And bringing this full circle, yes, AMD’s encoders do not have as high as quality as NVIDIA’s or Intel’s latest generation encoder but let’s not also forget that Plex’s transcoder is a two-sided coin with hardware decoding as well. Also, as I had noted earlier, their hardware encoder does not exhibit the pixel distortion issue with either HEVC or h.264 in Handbrake which is also built directly from ffmpeg. To the point, AMD’s decoder is totally a different story from their encoder (see here: Does anyone review video decoding quality any more? | AnandTech Forums: Technology, Hardware, Software, and Deals). Their decoder has long been considered superior to certainly Intel’s, and slightly better than NVIDIA’s.
For me at least, AMD’s decoder comes very, very close to MadVR which is considered the gold standard for decoders by HTPC enthusiasts. That would mean of course that there is less quality loss when you use hardware decoding on AMD hardware. So yes, you may get better raw encoding quality with Intel or NVIDIA but that is tested generally without hardware decoding active, in the context of game streaming where you are capturing raw video from your screen. In contrast, in our use case, Plex uses both hardware encoding and decoding in the transcoding process and you want the least quality loss when decoding in real time in hardware before the source frames are passed to the hardware encoder.
Meanwhile, three years after the introduction of AMD’s first Ryzen APUs, Plex still does not have proper support for AMD Video Core Next (VCN) hardware encoding. So much for the QNAP’s advertised 4K UHD transcoding…
It’s broken. And it has been that way since Navi and Ryzen 4000 series APUs and Plex support just keeps giving excuses ad infinitum. It comes down to them not adding proper support for VCN 2.0 and above, which was introduced with the RX 5000 series Navi GPUs in 2019. Two years later, no support. You can upvote for support here, though!
Why are you complaining about something they don’t support? I don’t understand this at all. Yes, it is frustrating, and it would be nice. But it is something explicitly not supported. If anything, its your fault for choosing an AMD GPU. Pick up an NVIDA card or use the integrated graphics on an Intel CPU. It isn’t an issue because it doesn’t work. Plex has very clearly designated AMD GPU’s as being as-is support, and receives no official support. Hence, this is not an issue. You are outside the scope of support.
This is a simple case of a user refusing to RTFM. Use the supported hardware, and you have no issues.
Apparently the issue is even worse with the 6000 series, transcoding with my old 1060 was working just fine and now that i got a rx 6700xt nothing is watchable with hw encoding enabled
Yep. Anything VCN 3.0 and above is severely impacted. Yet nothing but crickets. The most you will get is a lengthy non-response or excuse rather than any commitment.
Just chiming in after a long read across the forums regarding Plex hw acceleration with AMD GPUs. So far I can clearly see, that Plex is lacking responses and as well explanations or actually adressing the unsopported GPUs.
However, I swapped an NVIDIA RTX 3070 for an AMD RX 6800 XT in my Plex machine and I don’t see any issue so far? The hw transcoding works just fine and I tested this on multiple devices - remotely and locally and I cannot see any issues whatsoever. It may be highly subejctive, but the video quality and performance even seems to be better than with the RTX.
I gotta say that I’m on the beta OS of Windows 11 so I’m not sure what my experience and sightings are worth. But - personally - I don’t have any issue with HW acceleration and the AMD GPU at all. Maybe you can - calmly - explain what’s the status quo at the moment @Hifihedgehog and maybe they already changed something (?). Already, lol…
For those unaware and have not dived into the details described abundantly in the Plex forum, the issue described in detail above (the post I am replying to in particular is a regurgitation of numerous submitted reports) among the many examples given by users here and other threads still persists.
