Skip to main content

NVENC Not Detected

If Alchemist is running on a host with an NVIDIA GPU and NVENC isn't being used, the problem is almost always one of three things: the driver isn't visible to the process, FFmpeg wasn't built with NVENC support, or the startup probe failed for a specific codec. This page walks through all three in order.

Start with Settings → Hardware → Probe Log. Alchemist probes every plausible encoder at startup with a short FFmpeg test encode and records the stderr for every failure. That log is the authoritative answer to "why didn't NVENC work?" — most of the checks below are just narrowing down what the probe log already says.

1. Is the driver visible?​

On the host:

nvidia-smi

If that command fails or hangs, the driver is the problem, not Alchemist. Reinstall or reload the NVIDIA kernel module before continuing.

In Docker, nvidia-smi inside the container is the real test:

docker run --rm --gpus all nvidia/cuda:12.0.0-base-ubuntu22.04 nvidia-smi

If this fails, the container can't see the GPU. That's almost always a missing or misconfigured nvidia-container-toolkit. Reinstall it per the NVIDIA hardware guide and restart Docker.

2. Does FFmpeg expose NVENC?​

Inside the container (or on the host for a binary install):

ffmpeg -encoders | grep nvenc

You should see at least:

  • h264_nvenc
  • hevc_nvenc

On RTX 40 (Ada) series, you should also see av1_nvenc. If the list is empty, FFmpeg in that environment was not compiled with NVENC. For Docker, use the official Alchemist image — it ships FFmpeg with NVENC enabled. For a binary install, install an FFmpeg build that includes NVENC support (Debian/Ubuntu, Fedora, Arch, and Homebrew FFmpeg packages all enable it by default, but distro forks differ).

3. Does the probe actually succeed?​

Open Settings → Hardware → Probe Log. Each entry shows:

  • The encoder being tested (h264_nvenc, hevc_nvenc, av1_nvenc).
  • Whether the probe succeeded or failed.
  • The full FFmpeg stderr for failures.

Common probe-level failures:

  • "Driver does not support the required NVENC features" — update the NVIDIA driver; FFmpeg's NVENC build is ahead of your driver.
  • "No capable devices found" — GPU isn't visible to the process (go back to step 1).
  • av1_nvenc fails but hevc_nvenc / h264_nvenc succeed — your card is pre-Ada. AV1 NVENC requires an RTX 40 (Ada) card; RTX 30 (Ampere) and older decode AV1 but cannot encode it. Target HEVC or H.264 instead, or see Codecs.

4. Is the right vendor selected?​

If the probe succeeded but NVENC still isn't being used, check Settings → Hardware → Preferred Vendor. auto selects based on scoring; setting it to nvidia forces NVENC when it's available.

Leave Device Path empty. NVENC is selected from the driver and /dev/nvidiactl, not a render-node path.

5. What to check when the probe log is empty​

If no probes ran at all for NVENC, Alchemist didn't see the GPU as a candidate. That usually means:

  • The NVIDIA driver is not loaded (step 1).
  • Docker's --gpus all flag (or Compose deploy.resources block) isn't set. See GPU Passthrough.
  • You're running on a host that truly has no NVIDIA GPU — confirm with lspci | grep -i nvidia.

Still stuck?​

Check the related pages for the platform-specific paths, and the broader troubleshooting overview:

If the probe log points at a specific FFmpeg error that isn't covered here, that stderr is the right thing to paste into a GitHub issue: github.com/bybrooklyn/alchemist/issues.