 192-bit | |
Memory Bandwidth | 228 GB/s | |
Max Power Consumption | 70 W | |
Graphics Bus | PCI Express 4.0 x 16 | |
Display Connectors | mDP 1.4 (4)3 | |
Form Factor | 2.713”H x 6.6”L Dual Slot | |
Product Weight | 294 g (Low Profile Bracket) 306 g (ATX Bracket) | |
Thermal Solution | Active | |
NVENC | NVDEC | 1x | 1x (+AV1 decode) | |
1Peak rates are based on GPU boost clock. 2Effective TOPS/TFLOPS using the new sparsity feature. |