I'm unable to see any gpu attached with my g4dn.xlarge instances .

0

Hello I created g4dn.xlarge instances with which I should be provided with 1core NVIDIA T4 Tensor GPU , but when I run hardware monitor tool I saw there is no information related to any attachement of GPU , while we updated all drivers but till now we are unable to resolve problem .

Alpha
asked 8 months ago283 views
2 Answers
3

If you are running Windows:

By default, Amazon Quick Start Windows AMIs do not include device drivers for graphics processors. GPU-based instances can't access the GPU unless the NVIDIA drivers are installed. [1] (I know you stated this was already complete, but just in case I will include)

1. From http://www.nvidia.com/Download/Find.aspx , input the following for a G4 instance type:
	Product Type: Tesla
	Product Seriess: T-Series
	Product: T4
2. Click Search and install your desired driver version
3. Open the folder where you downloaded the driver and launch the installation file. Follow the instructions to install the driver and reboot your instance as required
4. Disable the built-in display adapter using Device Manager
5. Install these Windows features: Media Foundation and Quality Windows Audio Video Experience in the Server Manager
This is done by following the following instructions:
	a. Launch Server Manager (ServerManager.exe)
	b. Navigate to Manage
	c. Add Roles and Features
	d. Click next until you reach the Features section 
	e. Scroll Down and Select Media Foundation +Quality Windows Audio Video Experience
	f. Click Next and Install
6. Check Device Manager to verify that the GPU is working correctly
7. To achieve the best performance from your GPU, complete the optimization steps in the attached documentation[2]

When I initially launch the instance and pull up device manager there is a "3D Video Controller" listed under "other devices". I then installed the driver version 537.13 from the steps above. After the installation, the "3D Video Controller" is no longer listed and I now see "NVIDIA Tesla T4" under "Display adapters".

Now getting to this point I also noticed that task manager was not showing the GPU in the performance tab. If you are also running into the same issue follow these steps.

1. Open Device Manager
2. Right click "NVIDIA Tesla T4"  and select "Uninstall Device"
3. Select checkbox to delete the driver
4. Follow these instructions [4] to install the GRID driver

After installing this driver I was able to open Task manager and view GPU usage in the "Performance Tab"

[1] AWS Accelerated Computing - Installing NVIDIA Drivers https://docs.aws.amazon.com/AWSEC2/latest/WindowsGuide/install-nvidia-driver.html#nvidia-installation-options

[2] Optimizing GPU settings https://docs.aws.amazon.com/AWSEC2/latest/WindowsGuide/optimize_gpu.html

[3] GRID Driver install option https://docs.aws.amazon.com/AWSEC2/latest/WindowsGuide/install-nvidia-driver.html#nvidia-GRID-driver

AWS
answered 8 months ago
0

What platform are you running? This blog, details how to monitor the GPU with CloudWatch.

With Linux, a quick way to check is to run lshw -C video. Doing so on an instance of my own showed this response:

 *-display:1 UNCLAIMED
      description: 3D controller
      product: TU104GL [Tesla T4]
      vendor: NVIDIA Corporation
      physical id: 1e
      bus info: pci@0000:00:1e.0
      version: a1
      width: 64 bits
      clock: 33MHz
      capabilities: pm pciexpress msix bus_master cap_list
      configuration: latency=0
      resources: iomemory:10000-ffff iomemory:10000-ffff memory:c0000000-c0ffffff memory:100000000000-10000fffffff memory:100010000000-100011ffffff ```
AWS
answered 8 months ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions