Quick Sync Transcoding by Intel Generation for Plex and Jellyfin

Hardware transcoding is the main reason to care which Intel generation is inside a used mini PC. When a client can't play a file directly, your media server has to decode it and re-encode it in a format and bitrate the client accepts. Doing that on the CPU cores is slow and power-hungry, especially for 4K HEVC. Intel's Quick Sync Video engine does it in fixed-function hardware on the integrated GPU, so even a small 1-liter PC can serve transcodes. What it can decode and encode depends entirely on the generation.

What each generation can do

4th gen (Haswell) and older: H.264 decode and encode, plus MPEG-2 and VC-1 decode. No hardware HEVC, so 4K HEVC sources fall back to the CPU. Not a good media-server buy.
6th gen (Skylake): adds HEVC 8-bit decode and encode. HEVC 10-bit decode is only hybrid (partly on GPU shaders) and VP9 decode is partial, so 10-bit HDR files are a weak spot.
7th gen (Kaby Lake) through 10th gen (Coffee Lake, Comet Lake): full fixed-function HEVC 10-bit decode and encode, plus VP9 8-bit and 10-bit decode. This is the baseline for modern 4K HDR libraries. 10th-gen Ice Lake laptop chips (Gen11 graphics) add VP9 encode.
11th gen (Tiger Lake, Rocket Lake), 12th gen (Alder Lake), 13th gen (Raptor Lake) and the N-series (N100, N150, N305): add AV1 decode, plus HEVC 12-bit decode. No AV1 encode.
Core Ultra Series 1 (Meteor Lake): adds AV1 encode. Jellyfin notes this generation's encoder runs only in low-power (LP) mode, which current Jellyfin builds support.

The media engine is the same within each band regardless of CPU tier. A Core i3-8100T and a Core i7-8700T share the same Quick Sync capabilities, and an N100 has the same codec list as a 12th-gen Core i7. Higher CPU tiers help only when the CPU has to do software work alongside the GPU.

HDR tone mapping

When a 4K HDR file has to be transcoded for an SDR screen, the server must also tone-map it, or colors come out washed out. Jellyfin supports hardware HDR and Dolby Vision to SDR tone mapping on all Intel GPUs that have HEVC 10-bit decoding, meaning 7th gen and newer, plus the N-series. It offers two methods: OpenCL, which supports Dolby Vision Profile 5 and more tuning options, and QSV VPP, which uses less power but is currently Linux-only. On N-series chips under Linux, Jellyfin notes that Intel's OpenCL runtime added N95 and N100 support in its 23.x releases, so use an up-to-date driver.

Plex's tone-mapping article lists hardware tone mapping with Intel Quick Sync on Kaby Lake or newer processors and Plex Media Server 1.29.2 or newer, and adds saturation, contrast and brightness controls for Tiger Lake or newer. On Ubuntu with Coffee Lake or newer chips, Plex requires Ubuntu 20.04 or later.

Plex: Plex Pass and Intel requirements

Hardware-accelerated streaming in Plex is a premium feature that requires an active Plex Pass subscription. Without it, Plex transcodes on the CPU.
Plex's Windows requirements start at 2nd-gen Intel Core (Sandy Bridge, 2011) for Quick Sync, but HEVC encoding needs a 7th-gen Core i3/i5/i7 or newer and Plex Media Server 1.41.3 or newer. Plex warns that HEVC encoding results in fewer concurrent streams than H.264.
On Linux, Plex lists 64-bit Ubuntu 16.04 or later or 64-bit Fedora 26 or later as supported; other distributions may work but aren't supported. Hardware transcoding in Docker works on a Quick Sync host but needs extra container configuration.
If the PC also has a discrete graphics card, Plex notes that Quick Sync may become unavailable while the card is in use. That matters for small form factor desktops with a GPU, less so for 1-liter mini PCs.

Jellyfin is free and open source, so hardware transcoding needs no subscription. On Windows it uses QSV only; on Linux it supports both QSV (preferred on mainstream Intel GPUs) and VA-API, and Jellyfin notes that Linux QSV requires Broadwell (5th gen) or newer.

AMD: Jellyfin works, Plex is 'as is'

AMD mini PCs can transcode, but the support picture differs by server. Jellyfin supports AMD through AMF on Windows and, on Linux, through VA-API using Mesa's open-source driver, which it prefers over AMF there. What the hardware can do depends on the video engine generation:

Ryzen 2000/3000 APUs and Ryzen Embedded V1000/R1000 (VCN 1.0): H.264 and HEVC decode and encode, VP9 decode, 10-bit HEVC decode. No 10-bit HEVC encode and no AV1.
Ryzen 4000 and 5000 (VCN 2.x): Jellyfin lists Ryzen 4000 APUs as the first with 10-bit HEVC encode. Still no AV1.
Ryzen 6000 and 7035 (Rembrandt, VCN 3.1): adds AV1 decode.
Ryzen 7040 and 8040 (Phoenix, Hawk Point, VCN 4.0): adds AV1 encode.
Pre-Ryzen A-series (Bristol Ridge): no 10-bit HEVC decode; Jellyfin lists 10-bit HEVC decode only from Polaris-era Radeon GPUs onward.

Jellyfin supports hardware tone mapping on AMD GPUs with HEVC 10-bit decoding, via OpenCL or Vulkan. It is also candid about quality: it says AMD's hardware encoders are "currently no match for Intel QSV and NVIDIA NVENC", though newer VCN versions have improved.

Plex is more limited. Its support article says its hardware transcoding has technical support for many dedicated AMD graphics cards but that Plex hasn't done official, full testing, that AMD support is provided "as is", and that it recommends Intel Quick Sync or an NVIDIA GPU instead. Its Linux requirements mention only Intel and NVIDIA. If Plex is your server, an Intel mini PC is the safer choice.

How many streams?

Honest answer: it depends on the source codec, resolution, bitrate, tone mapping and subtitles, and we haven't found trustworthy per-generation stream counts. Neither server imposes an artificial cap on Intel: Plex says Intel hardware encoding on Windows and Linux has no artificial limit on simultaneous videos, and Jellyfin says there is no concurrent encoding session limit on Intel iGPUs. Jellyfin describes mainstream iGPUs such as the UHD 620 and 630, Iris 640 and 655, and Gen 11 graphics as capable of multiple simultaneous 4K HDR HEVC 10-bit transcodes. The easiest way to raise capacity is to avoid transcoding: clients that direct-play your files need far less from the server.

Which generation to buy

Windows 11 plus Plex or Jellyfin: 8th gen or newer. It is the oldest Intel generation Microsoft supports for Windows 11 and has the full HEVC 10-bit and VP9 engine. Business tiny PCs such as the ThinkCentre M720q and EliteDesk 800 G4 Mini fit here. See Windows 11 CPU generations.
Linux-only Jellyfin or Plex: 7th gen works just as well for transcoding, since its media engine matches 8th gen.
AV1 sources: 11th gen or newer for decode, or an N-series box such as the Beelink EQ12. See N100 vs N150 vs N305.
AV1 encode: Core Ultra Series 1 or newer.
Avoid 4th gen and older for transcoding, and treat 6th gen as an H.264 and 8-bit HEVC machine.

Power and noise

Fixed-function transcoding is efficient, so the box's baseline power and fan matter more than transcoding load for a server that runs all day. Measured figures vary by model. CNX Software measured an N100-based Beelink EQ12 at 12.1 W idle in Windows and a maximum of 34.3 dBA, and a Ryzen 5 5600H Beelink SER5 at 5.3 W idle in Windows with its fan peaking at 38 dBA under full CPU stress. Check reviews of the exact model rather than assuming an N-series box idles lower than a Core or Ryzen one.

Checks before you rely on hardware transcoding

Make sure the integrated GPU is enabled in the BIOS; on desktops with a graphics card it may be disabled by default.
On Linux, confirm the iGPU device appears (usually /dev/dri/renderD128) and pass it into your Docker container or VM.
In Plex, confirm Plex Pass is active and enable hardware acceleration in the transcoder settings; in Jellyfin, pick QSV or VA-API and tick only the codecs your generation supports.
Play a 4K HDR test file to a client that forces a transcode, and check the dashboard shows hardware transcoding.

For full server requirements, see the media server use page.