diff --git a/README.md b/README.md
index bfcceaf18dd4b8b45f33853e4e8e1911a0809987..eef34c2ac2d030330917f9141640e4a88bf83a31 100644
--- a/README.md
+++ b/README.md
@@ -282,7 +282,13 @@ This can be done by setting the environment variable
 compute capabilities of the target GPU architectures that kernels will
 be built for. Entries can be separated using semicola. Compute
 capabilities for various CUDA-enabled GPUs can be found on this page:
-https://developer.nvidia.com/cuda-gpus
+https://developer.nvidia.com/cuda-gpus  
+You can also manually check your CUDA-enabled GPU's compute capability
+using:
+
+```shell
+nvidia-smi --query-gpu=compute_cap --format=csv,noheader
+```
 
 For example, if we want to build kernels that target NVIDIA V100
 (compute capability 7.0), NVIDIA A100 (compute capability 8.0), and