The damn N5105 can't be decoded and the damn solution

[dan123 0]

Since the launch of the N5105 processor-based server from QNAP it has become clear that Emby has had many issues regarding the inability to hardware transcode.

I found several posts in the forum, the following in chronological order.

• https://emby.media/community/index.php?/topic/107087-hw-acceleration-not-working/
• https://emby.media/community/index.php?/topic/107440-why-ts-464c-can-only-use-software-decoding？/
• https://emby.media/community/index.php?/topic/109875-anyone-using-n5105-qnap-nas-with-emby/
• https://emby.media/community/index.php?/topic/110116-hw-acceleration-not-working-ts464c/

After when I purchased the same model, the same problem occurred. A friend of mine, which purchased a similar model found that the hard solution worked correctly under PLEX, so I have reason to believe that it was not a hardware defect, but just a software problem that caused it.

Interestingly, since April I have seen that luke has been aware of the problem and indicated that it would be fixed, but why has it not worked?

Oh well, maybe it's a tricky enough problem ......

But it actually has a solution so clean that even a monkey could fix everything!

That is, in the 'Transcode' option, turn off VAAPI and turn on QuickSync only!

You will then be surprised to find that everything starts working.

I don't know why luke didn't try to fix this, or offer any meaningful suggestions. But for all users of this processor, if you are experiencing this problem, yes, this is a monkey-level solution, no need to listen to any 'problem logging' or 'suggestions' from luke.

Edited July 5, 2022 by dan123

[Luke 37251]

Hi, thanks for sharing your findings. We'll take a look at it.

[softworkz 3349]

@dan123 There exists a bunch of issues with Intel GPU media acceleration, primarily regarding the newer CPU generations like Jasper Lake (yours) Elkhart Lake, Alder Lake, etc.

It's not a single problem and there doesn't exist a single solution. Your case "VAAPI defunct but QSV working" is not even a common one, i.e. nothing like that we would have seen it frequently. It also very much depends on the distribution, not just on the drivers we're shipping, and we even had a case with TigerLake where OS, and software were identical, same for the CPU. The only difference was that one device was a Laptop and the other was a kind of NUC box.

From those user reports

As we got close contact to Intel regarding media hwa, and talked to them about those issues, it became clear that there's no point in following up those cases with "try this"/"try that" because it would be just fishing in the dark. Intel confirmed those issues and most of them are supposed to be solved when using the latest versions of libva, iHD driver, i965 kernel mode driver and GuC/HuC firmware.

The problem was, that during the past few months, we were stabilizing for the next stable release and during that final stage we couldn't update all those Intel bits in our packages.

Though, we're free to do now, and the next beta (or maybe one after) will have all these things update and then we'll be good to look into all those (remaining) issues.

Thanks for your understanding!

[dan123 0]

On 7/8/2022 at 2:38 AM, softworkz said:

@dan123 There exists a bunch of issues with Intel GPU media acceleration, primarily regarding the newer CPU generations like Jasper Lake (yours) Elkhart Lake, Alder Lake, etc.

It's not a single problem and there doesn't exist a single solution. Your case "VAAPI defunct but QSV working" is not even a common one, i.e. nothing like that we would have seen it frequently. It also very much depends on the distribution, not just on the drivers we're shipping, and we even had a case with TigerLake where OS, and software were identical, same for the CPU. The only difference was that one device was a Laptop and the other was a kind of NUC box.

From those user reports

As we got close contact to Intel regarding media hwa, and talked to them about those issues, it became clear that there's no point in following up those cases with "try this"/"try that" because it would be just fishing in the dark. Intel confirmed those issues and most of them are supposed to be solved when using the latest versions of libva, iHD driver, i965 kernel mode driver and GuC/HuC firmware.

The problem was, that during the past few months, we were stabilizing for the next stable release and during that final stage we couldn't update all those Intel bits in our packages.

Though, we're free to do now, and the next beta (or maybe one after) will have all these things update and then we'll be good to look into all those (remaining) issues.

Thanks for your understanding!

Thank you for your explanation and I hope it will be resolved.

The performance on Jasper Lake is really weird and unstable.

Edited July 12, 2022 by dan123

[dtuando 1]

I updated my Mini GMKtek PC from the J4125 to the N5105 in hopes of better quicksync. But this is the end result. Is this fix coming in the near future? or is it currently out in a beta?

[Luke 37251]

25 minutes ago, dtuando said:

I updated my Mini GMKtek PC from the J4125 to the N5105 in hopes of better quicksync. But this is the end result. Is this fix coming in the near future? or is it currently out in a beta?

We are looking into it. Thanks.

[dtuando 1]

Thanks, I tried everything I could find. I disabled the encoder and the decoder started to work.....

with further research running as root

echo "options i915 enable_guc=2" > /etc/modprobe.d/i915.conf

and then running update-grub folowed by a reboot solved my issues and all is well now.

[Luke 37251]

On 7/23/2022 at 4:18 PM, dtuando said:

Thanks, I tried everything I could find. I disabled the encoder and the decoder started to work.....

with further research running as root

echo "options i915 enable_guc=2" > /etc/modprobe.d/i915.conf

and then running update-grub folowed by a reboot solved my issues and all is well now.

We know about these issues and we are currently discussing ways to detect those situations and inform users without automatically making modifications to the kernel parameters.

[Micro8Me 1]

deleted

Edited November 23, 2023 by Micro8Me