Discrete GPU (dGPU) Support

The IGX Thor Developer Kit comes equipped with a discrete GPU: the NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition. Both the integrated GPU (iGPU) and dGPU share the same device driver on the IGX Thor platform.

By default, display and graphics output are enabled on the iGPU. During operating system installation, you must connect your monitor to the iGPU display port. While the system uses the iGPU for display and graphics by default, the dGPU is available for compute tasks.

Using the dGPU for Compute Tasks

NVIDIA_VISIBLE_DEVICES environment variable can be used to specify the GPU to use for compute tasks. For example, to use the dGPU for compute tasks, you can set the following environment variable:

export CUDA_VISIBLE_DEVICES=0

Note

CUDA default order using “compute power”, so 0 is the more powerful GPU (dGPU), 1 is the less powerful GPU (iGPU).

An example of using the dGPU for compute tasks is as follows:

git clone --depth 1 https://github.com/NVIDIA/cuda-samples.git
cd cuda-samples
mkdir build
cd build
cmake -DCMAKE_CUDA_COMPILER=/usr/local/cuda-13.0/bin/nvcc ..
cd Samples/1_Utilities/deviceQuery
make
./deviceQuery

You should see output similar to the following:

./deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition"
CUDA Driver Version / Runtime Version          13.0 / 13.0
CUDA Capability Major/Minor version number:    12.0
Total amount of global memory:                 97249 MBytes (101972639744 bytes)
...

Note

You may need to install cmake and cuda-toolkit-13-0 to build the cuda-samples.

Enabling dGPU Display and Graphics

To use the dGPU for display and graphics workloads, you need to complete a manual configuration to switch display output from the iGPU to the dGPU. The steps are as follows:

1. Connect your monitor to the iGPU display port.

2. Boot up the system.

3. Update environment and configuration files to enable dGPU graphics:

   • Add the following line to /etc/environment to disable iGPU graphics:

   export NV_DISABLE_IGPU_GRAPHICS=1
   

   • Edit the /etc/X11/xorg.conf.pci file to specify the dGPU as the display device. Your configuration should look similar to the following:

   # Disable extensions not useful on Tegra.
   Section "Module"
      Disable     "dri"
      SubSection  "extmod"
         Option  "omit xfree86-dga"
      EndSubSection
   EndSection
   
   Section "Device"
      Identifier  "Tegra0"
      Driver      "nvidia"
      Option      "AllowEmptyInitialConfiguration" "true"
      #    BusID       "PCI:1:0:0"
      BusID       "PCI:9@4:0:0"
   EndSection
   

4. Power off the system and move the monitor connection to the dGPU display port.

5. Power on the system.

6. Verify that the display is now connected to and driven by the dGPU.