The first time you use Feature Assistant, set parameters for GPU selection, processing precision, and the server-client communication port before you start this tutorial.

  1. From the ENVI main menu, select File > Preferences. The Preferences dialog appears.

  2. Scroll to the bottom of the Preferences options and select Feature Assistant to display the preferences:

  3. Set the preferences as follows:

    • GPU Identifier: Keep the default of 0 to use the primary NVIDIA GPU. Or, if multiple supported GPUs are installed, specify the GPU ID of the device to use.

      To view available NVIDIA GPUs and their associated device IDs, run the following command from a Command Prompt (CMD), PowerShell, or Linux Terminal:

      nvidia-smi --query-gpu=index,name,driver_version,memory.total,compute_cap --format=csv,noheader

      Example output:

      0, NVIDIA RTX 6000 Ada Generation, 596.41, 49140 MiB, 8.9 (where 0 selects RTX 6000 Ada Generation)

      1, NVIDIA RTX A4500, 596.41, 20470 MiB, 8.6 (where 1 selects RTX A4500)

      Note: Before configuring the GPU Identifier, verify that the selected GPU meets all Feature Assistant hardware, driver, and CUDA requirements. See the "System Requirements" section in ENVI Feature Assistant Help for supported NVIDIA GPU specifications and installation prerequisites. Feature Assistant will verify GPU settings for the specified device ID and report compatibility issues.

    • Float 16-bit: Feature Assistant supports Float 16 (Half Precision) and Float 32 (Single Precision) inference modes. The default is Float 32. The selection affects processing speed, GPU memory consumption, and detection accuracy. If your GPU supports F16, you can change the preference to Yes and compare results between the two settings. Otherwise, keep the setting at No to use FP32 single precision processing.
    • Port: The default port for the server is 8765; change the port if the default port is not available
  4. Click OK.