Jellyfin Transcoding and Encoding, on AMD

Apps and Virtualization
1

Yes I know, another one, but not quite.

I’ve been looking at a lot of Mini-PC + Occulink/Thunderbolt4 NVME DAS builds that I believe are going to work well for a TrueNAS Scale minirack build. All the AMD ones kind of have the specs I want, just a nice flat out 16 core, 32 thread package with extra room for NVME drives galore, but I’d really like to avoid using the PCIE slot for a ARC GPU. I host a lot of media, so out of this build, that will be important, but I also host a few game servers.

Most if not every single thread I’ve reviewed mostly just ends with ‘go with intel quicksync just because, that is the way, supportability’. Yet those threads never really go into details, I haven’t found very recent proof on, why.

I understand quicksync is a dedicated hardware core on the die, so that’s different from utilizing a few cores on the iGPU. I understand that even Jellyfin KB’s have recommendations, yet state supportability.

I haven’t really been able to find anything on the Ryzen 9000 series feedback from users, or even the Ryzen AI 370 which is RDNA 3.5 from users, then we have Strix Halo coming out soon.

What I have found is benchmarks showing at least the AI 370 being extremely good at these things. AV1, HEVC @ 4k all seemed fine, if not better than even the Meteor-Lake equivalent.

Now I understand, well, speed is different from quality. Is it so much that the issue is Jellyfin supporting the AMD CPU’s for the quality? I.E. tone mapping? or vice versa, that AMD iGPU is not good enough or well, is it just …we don’t have much real world data to even comment yet? I feel like we have a FEW real threads out there that do have a bit of real world data but it’s like 4000g/5000g series, not even 8000g series.

Do I need to be a guinea pig? Haha

I’m just wondering if anyone out there has experience with the very new AMD CPU’s and utilizing it as an all around NAS/APP box + Media server and possibly multiple transcode sessions.

3

Personally I think it’s just brand name recognition, the fact that QuickSync has been consistently ahead of AMD’s offerings in performance, and people’s reluctance to change, so we end up not getting much updated info.

I built a TrueNAS box fairly recently using a Ryzen Pro 4650G and it supports transcoding just fine in Jellyfin, AFAICT. I haven’t done a back-to-back test with an Intel chip to see if there’s a quality difference, but the Ryzen does support it, even at 4k. I wouldn’t recommend trying to transcode 8 4K streams simultaneously, or even 4, but for a single stream it works, and that’s all I need, because there’s 2 of us here with one TV and we watch it together.

So if my relatively old 4650G CPU can handle it, I’m sure the newer ones can handle it even better.

One note: my 5+ year old CPU doesn’t support AV1 transcoding, but newer ones might. I’m not worried about this currently, but in the future if it’s an issue I’ll look into upgrading.

4

Yeah I think RDNA 3 does? Still not QUITE as good, but RDNA 3.5 on the AI 370’s seems to be, very very good, especially at 4k.

Though that’s short lived because Strix Halo will include RDNA 4 on them. So we’ll see how those do soon.

5

Quoting from AMD GPU | Jellyfin - emphasis added by me

Due to the lack of B-frame support, the encoding quality of the AMD H.264 hardware encoder has been unsatisfactory. Although RX 6000/VCN3.0 brings back the B-frame support, the quality improvement is not great.

The AMD HEVC encoder is far better than the AMD H.264 encoder, and the new AMD AV1 encoding support on RX 7000/VCN4.0 seems to be the savior of AMD encoding quality. Nonetheless they are currently no match for Intel QSV and NVIDIA NVENC. VCN4.0 additionally improved the encoding speed drastically.

1 Like
6

Interesting, very large judgements that provide only surface level statements.

Really nice that Jellyfin KB’s are telling us what media engine features were and weren’t there, but now are, and providing in-depth information such as “far better” and “no match”. Amazing.

7

Yeah, the more I dig, especially within Puget forums. AMD as a platform just seem better, Quicksync, over the last 2 gens, just isn’t giving THAT much more of a advantage for encoding/transcoding.