diff --git a/README.md b/README.md
index d628736..63a5655 100644
--- a/README.md
+++ b/README.md
@@ -8,17 +8,12 @@ TensorRTx has a brother project [Pytorchx](https://github.com/wang-xinyu/pytorch
 
 ## Test Environment
 
-Jetson TX1
+1. Jetson TX1 / Ubuntu16.04 / cuda9.0 / cudnn7.1.5 / tensorrt4.0.2 / nvinfer4.1.3
 
-Ubuntu16.04
+2. GTX1080 / Ubuntu16.04 / cuda10.0 / cudnn7.6.5 / tensorrt7.0.0 / nvinfer7.0.0
 
-cuda9.0
+Currently, TX1 ans x86 GTX1080 were tested. trt4 api were using, some api are deprecated in trt7, but still can compile successfully.
 
-cudnn7.1.5
-
-tensorrt4.0.2/nvinfer4.1.3
-
-Currently, I only test it on TX1, But I think it will be totally OK on TX2 with same software version. And also it will be easy be ported to x86.
 
 ## Models