3080 support

[zhangyux15]

Description

I run tensorrt sample with 3080 failed, but works for 2080ti by setdevice.

Environment

TensorRT Version: 7.1.3.4 GPU Type: 3080 Nvidia Driver Version: 456.38 CUDA Version: 11.0 CUDNN Version: 8.0 Operating System + Version: Windows Python Version (if applicable): TensorFlow Version (if applicable): PyTorch Version (if applicable): Baremetal or Container (which commit + image + tag):

Relevant Files

TensorRT.sample_onnx_mnist

Steps To Reproduce

I just run the onnx sample:

RUNNING TensorRT.sample_onnx_mnist # E:\TensorRT-7.1.3.4.Windows10.x86_64.cuda-11.0.cudnn8.0\TensorRT-7.1.3.4\samples\sampleOnnxMNIST\....\bin\sample_onnx_mnist.exe [09/27/2020-02:51:07] [I] Building and running a GPU inference engine for Onnx MNIST

Input filename: ../../data/mnist/mnist.onnx ONNX IR version: 0.0.3 Opset version: 8 Producer name: CNTK Producer version: 2.5.1 Domain: ai.cntk Model version: 1 Doc string:

[09/27/2020-02:51:27] [W] [TRT] onnx2trt_utils.cpp:220: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32. [09/27/2020-02:51:28] [E] [TRT] C:\source\rtSafe\cuda\caskUtils.cpp (98) - Assertion Error in nvinfer1::rt::task::trtSmToCask: 0 (Unsupported SM.) &&&& FAILED TensorRT.sample_onnx_mnist # E:\TensorRT-7.1.3.4.Windows10.x86_64.cuda-11.0.cudnn8.0\TensorRT-7.1.3.4\samples\sampleOnnxMNIST\....\bin\sample_onnx_mnist.exe

[zerollzeng]

I am wondering what's the TensorRT 8.0 release date, in next major version the 30x series device should be supported.

[mk-nvidia]

We're in the process of releasing an updated TensorRT version 7.2.x very soon which will add support for RTX 30xx GPUs, thanks for your patience.

[zerollzeng]

So when will 8.x cversion comes out? 😃

Mayank Kaushik notifications@github.com 于 2020年10月1日周四 上午12:22写道：

We're in the process of releasing an updated TensorRT version 7.2.x very soon which will add support for RTX 30xx GPUs, thanks for your patience.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/NVIDIA/TensorRT/issues/800#issuecomment-701497435, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJED7GBMPCFWXDWWH7SNXVLSINLK7ANCNFSM4R3B3OTQ .

[xkszltl]

@mk-nvidia Any timeline/ETA for the very soon 7.2.x?

[mk-nvidia]

@xkszltl 7.2.1 is now live.

[sherlockchou86]

Seems 7.2.1 would not work with cuda 11.1 on 3080. Anybody has tested it successfully on 3080?

[dirtmaxim]

@sherlockchou86 I tested, it works. I built TensorFlow with 7.2.1 support https://github.com/yaroslavvb/tensorflow-community-wheels/issues/168

[sherlockchou86]

it works now！my mistake as the driver was not installed successfully with the right version

[pustar]

We're in the process of releasing an updated TensorRT version 7.2.x very soon which will add support for RTX 30xx GPUs, thanks for your patience.

I am facing 'Assertion Error in nvinfer1::rt::task::trtSmToCask: 0 (Unsupported SM.)' problem, with RTX 3060, cuda 11.0, win10, cudnn 8.1

[pustar]

it works now！my mistake as the driver was not installed successfully with the right version

Would you please tell me the proper version of cuda and cudnn which work fine with Tensorrt 7.2.4? @sherlockchou86

[pustar]

Which version of tensorflow did you use? Which version of cuda and tensorrt? Can you tell me? Thank you very much @dirtmaxim

@sherlockchou86 I tested, it works. I built TensorFlow with 7.2.1 support yaroslavvb/tensorflow-community-wheels#168