Architecture |
NVIDIA Ampere Architecture |
Process Size |
8nm | Samsung |
Transistors |
28.3 Billion |
Die Size |
628.4 mm2 |
CUDA Cores |
7168 |
Tensor Cores |
224 |
RT Cores |
56 |
Single Precision Performance |
23.7 TFLOPS |
RT Core Performance |
46.2 TFLOPS |
Tensor Performance |
182.2 TFLOPS |
GPU Memory |
20GB GDDR6 with ECC |
Memory Interface |
320-bit |
Memory Bandwidth |
640 GB/sec |
Display Connectors |
4x DisplayPort 1.4a |
NVENC | NVDEC |
1x | 2x (+ AV1 decode) |
System Interface |
PCI Express 4.0 x16 |
Form Factor |
4.4" H x 10.5" L Dual Slot |
Thermal Solution |
Active Fansink |
Maximum Power Consumption |
200 W |
Power Connector |
1x 8-pin PCIe |
Max Digital Resolution |
7680 x 4320 x36 bpp at 60 Hz |
NVIDIA 3D Vision and 3D Vision Pro |
Optional 3-pin mini DIN connector bracket |
Frame Lock |
Compatible with NVIDIA Quadro Sync II |
NVLink |
2-way low profile (2-slot and 3-slot bridges) connects 2x NVIDIA RTX A4500 |
NVLink Interconnect |
112.5 GB/sec bidirectional |