I need an application that use Tesla T4

I need a an application that shows that the Tesla T4 installed in my desktop computer works.

You can compile and run CUDA samples to confirm that Tesla T4 works. CUDA samples are available at GitHub - NVIDIA/cuda-samples: Samples for CUDA Developers which demonstrates features in CUDA Toolkit.

Our desktop computer MSI Tomahawk Arctic Z270 is running Windows 10, and Tesla T4 GPU shows up in Device Manager > Display adapters. When NVDIA Tesla T4 is clicked, NVIDIA T4 Properties popup is displayed, and General > Device status shows “The device is working properly.”

When dxdiag is run, page DirectX Diagnostic Tool is opened.

A file obtained as “Save All Information” created by DirectX Diagnostic Tool contains the following information:

Display Devices

       Card name: Intel(R) HD Graphics 530
    Manufacturer: Intel Corporation
       Chip type: Intel(R) HD Graphics Family

(Nvidia Tesla T4 is not listed as a Display Device.
(However it is listed as a System Device.)

System Devices

 . . .

Name: NVIDIA Tesla T4
Device ID: PCI\VEN_10DE&DEV_1EB8&SUBSYS_12A210DE&REV_A1\4&2C42DEA8&0&0008
. . .
Driver: C:\Program Files\NVIDIA
. . .
(Intel (R) HD Graphics 530 is not listed as a System Device).
. . .


I do not know yet whether our Tesla T4 can work correctly or not. I will appreciate any useful information and an application that uses Tesla T4.

I started to install CUDA. However, I did not install Visual Studio completely. Therefore, most of the CUDA applications were not installed. I want to reinstall Visual Studio and CUDA.

Should I remove the installed Visual Studio and CUDA apps before they are reinstalled?

Yo can reinstall Vsual Studio and CUDA without removing them. They are automatically removed when their reinstallation starts.

It is important to install Visual Studio before CUDA. Otherwise, one file that CUDA installs in a directory of Visual Studio will be erased when the Visual Studio is reinstalled. in other words, When Visual Studio is reinstalled, CUDA must be reinstalled. I learned these during the last seeral days by making mistakes.

Thank you for your reply. I downloaded all the samples from

and build .exe files and executed them. by the following steps.

Step 0. As of May 2022, sample files of GetHub uses CUDA 11.6 although the latest version of CUDA is CUDA 11.7. Therefore, CUDA 11.6 must be be downloaded from the archived Versions page and installed. The Community Version of Visual Studio is free, and it must be installed before CUDA is installed.

Step 1. Repository of samples at GitHUb can be downloaded by Visual Studio. Create an empty directory as C:\Users\xxxx\source\repos\yyyy
where xxxx is your name and yyyy is the dirctory, e.g., GitHub Samples, where downloaded samples are stored.

Step 2. When Visual Studio is opened without any projct, you can download samples from GitHub by clicking “Clone a repository” and by providing the following data:

Repository location

Path
C:\Users\xxxx\source\repos\yyyy

Downloaded samples are grouped in the following subdirectories in
C:\Users\xxxx\source\repos\yyyy\Samples

0_Introduction
1_utilities
2_Concepts_and_Techniques
3_CUDA_Features
4_CUDA_Libraries
5_Domain_Specific
6-Performance

A downloaded sample can be slected by Visual Studio for creating an .exe file and executing.

Step 3. File > Open > (xxxx > source > repos > yyyy > samples in a popup)
Then select a group of samples, e.g., 3_CUDA_Features.

Step 4. Select a sample, say, cdpQuadtree > cdQuadtree_vs2022.sln > Open

Step 5. Build > Build Solution generates cdpQuadtree.exe.

Step 6. Debug > Start Without Debugging executes cdpQuadtree.exe.

My current problem is CUDA Error: code=304(cudaErrorOperatingSystem) “cudaGraphicsGLRegisterBuffer( &cuda_pbo_resource, pbo, cudaGraphicsMapFlagsWriteDiscard)”.
This error occurs when NTsla T4 tries to connect to the display adapter.

The problem discussed in this thread occured when I was using Intel(R) HD Graphics 530 as a graphics card. I replaced it by Nvidia GeForce GT 720. Then graphics output was correctly displaced.

I am now using CUDA 11.8, and GitHub Samples that produce graphical output do not work because of OpenGL problems.