内容摘自官方文档:
https://docs.nvidia.com/deeplearning/tensorrt/support-matrix/index.html#hardware-precision-matrix
The following table lists NVIDIA hardware and which precision modes that each hardware supports. TensorRT supports all NVIDIA hardware with capability SM 5.0 or higher. It also lists the availability of DLA on this hardware. Refer to the following tables for the specifics.
下表列出了 NVIDIA 硬件以及每个硬件支持的精度模式。 TensorRT 支持所有具有 SM 5.0 或更高功能的 NVIDIA 硬件。 它还列出了此硬件上 DLA 的可用性。 具体请参考下表。
Note: Support for CUDA compute capability version 3.0 has been removed. Support for CUDA compute capability versions below 5.0 may be removed in a future release and is now deprecated.
注意:已删除对 CUDA 计算能力版本 3.0 的支持。 未来版本中可能会删除对 5.0 以下的 CUDA 计算能力版本的支持,现在已弃用。
关于 CUDA Compute Capability 可以参考:
https://blog.csdn.net/HaoZiHuang/article/details/125834480
或者
https://developer.nvidia.com/cuda-gpus
CUDA Compute Capability | Example Device | TF32 | FP32 | FP16 | INT8 | FP16 Tensor Cores | INT8 Tensor Cores | DLA |
---|---|---|---|---|---|---|---|---|
8.7 | NVIDIA DRIVE AGX Orin™ | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
8.6 | NVIDIA A10 | Yes | Yes | Yes | Yes | Yes | Yes | No |
8.0 | NVIDIA A100/GA100 GPU | Yes | Yes | Yes | Yes | Yes | Yes | No |
7.5 | NVIDIA T4 | No | Yes | Yes | Yes | Yes | Yes | No |
7.2 | Jetson AGX Xavier | No | Yes | Yes | Yes | Yes | Yes | Yes |
7.0 | NVIDIA V100 | No | Yes | Yes | Yes | Yes | No | No |
6.1 | NVIDIA P4 | No | Yes | No | Yes | No | No | No |
6.0 | NVIDIA P100 | No | Yes | Yes | No | No | No | No |
5.2 | NVIDIA M4 | No | Yes | No | No | No | No | No |
5.0 | Quadro K2200 | No | Yes | No | No | No | No | No |
CUDA Compute Capability | Example Device | FP32 | FP16 | INT8 | FP16 Tensor Cores | INT8 Tensor Cores | DLA |
---|---|---|---|---|---|---|---|
3.7 | NVIDIA K80 | Yes | No | No | No | No | No |
3.5 | NVIDIA K40 | Yes | No | No | No | No | No |
CUDA Compute Capability | Example Device | FP32 | FP16 | INT8 | FP16 Tensor Cores | INT8 Tensor Cores | DLA |
---|---|---|---|---|---|---|---|
3.0 | NVIDIA K10 | Yes | No | No | No | No | No |
最后必须说明一下:
根据老黄的文档,这个计算能力不是计算性能,他只是标识了gpu能做什么样的工作。
如果计算能力在5.2(不含)以下,根本就不能用cuda。
要比较计算性能,可以参考flops参数跟显卡的售价。
摘自知乎:
https://www.zhihu.com/question/44769165/answer/318719848
小白要注意辨别,CUDA版本和显卡CUDA Compute Capability 二者不是一个东西!!!!