GPU 型号与计算能力

在使用 Vllm 时，使用 V100 显卡和 A100 显卡会显示，V100 不支持 BF16 精度。那么具体的 GPU 型号和支持精度的关系是怎么样的？

Hardware and Precision

备注

Support Matrix — NVIDIA TensorRT Documentation

CUDA Compute Capability	Example Devices	TF32	FP32	FP16	FP8	FP4	BF16	INT8	FP16 Tensor Cores	INT8 Tensor Cores	DLA
12.0	NVIDIA RTX 5090	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
10.0	NVIDIA B200	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
9.0	NVIDIA H100	Yes	Yes	Yes	Yes	Yes 5	Yes	Yes	Yes	Yes	No
9.0	NVIDIA GH200 480 GB	Yes	Yes	Yes	Yes	Yes 5	Yes	Yes	Yes	Yes	No
8.9	NVIDIA L40S	Yes	Yes	Yes	Yes	Yes 5	Yes	Yes	Yes	Yes	No
8.7	NVIDIA DRIVE AGX Orin	Yes	Yes	Yes	No	No	No	Yes	Yes	Yes	Yes
8.6	NVIDIA A10	Yes	Yes	Yes	No	No	Yes	Yes	Yes	Yes	No
8.0	NVIDIA A100	Yes	Yes	Yes	No	No	Yes	Yes	Yes	Yes	No
7.5	NVIDIA T4	No	Yes	Yes	No	No	No	Yes	Yes	Yes	No

GPU Compute Capability

备注

CUDA GPUs - Compute Capability | NVIDIA Developer

GPU	Compute Capability
NVIDIA H100	9.0
NVIDIA L4	8.9
NVIDIA L40	8.9
NVIDIA A100	8.0
NVIDIA A40	8.6
NVIDIA A30	8.0
NVIDIA A10	8.6
NVIDIA A16	8.6
NVIDIA A2	8.6
NVIDIA T4	7.5
NVIDIA V100	7.0
Tesla P100	6.0
Tesla P40	6.1
Tesla P4	6.1
Tesla M60	5.2
Tesla M40	5.2
Tesla K80	3.7
Tesla K40	3.5
Tesla K20	3.5
Tesla K10	3.0