NVIDIA 显卡硬件支持的精度模式
很多炼丹师不知道自己英伟达显卡支持哪些精度模式，本文整理了NVIDIA官网的数据，为你解开疑惑。
1. 首先了解CUDA计算能力及其支持的精度模式；
2. 查看自己显卡（或其它NVIDIA硬件）的计算能力值为多少。
表1 CUDA计算能力及其支持的精度模式
| CUDA Compute Capability | TF32 | FP32 | FP16 | INT8 | FP16 Tensor Cores | INT8 Tensor Cores | DLA |
| 9 | Yes | Yes | Yes | Yes | Yes | Yes | No |
| 8.9 | Yes | Yes | Yes | Yes | Yes | Yes | No |
| 8.7 | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 8.6 | Yes | Yes | Yes | Yes | Yes | Yes | No |
| 8 | Yes | Yes | Yes | Yes | Yes | Yes | No |
| 7.5 | No | Yes | Yes | Yes | Yes | Yes | No |
| 7.2 | No | Yes | Yes | Yes | Yes | Yes | Yes |
| 7 | No | Yes | Yes | Yes | Yes | No | No |
| 6.1 | No | Yes | Yes | Yes | No | No | No |
| 6 | No | Yes | Yes | No | No | No | No |
表2 NVIDIA 硬件（包含显卡、嵌入式板卡等）对应的计算能力
| GPU | Compute Capability |
| NVIDIA H100 | 9 |
| NVIDIA L4 | 8.9 |
| NVIDIA L40 | 8.9 |
| RTX 6000 | 8.9 |
| GeForce RTX 4090 | 8.9 |
| GeForce RTX 4080 | 8.9 |
| GeForce RTX 4070 Ti | 8.9 |
| GeForce RTX 4070 | 8.9 |
| GeForce RTX 4060 | 8.9 |
| GeForce RTX 4050 | 8.9 |
| Jetson AGX Orin | 8.7 |
| Jetson Orin NX | 8.7 |
| Jetson Orin Nano | 8.7 |
| NVIDIA A40 | 8.6 |
| NVIDIA A10 | 8.6 |
| NVIDIA A16 | 8.6 |
| NVIDIA A2 | 8.6 |
| RTX A6000 | 8.6 |
| RTX A5000 | 8.6 |
| RTX A4000 | 8.6 |
| RTX A3000 | 8.6 |
| RTX A2000 | 8.6 |
| GeForce RTX 3090 Ti | 8.6 |
| GeForce RTX 3090 | 8.6 |
| GeForce RTX 3080 Ti | 8.6 |
| GeForce RTX 3080 | 8.6 |
| GeForce RTX 3070 Ti | 8.6 |
| GeForce RTX 3070 | 8.6 |
| Geforce RTX 3060 Ti | 8.6 |
| Geforce RTX 3060 | 8.6 |
| GeForce RTX 3050 Ti | 8.6 |
| GeForce RTX 3050 | 8.6 |
| NVIDIA A100 | 8 |
| NVIDIA A30 | 8 |
| NVIDIA T4 | 7.5 |
| Quadro RTX 8000 | 7.5 |
| Quadro RTX 6000 | 7.5 |
| Quadro RTX 5000 | 7.5 |
| Quadro RTX 4000 | 7.5 |
| RTX 5000 | 7.5 |
| RTX 4000 | 7.5 |
| RTX 3000 | 7.5 |
| T2000 | 7.5 |
| T1200 | 7.5 |
| T1000 | 7.5 |
| T600 | 7.5 |
| T500 | 7.5 |
| T400 | 7.5 |
| GeForce GTX 1650 Ti | 7.5 |
| NVIDIA TITAN RTX | 7.5 |
| Geforce RTX 2080 Ti | 7.5 |
| Geforce RTX 2080 | 7.5 |
| Geforce RTX 2070 | 7.5 |
| Geforce RTX 2060 | 7.5 |
| Jetson AGX Xavier | 7.2 |
| Jetson Xavier NX | 7.2 |
| NVIDIA V100 | 7 |
| Quadro GV100 | 7 |
| NVIDIA TITAN V | 7 |
| Jetson TX2 | 6.2 |
| Tesla P40 | 6.1 |
| Tesla P4 | 6.1 |
| Quadro P6000 | 6.1 |
| Quadro P5200 | 6.1 |
| Quadro P5000 | 6.1 |
| Quadro P4200 | 6.1 |
| Quadro P4000 | 6.1 |
| Quadro P3200 | 6.1 |
| Quadro P3000 | 6.1 |
| Quadro P2200 | 6.1 |
| Quadro P2000 | 6.1 |
| Quadro P1000 | 6.1 |
| Quadro P620 | 6.1 |
| Quadro P600 | 6.1 |
| Quadro P500 | 6.1 |
| Quadro P400 | 6.1 |
| P620 | 6.1 |
| P520 | 6.1 |
| NVIDIA TITAN Xp | 6.1 |
| NVIDIA TITAN X | 6.1 |
| GeForce GTX 1080 Ti | 6.1 |
| GeForce GTX 1080 | 6.1 |
| GeForce GTX 1070 Ti | 6.1 |
| GeForce GTX 1070 | 6.1 |
| GeForce GTX 1060 | 6.1 |
| GeForce GTX 1050 | 6.1 |
| Tesla P100 | 6 |
| Quadro GP100 | 6 |
| Jetson Nano | 5.3 |
通过以上两表，可了解每个硬件支持的精度模式。
参考：
Support Matrix :: NVIDIA Deep Learning TensorRT Documentation
CUDA GPUs - Compute Capability | NVIDIA Developer