Transcoding accelerators

Transcoding accelerators #

Introduced in Nimbra Edge R3.25

Supported accelerators #

Nimbra Edge supports various hardware accelerators for accelerating video transcoding. The available accelerators depend on the type of appliance.

Kubernetes video appliances #

Supported accelerators:

  • NVIDIA GPUs with driver version 570 or above

Appliances installed with the connectit installer #

Supported accelerators:

  • NVIDIA GPUs with driver version 570 or above
  • NETINT Quadra T1U, T1A, T2A

Encoding Capabilities #

The table below shows the encoding capabilities of the supported accelerators and CPU software encoding.

Feature NVIDIA GPU NVIDIA GPU (Blackwell generation or newer) NETINT Quadra CPU (Software)
H.262 (MPEG-2)
H.264 (AVC)
H.265 (HEVC)
4:2:0 8-bit
4:2:2 8-bit ⚠️ ¹
4:4:4 8-bit ⚠️ ¹
4:2:0 10-bit ⚠️ ² ⚠️ ¹
4:2:2 10-bit ⚠️ ¹
4:4:4 10-bit ⚠️ ² ⚠️ ¹
Progressive
Interlaced ⚠️ ⁴ ⚠️ ³
Resolution Scaling
Overlays
Deinterlacing

Legend:

  • ✅ Fully supported
  • ⚠️ Partial support
  • ❌ Not supported

¹ H.264 and H.265 only
² H.265 only
³ H.262 and H.264 only
⁴ H.264 only, no 4:4:4 pixel format, no resolution scaling

Limitations #

When transcoding parameters are not supported by the available accelerator, transcoding will automatically fall back to CPU software encoding. This is represented in the Encoder Compatibility section:

encoder compatibility

CPU software encoding is significantly slower than hardware acceleration and may not achieve acceptable performance, especially at higher resolutions and bitrates. Use hardware accelerators whenever possible for production workloads, or deploy on high-performance CPU nodes dedicated to transcoding tasks.

Accelerator Details #

NVIDIA GPU #

For more information on specific devices, see the NVIDIA Video Encode and Decode Support Matrix.

NETINT Quadra #

The specific capacity of each device is specified on the NETINT website.