Hardware Acceleration

Configure GPU-accelerated transcoding with Intel QSV, NVIDIA NVENC, AMD AMF, and VA-API

Hardware acceleration uses your GPU to transcode media, providing 10-20x faster performance compared to CPU-only encoding while significantly reducing CPU usage.

Supported Hardware Acceleration Types

Jellyfin supports multiple hardware acceleration platforms:

Intel Quick Sync

Intel integrated graphics (6th gen+)

NVIDIA NVENC

NVIDIA GPUs (GTX 900+ series)

AMD AMF

AMD GPUs (GCN architecture+)

VA-API

Linux video acceleration API

VideoToolbox

Apple hardware (macOS/iOS)

Video4Linux

V4L2 M2M (Raspberry Pi, etc.)

Rockchip MPP

Rockchip SoC devices

Software

CPU-only encoding (no GPU)

Hardware Acceleration Configuration

Enabling Hardware Acceleration

In encoding.xml:

<HardwareAccelerationType>qsv</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>

Available types:

• none - Software encoding only
• qsv - Intel Quick Sync Video
• nvenc - NVIDIA NVENC
• amf - AMD Advanced Media Framework
• vaapi - VA-API (Linux)
• videotoolbox - Apple VideoToolbox
• v4l2m2m - Video4Linux M2M
• rkmpp - Rockchip Media Process Platform

Intel Quick Sync Video (QSV)

Requirements

1. Hardware Requirements

   • Intel CPU with integrated graphics (6th generation or newer recommended)
   • Intel Iris Xe or UHD Graphics
   • Compatible chipset with Quick Sync support
2. Driver Requirements

   Linux:

   # Install Intel media drivers
   sudo apt-get install intel-media-va-driver-non-free
   

   Windows: Install latest Intel graphics drivers from Intel’s website.

3. Device Permissions (Linux)

   Add Jellyfin user to video and render groups:

   sudo usermod -aG video jellyfin
   sudo usermod -aG render jellyfin
   sudo systemctl restart jellyfin
   

QSV Configuration

<HardwareAccelerationType>qsv</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>
<QsvDevice></QsvDevice>
<EnableIntelLowPowerH264HwEncoder>false</EnableIntelLowPowerH264HwEncoder>
<EnableIntelLowPowerHevcHwEncoder>false</EnableIntelLowPowerHevcHwEncoder>

Standard QSV

For most Intel GPUs (6th-11th gen):

<HardwareAccelerationType>qsv</HardwareAccelerationType>
<EnableIntelLowPowerH264HwEncoder>false</EnableIntelLowPowerH264HwEncoder>
<EnableIntelLowPowerHevcHwEncoder>false</EnableIntelLowPowerHevcHwEncoder>

Low Power Mode

For Intel Xe graphics (11th gen+):

<HardwareAccelerationType>qsv</HardwareAccelerationType>
<EnableIntelLowPowerH264HwEncoder>true</EnableIntelLowPowerH264HwEncoder>
<EnableIntelLowPowerHevcHwEncoder>true</EnableIntelLowPowerHevcHwEncoder>

Low power mode provides better efficiency on newer architectures.

Supported Codecs (QSV)

Codec	Decode	Encode	Notes
H.264	✓	✓	All QSV generations
HEVC	✓	✓	6th gen+
VP9	✓	✗	9th gen+ (decode only)
AV1	✓	✓	12th gen+
MPEG-2	✓	✓	All generations
VC-1	✓	✗	Decode only

QSV Tone Mapping

<EnableVppTonemapping>true</EnableVppTonemapping>
<VppTonemappingBrightness>16</VppTonemappingBrightness>
<VppTonemappingContrast>1</VppTonemappingContrast>

Intel VPP (Video Post Processing) provides hardware-accelerated HDR to SDR tone mapping on supported platforms.

NVIDIA NVENC

Requirements

1. Hardware Requirements

   • NVIDIA GPU with NVENC support:
     • GeForce GTX 900 series or newer
     • RTX series (best performance)
     • Quadro/Tesla professional GPUs
2. Driver Requirements

   Linux:

   # Install NVIDIA drivers
   sudo apt-get install nvidia-driver-535
   
   # Verify installation
   nvidia-smi
   

   Windows: Install latest GeForce or Studio drivers from NVIDIA’s website.

3. NVENC Session Limits

   Consumer GPUs have encoding session limits:

   • GTX/RTX cards: 3 concurrent sessions (can be patched)
   • Quadro/Tesla: Unlimited sessions

   Use nvidia-patch to remove session limits on consumer cards.

NVENC Configuration

<HardwareAccelerationType>nvenc</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>
<EnableEnhancedNvdecDecoder>true</EnableEnhancedNvdecDecoder>

Enhanced NVDEC

<EnableEnhancedNvdecDecoder>true</EnableEnhancedNvdecDecoder>

Enables enhanced NVDEC decoder for:

• Dolby Vision to SDR tone mapping
• Better HDR support
• Improved compatibility

Supported Codecs (NVENC)

Codec	Decode	Encode	Generation
H.264	✓	✓	All NVENC
HEVC	✓	✓	Maxwell+ (GTX 900+)
VP9	✓	✗	Pascal+ (GTX 10xx+)
AV1	✓	✓	Ada Lovelace (RTX 40xx)
MPEG-2	✓	✗	Decode only
VC-1	✓	✗	Decode only

CUDA Tone Mapping

NVENC supports CUDA-accelerated tone mapping:

<EnableTonemapping>true</EnableTonemapping>
<TonemappingAlgorithm>bt2390</TonemappingAlgorithm>

CUDA tone mapping requires:

• FFmpeg built with CUDA support
• Compatible NVIDIA GPU (Pascal or newer)

AMD AMF

Requirements

1. Hardware Requirements

   • AMD GPU with GCN architecture or newer:
     • Radeon RX 400 series or newer (recommended)
     • Radeon VII
     • RX 5000/6000/7000 series
2. Driver Requirements

   Linux:

   # Install AMD drivers and VA-API
   sudo apt-get install mesa-va-drivers
   

   Windows: Install AMD Adrenalin drivers from AMD’s website.

AMF Configuration

<HardwareAccelerationType>amf</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>

Supported Codecs (AMF)

Codec	Decode	Encode	Notes
H.264	✓	✓	All GCN GPUs
HEVC	✓	✓	GCN 3.0+
AV1	✓	✓	RDNA 3 (RX 7000+)
VP9	✓	✗	Decode only

On Linux, AMF uses VA-API as the backend. Consider using VA-API directly for better compatibility.

VA-API (Linux)

Requirements

1. Install VA-API Drivers

   Intel

   sudo apt-get install intel-media-va-driver-non-free
   

   AMD

   sudo apt-get install mesa-va-drivers
   

   NVIDIA (Nouveau)

   sudo apt-get install mesa-va-drivers
   

   NVIDIA proprietary drivers don’t support VA-API. Use NVENC instead.
2. Verify VA-API

   # Install vainfo utility
   sudo apt-get install vainfo
   
   # Check VA-API support
   vainfo
   

   Should show available profiles and entrypoints.

3. Device Permissions

   # Check render device
   ls -l /dev/dri/renderD128
   
   # Add user to groups
   sudo usermod -aG video jellyfin
   sudo usermod -aG render jellyfin
   sudo systemctl restart jellyfin
   

VA-API Configuration

<HardwareAccelerationType>vaapi</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>
<VaapiDevice>/dev/dri/renderD128</VaapiDevice>

The default VA-API device is /dev/dri/renderD128, which works for most single-GPU systems.

Multiple GPUs

For systems with multiple GPUs:

# List render devices
ls -l /dev/dri/

# Output:
# renderD128  (usually Intel iGPU)
# renderD129  (usually discrete GPU)

Specify the desired device:

<VaapiDevice>/dev/dri/renderD129</VaapiDevice>

VA-API Tone Mapping

<EnableTonemapping>true</EnableTonemapping>
<TonemappingAlgorithm>bt2390</TonemappingAlgorithm>

VA-API supports hardware tone mapping on Intel GPUs (9th gen+).

Apple VideoToolbox

Requirements

• macOS device with Apple Silicon or Intel CPU with integrated graphics
• macOS 10.13 or newer

VideoToolbox Configuration

<HardwareAccelerationType>videotoolbox</HardwareAccelerationType>
<EnableHardwareEncoding>true</EnableHardwareEncoding>
<EnableVideoToolboxTonemapping>true</EnableVideoToolboxTonemapping>

Supported Codecs (VideoToolbox)

Codec	Decode	Encode	Notes
H.264	✓	✓	All devices
HEVC	✓	✓	A10+ / 6th gen Intel+
VP9	✓	✗	Decode only

Docker Configuration

Intel QSV (Docker)

docker-compose.yml

services:
  jellyfin:
    image: jellyfin/jellyfin:latest
    devices:
      - /dev/dri:/dev/dri
    group_add:
      - "109"  # render group ID
    volumes:
      - /path/to/config:/config
      - /path/to/media:/media

Find your render group ID with: getent group render | cut -d: -f3

NVIDIA (Docker)

docker-compose.yml

services:
  jellyfin:
    image: jellyfin/jellyfin:latest
    runtime: nvidia
    environment:
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=compute,video,utility
    volumes:
      - /path/to/config:/config
      - /path/to/media:/media

Requires NVIDIA Container Toolkit installed on the host.

AMD VA-API (Docker)

services:
  jellyfin:
    image: jellyfin/jellyfin:latest
    devices:
      - /dev/dri:/dev/dri
    group_add:
      - "109"  # render group
      - "44"   # video group
    volumes:
      - /path/to/config:/config
      - /path/to/media:/media

Performance Comparison

Typical 1080p H.264 transcoding performance:

Method	Speed	CPU Usage	Power
Software (libx264)	1x	100%	High
Intel QSV	10-15x	10-20%	Low
NVIDIA NVENC	15-20x	5-10%	Medium
AMD AMF	10-15x	10-15%	Medium
VA-API	10-15x	10-20%	Low

Speed multipliers are approximate and vary based on hardware, source material, and encoding settings.

Troubleshooting

1. Hardware Acceleration Not Working

   Check:

   # Verify device exists
   ls -l /dev/dri/
   
   # Check permissions
   id jellyfin
   
   # Test VA-API (if applicable)
   sudo -u jellyfin vainfo
   
   # Check Jellyfin logs
   journalctl -u jellyfin -f
   
2. Permission Denied Errors

   # Add user to required groups
   sudo usermod -aG video,render jellyfin
   sudo systemctl restart jellyfin
   
   # Verify group membership
   groups jellyfin
   
3. Quality Issues

   • Reduce CRF value for higher quality
   • Check if source supports hardware decoding
   • Verify hardware encoder supports desired codec
   • Try different tone mapping algorithm
4. NVIDIA Session Limit

   If you hit the 3-session limit on consumer GPUs:

   • Apply nvidia-patch
   • Use a Quadro/Tesla GPU (no limits)
   • Reduce concurrent transcoding sessions
5. Docker Hardware Access

   Ensure:

   • Devices are properly mounted (--device)
   • Correct group IDs are added (--group-add)
   • Runtime is specified for NVIDIA (--runtime nvidia)

Best Practices

Match Hardware to Workload

Use Intel QSV for efficiency, NVIDIA for maximum performance, and VA-API for compatibility.

Enable Hardware Decoding

Configure hardware decoding for codecs your GPU supports to reduce CPU usage further.

Monitor Temperatures

Ensure adequate cooling, especially during sustained transcoding sessions.

Test Before Deployment

Verify quality and compatibility with your media library before enabling for all users.

Codec Selection by Platform

Intel QSV

Best for:

• H.264/HEVC transcoding
• Low power consumption
• Integrated graphics systems
• Multiple concurrent streams (no session limits)

NVIDIA NVENC

Best for:

• Maximum transcoding speed
• High-quality HEVC encoding
• Modern RTX cards for AV1
• Systems with dedicated GPU

AMD AMF

Best for:

• AMD-based systems
• H.264/HEVC encoding
• RX 7000 series for AV1

VA-API

Best for:

• Linux environments
• Intel/AMD GPUs
• Open-source stack preference

Next Steps

Transcoding Settings

Fine-tune quality, codecs, and performance options

Network Configuration

Configure remote access and bandwidth limits