Evaluation of HEVC decoders (SW, Hybrid and HW)

[xooyoozoo]

Can you name a few ?

Are there any free analysers capable of doing that ?

Intel, Solveig, Parabola all have trials.

I don't think there are any free analyzers for looking at parameter sets. There was a GPL'ed tool a while back that dumped AVC parameter sets, but I don't think the creators have made a HEVC version.

[NikosD]

Thanks for the reply.

I tried all 3 demos of Intel, Solveig, Parabola on Gravity HEVC trailer and none of them worked.

They all have such restrictions in demo apps, that didn't allow them to provide the encoding parameters (or even open the video file in case of Intel)

[P.J]

Ok, Samsung NX1 clips didn't run with 750Ti using PotPlayer.
I also tried 10-bit Samsung UHD clip with 750Ti but it looks like this: :(

http://7.t.imgbox.com/58q8JLof.jpg (http://imgbox.com/58q8JLof)


No, HD 4600 can use 1GB RAM, that's the max setting of BIOS and if you see my first post with the benchmark details, I have specific numbers for GPU RAM utilisation during benchmarking.

I did the test again monitoring real main memory RAM of my system and GPU RAM and during playing back of one of the samsung clips, GPU RAM went as high as 1.5GB

How do you check the GPU RAM?

[ndkamal]

I have an ultrabook with Geforce 840M, and I can't play HEVC files with this chipset, when I try, I have a strange picture like this.

http://9.t.imgbox.com/PRf8wg3f.jpg (http://imgbox.com/PRf8wg3f)

And on the decoder device section, I have not all the formats supported by the Geforce 840M :

http://7.t.imgbox.com/kaqmbOTj.jpg (http://imgbox.com/kaqmbOTj)

Apparenty Optimus has problem with Maxwell chipset. :-(.

[mzso]

Hello!
Is there an OpenCL decoder that does anything useful with an NVIDIA card (gts450)?
I tried the Lentoid HEVC Decoder (It's the same as "Strongene", right? ). The only difference was that it used a shade more CPU.

[wanezhiling]

Strongene PC HEVC/H.265 Decoder Plugin For Intel Media SDK (new) (http://www.strongene.com/en/downloads/downloadCenter.jsp)

[JohnLai]

Great news. Credit to Anandtech article http://www.anandtech.com/show/8811/nvidia-tegra-x1-preview

" For video encode and decode, we see support for 4K60 H.265 and VP9, with support for 10 bit color on decode on H.265"

10 bit as well? I wonder if upcoming GTX960 supports it......

[mzso]

Strongene PC HEVC/H.265 Decoder Plugin For Intel Media SDK (new) (http://www.strongene.com/en/downloads/downloadCenter.jsp)

If you addressed this to me I don't know what you mean.

It's an SDK which is of no use to me.

[nevcairiel]

10 bit as well? I wonder if upcoming GTX960 supports it......

Doubtful, its the same architecture as the 970/980, which don't even have full 8-bit support yet, and it would be nothing but a miracle if they update the hardware decoder in the 960 just 2 month after the earlier launch.

However, the existence of a full HW 10-bit decoder will at least mean it'll be available in the next gen.

[nevcairiel]

Your quote is still speculation, not a fact that the 960 will have it though.

[NikosD]

How do you check the GPU RAM?


I mainly use the gadget of HWiNFO, HWiNFO monitor.
It has same figures as GPU-Z


Hello!
Is there an OpenCL decoder that does anything useful with an NVIDIA card (gts450)?
I tried the Lentoid HEVC Decoder (It's the same as "Strongene", right? ). The only difference was that it used a shade more CPU.

Strongene and Lentoid are the same thing, but Strongene/Lentoid has two kind of decoders.

CPU decoder only and Hybrid CPU/ GPU (OpenCL) decoder.

Also, you could try latest PowerDVD, I think it has an option for OpenCL H.265 GPU assisted video decoding.


Strongene PC HEVC/H.265 Decoder Plugin For Intel Media SDK (new) (http://www.strongene.com/en/downloads/downloadCenter.jsp)


Thanks.

It seems that this is not a stand-alone decoder, but a plugin for developer's use and integration in Intel's MSDK.

You could post it to Eric or Nevcairiel to tell us their opinions about how that could be useful.

[NikosD]

Tegra X1 with its 4K60 fps HW decoder for VP8/VP9 and H.264/H.265 plus 10bit H.265 HW support, is a very interesting product.

I think it's possible for 960 card to have such decoder inside.

We have to wait and see later this month (Jan 22 - I think) for that product arrival.

[mzso]

Strongene and Lentoid are the same thing, but Strongene/Lentoid has two kind of decoders.

CPU decoder only and Hybrid CPU/ GPU (OpenCL) decoder.

Also, you could try latest PowerDVD, I think it has an option for OpenCL H.265 GPU assisted video decoding.


I see. I only noticed in the add filter dialog Lentoid HEVC Decoder and Lentoid HEVC Source (for raw HEVC streams?). So does it have some totally different name? Because only these two start with Lentoid.

Powerdvd's decoder is not available to other apps?

[P.J]

Also, you could try latest PowerDVD, I think it has an option for OpenCL H.265 GPU assisted video decoding.

I don't think...

[ibius]

Cyberlink Video Decoder from PowerDVD v14 does support OpenCL H.265 hw-assisted video decoding and it is available to other DS players, it just has a 'do not use' merit.
You also have to set the decoder to use HAM-mode, don't remember if DXVA-mode would also use OpenCL hybrid decoding. When playing a HEVC video go to the Filters-tab, press and hold 'Ctrl'+'Left Mouse' on the 'Cyberlink Video Decoder (PDVD Generic)', it will open advanced/detailed settings view, the Profile column will tell you if the decoder is using OpenCL or not.
It wont use OpenCL on my GTX 550Ti, but does work on my brothers AMD Radeon HD 5700, too bad, because mine does a lot better on all my samples and tests.
But even in SW-mode I think it does even better than Lentoid and it also supports 10-bit, which Lentoid does not, and both are a LOT faster than LAVVideo x86, on my brothers x64-machine also LAVVideo x64 was slower than Cyberlink and Lentoid, of course it has an old AMD Phenom X4 2.6 Ghz with up to SSE4A instructions.

P.S. last LAV build I tested was 0.63.0.12, don't know how much faster 0.63.0.41 is, I did read it should have better error handling now.

[P.J]

CLCvd.ax? No luck with 750Ti too :(

http://i.imgur.com/h9FyKY0.png

[mzso]

Cyberlink Video Decoder from PowerDVD v14 does support OpenCL H.265 hw-assisted video decoding and it is available to other DS players, it just has a 'do not use' merit.
You also have to set the decoder to use HAM-mode, don't remember if DXVA-mode would also use OpenCL hybrid decoding. When playing a HEVC video go to the Filters-tab, press and hold 'Ctrl'+'Left Mouse' on the 'Cyberlink Video Decoder (PDVD Generic)', it will open advanced/detailed settings view, the Profile column will tell you if the decoder is using OpenCL or not.
It wont use OpenCL on my GTX 550Ti, but does work on my brothers AMD Radeon HD 5700, too bad, because mine does a lot better on all my samples and tests.
But even in SW-mode I think it does even better than Lentoid and it also supports 10-bit, which Lentoid does not, and both are a LOT faster than LAVVideo x86, on my brothers x64-machine also LAVVideo x64 was slower than Cyberlink and Lentoid, of course it has an old AMD Phenom X4 2.6 Ghz with up to SSE4A instructions.

P.S. last LAV build I tested was 0.63.0.12, don't know how much faster 0.63.0.41 is, I did read it should have better error handling now.

I finally figured out that I was using the software decoder. Here the OpenCL decoder just causes an instant crash.

Hmmm. I don't see the cyberlink decoder appearing anywhere in the filters window...

[ibius]

CLCvd.ax? No luck with 750Ti too :(

Yes CLCvd.ax, so it looks like both Cyberlink and Strongene (says so on their site) implemented HEVC OpenCL partial hw-acceleration only for AMD, yet.?
I don't know if its because of NVIDIA's lagging behind OpenCL performance or what, but AMD systems have newer OpenCL version as well.

I finally figured out that I was using the software decoder. Here the OpenCL decoder just causes an instant crash.

Hmmm. I don't see the cyberlink decoder appearing anywhere in the filters window...

Do you mean 'Strongene/Lentoid OpenCL HEVC Decoder'? If so same here, on my brothers AMD machine (HD 5700) Strongene's OpenCL decoder is a mess, out of about two dozen samples it did play 1, I think, without crashing.

Try setting higher merit for 'Cyberlink Video Decoder (PDVD Generic)' globally or just within your player, at least 00600000 or 00800000, and block those that interfere.

[nevcairiel]

Yes CLCvd.ax, so it looks like both Cyberlink and Strongene (says so on their site) implemented HEVC OpenCL partial hw-acceleration only for AMD, yet.?
I don't know if its because of NVIDIA's lagging behind OpenCL performance or what, but AMD systems have newer OpenCL version as well.

NVIDIA and Intel support partial acceleration through DXVA2, which is a far more favorable solution than a proprietary OpenCL decoder, because once they support full acceleration as well, software will not need to change at all.

[NikosD]

Nev, what about that HEVC plugin for Intel's MSDK by Strongene that Wanezhiling posted above ?

Could that be useful and easy to integrate it in LAV filters like Quick sync decoder ?

[nevcairiel]

Its a software decoding module. No interest.

[Yups]

HEVC Hybrid decoding works quite on my Haswell GT2. There are some limitations though. If the GPU load is very high it stutters. This video for example: bbb-3840x2160-cfg02.mkv
http://www.libde265.org/downloads-videos/

It stutters, even though CPU load is relatively low roughly 20%. GPU load is over 90% on this video. 3840x2160 60 fps is too much.

[Yups]

I saw the GPU memory usage to 1.5 GB which is unexplainable because from BIOS I can assign up to 1 GB GPU memory!




It doesn't matter how much of your system RAM is dedicated to your iGPU, Intel GPUs can use up to 1.6-1.7GB of your system RAM.

[NikosD]

This is a new updated post in order to include the CyberLink's PowerDVD v14 H.265 decoder, which is both a 32bit CPU and Hybrid (OpenCL) decoder.

Also, I added a 10bit HEVC clip and Microsoft's MFT Windows 10 HEVC decoder in both x86 and x64 versions.

I used latest versions of DXVA Checker x86/x64 v3.3.1 in benchmark playback performance mode at 1280x720 scaling resolution, LAV x86/x64 0.63.41, and Win 10 beta build 9879

Tests include five systems and the first one has a modern CPU Core i7-4790 with Win 8.1 x64 and Intel HD 4600 iGPU@1.5GHz with v.3960 drivers.
I used Threads=16 for Core i7-4790 using LAV x86/x64 CPU decoder.


Core i7-4790(AVX2) - iGPU HD4600


1.ProRes-1080p@30fps-2Mbps


LAV x64 CPU 177/247/264 CPU 78% Power:55W

LAV x86 CPU 125/151/198 CPU 91% Power:58W

LAV x64 DXVA 100/151/176 CPU 15% Power:50W

LAV x86 DXVA 99/149/168 CPU 17% Power:51W

PowerDVD OpenCL 95/135/152 CPU 23% Power:56W

PowerDVD CPU 79/132/150 CPU 32% Power:36W

Lentoid OpenCL 90/111/120 CPU 20% Power:36W

Lentoid CPU 86/108/110 CPU 16% Power:35W



2.Gravity - 2048x858@24fps-5Mbps-10bit


LAV x64 CPU 105/148/225 CPU 65% Power:51W

LAV x86 CPU 169/106/227 CPU 68% Power:51W

PowerDVD CPU 66/101/197 CPU 55% Power:42W



3.Beauty-2160p@30fps-12.3Mbps


LAV x64 CPU 51/64/66 CPU 80% Power:50W

Lentoid OpenCL 25/43/43 CPU 42% Power:55W

LAV x86 CPU 24/41/46 CPU 93% Power:39W

LAV x64 DXVA 35/37/40 CPU 14% Power:45W

LAV x86 DXVA 34/36/38 CPU 15% Power:45W

Lentoid CPU 10/33/35 CPU 24% Power:40W

PowerDVD CPU 23/29/31 CPU 32% Power:33W


Second system is a Sandybridge Core i5-2400 CPU under Win 8.1 x64 using an ATI Radeon 5750 (14.12 drivers) card with OpenCL support.
I used LAV x86 with Threads=Auto (~1% difference than threads=16), but I used LAV x64 with Threads=16 (more than 20% faster than Auto)



Core i5-2400(AVX) - ATI HD5750


1.ProRes-1080p@30fps-2Mbps


LAV x64 CPU 226/269/318 CPU 91%

Lentoid CPU 128/154/202 CPU 55%

PowerDVD CPU 80/149/194 CPU 52%

LAV x86 CPU 78/108/165 CPU 91%

Lentoid OpenCL 87/107/127 CPU 49% GPU 89%

PowerDVD OpenCL 64/105/141 CPU 23% GPU 83%



2.Gravity - 2048x858@24fps-5Mbps-10bit


LAV x64 CPU 79/138/290 CPU 91%

PowerDVD CPU 56/96/285 CPU 66%

LAV x86 CPU 51/86/264 CPU 89%



3.Beauty-2160p@30fps-12.3Mbps


LAV x64 CPU 43/62/75 CPU 88%

Lentoid CPU 23/38/40 CPU 61%

PowerDVD CPU 25/33/36 CPU 48%

LAV x86 CPU 20/29/31 CPU 91%

Lentoid OpenCL 8/27/30 CPU 47% GPU 80%

PowerDVD OpenCL 18/26/33 CPU 40% GPU 52%



Third system is an old but quad core CPU Core 2 Quad@2.26GHz (266MHzx8.5) with no OpenCL option under Windows 8.1 x64 system using a Fermi Nvidia card - GT610 (347.09 drivers)
I tested a 4K clip, although it's a rather slow CPU not really capable of even 4K@24 fps with any of the decoders.
I used Threads=12 for Core 2 Quad using LAV x86/x64 CPU decoder.

The pure decode benchmarks results, mainly due to slow Nvidia GT610 card and PCI-E v1.0 x4 motherboard, are ~65% faster than the below playback benchmark mode results.
So, for quad cores with better m/b and GPU than mine, the results will be a lot faster in playback mode than these.


C2Q@2.26GHz(SSE4.1) - GT610



1.ProRes-1080p@30fps-2Mbps


LAV x64 46/53/54 CPU 59%

Lentoid 26/45/46 CPU 49%

LAV x86 35/43/52 CPU 90%

PowerDVD 34/38/38 CPU 50%



2.Gravity - 2048x858@24fps-5Mbps-10bit


LAV x64 25/40/58 CPU 85%

LAV x86 23/35/58 CPU 87%

PowerDVD 22/31/38 CPU 69%



3.Beauty-2160p@30fps-12.3Mbps


LAV x64 4/13/13 CPU 62%

LAV x86 CPU 2/12/13 CPU 91%

Lentoid 1/11/12 CPU 49%

PowerDVD 3/9/9 CPU 53%


Fourth system is an old CPU Core 2 Duo - E7300 CPU@2.83GHz (283MHz x 10) with no OpenCL option under Windows 8.1 x64 system using a Fermi Nvidia card - GT440 (347.09 drivers)
The pure decode benchmarks results, mainly due to PCI-E v1.0 x4 motherboard, are a lot faster than the below playback benchmark mode results.
I didn't test any 4K clips, because it's a slow CPU not really capable of even 4K@24 fps with any of the decoders.



C2D@2.83GHz(SSE4.1) - GT440



1.ProRes-1080p@30fps-2Mbps


LAV x64 35/43/50 CPU 86%

Lentoid 29/41/48 CPU 80%

PowerDVD 28/36/39 CPU 85%

LAV x86 20/27/40 CPU CPU 92%



2.Gravity - 2048x858@24fps- Mbps-10bit


LAV x64 20/30/52 CPU 92%

PowerDVD 17/26/41 CPU 83%

LAV x86 14/24/51 CPU CPU 92%


Fifth system is an old CPU Core 2 Duo - T7600 CPU@2.33GHz with no OpenCL option under Windows 10 build 9879 OS.
I didn't test any 4K clips, because it's a slow CPU not really capable of even 4K@10 fps with any of the decoders.


C2D@2.33GHz(SSSE3)-T7600



1.ProRes-1080p@30fps-2Mbps


PowerDVD 22/34/50 CPU 60%

Lentoid 24/31/40 CPU 54%

LAV x64 21/30/54 CPU 89%

LAV x86 19/27/55 CPU CPU 89%



2.Gravity - 2048x858@24fps-5Mbps-10bit


MS MFT x64 21/34/57 CPU 88%

MS MFT x86 19/32/57 CPU 87%

PowerDVD 13/24/56 CPU 70%

LAV x64 16/21/22 CPU 67%

LAV x86 13/20/23 CPU CPU 74%



Comments

There are no Hybrid 10bit HEVC decoders yet, so only CPU can decode 10bit clips.

Intel's Haswell iGPUs (don't know about Ivy) are the only GPUs capable of using both types of hybrid decoders (DXVA and OpenCL)
AMD uses OpenCL only (no DXVA) and Nvidia uses DXVA only (no OpenCL) and only on newer cards.

Lentoid (Strongene) seems to be the fastest x86 decoder, but not with HT CPUs. It looks like a problem with multi-threading. Also is the only decoder not capable of 10bit decoding.

MS MFT Windows 10 decoder can be used only with MP4 container and looks like by far fastest than anything, at least using older SSEx CPUs

PowerDVD v14 decoder can't use hybrid OpenCL decoder for clips above 1080p, at least on my Intel system, but with Radeon 5750 worked fine at that resolution.

LAV x86 has almost always the best CPU utilization (most of the times near 90% on any CPU), but due to different optimizations of LAV x64, the latter is always faster.

On a new modern CPU like Core i7-4790, LAV x64/x86 seems faster than anything else even in both CPU and hybrid decoders. Especially LAV x64 CPU is by far faster than anything (haven't tried MS MFT decoder).

Lentoid OpenCL decoder is very fast on 4K clips.

PowerDVD v14 is the slowest decoder of all on 4K clips, but faster than Lentoid on 1080p.

On the other hand, for an old CPU like Core 2 Duo with SSSE3, MS MFT is clearly ahead of all other decoders regarding CPU performance.
PowerDVD is faster than both LAV and Lentoid.

LAV in both versions is the slowest decoder and can't keep up in realtime decoding for both clips using x86 version on a Core 2 Duo processor. LAV x64 can't keep up for Gravity, too.

The Core i5-2400 system manages to be faster than Core i7-4790 (!) on the easier to decode 1080p clip and very close to 4K clip decoding, due to the faster discrete card (Radeon 5750) than the iGPU of Core i7.

Once again the importance of a fast card is obvious when using CPU decoding real-world benchmark tests, using a renderer in playback performance mode and not pure decoding performance with a null renderer.

Using a 45nm Core 2 Duo with SSE4.1, LAV x64 is once again faster and LAV x86 is the slowest.

For quad core SSE4.1 CPUs, PowerDVD doesn't scale good, so it's once again the slowest of all decoders, where LAV x64 is once more faster, but not far from LAV x86 and Lentoid.

[NikosD]

Its a software decoding module. No interest.

It doesn't matter how much of your system RAM is dedicated to your iGPU, Intel GPUs can use up to 1.6-1.7GB of your system RAM.

Thanks.

[mzso]

Do you mean 'Strongene/Lentoid OpenCL HEVC Decoder'? If so same here, on my brothers AMD machine (HD 5700) Strongene's OpenCL decoder is a mess, out of about two dozen samples it did play 1, I think, without crashing.

Try setting higher merit for 'Cyberlink Video Decoder (PDVD Generic)' globally or just within your player, at least 00600000 or 00800000, and block those that interfere.

Even with AMD? Huh...

It doesn't even appear in the available codecs list neither in Potplayer nor in MPC-HC. (checked with x64 mpc-hc too) I don't know why. I installed pdvd. Do you know which is the codec's file and what are it's dependencies? Maybe I could add it manually....

[ibius]

Even with AMD? Huh...

It doesn't even appear in the available codecs list neither in Potplayer nor in MPC-HC. (checked with x64 mpc-hc too) I don't know why. I installed pdvd. Do you know which is the codec's file and what are it's dependencies? Maybe I could add it manually....

Well at least with older AMD HD5700, but according to NikosD's test results on the first page Strongene's OpenCL runs on iGPU HD4600 as well, although not much better, it crashed on 3 samples out of 6.
Nev says Intel supports partial DXVA2 decoding like NVIDIA, but both Strongene's and Cyberlink's OpenCL implementation works on AMD and Intel but not even on 750Ti, so I guess NVIDIA's OpenCL is for sh*t.

No need to check MPC-HC x64, Cyberlink Video Decoder is 32-bit only like PowerDVD itself.
If you don't even see the filter, find CLCvd.ax and register it manually, with PowerDVD installed you should have the decoder and it's dependencies in PowerDVD\Common\VideoFilter folder.

[mzso]

Well at least with older AMD HD5700, but according to NikosD's test results on the first page Strongene's OpenCL runs on iGPU HD4600 as well, although not much better, it crashed on 3 samples out of 6.
Nev says Intel supports partial DXVA2 decoding like NVIDIA, but both Strongene's and Cyberlink's OpenCL implementation works on AMD and Intel but not even on 750Ti, so I guess NVIDIA's OpenCL is for sh*t.

No need to check MPC-HC x64, Cyberlink Video Decoder is 32-bit only like PowerDVD itself.
If you don't even see the filter, find CLCvd.ax and register it manually, with PowerDVD installed you should have the decoder and it's dependencies in PowerDVD\Common\VideoFilter folder.

Cool. That helped. The HAM setting should be the one using OpenCl, right? Not sure if it's doing it though. It's maybe a shade faster than lav based on CPU usage, hard to tell. And I don't see any GPU usage increase.

[P.J]

so I guess NVIDIA's OpenCL is for sh*t.

Well explained :p :thanks:

[NikosD]

Is there someone who knows what's going on with DivX 10 HEVC decoder ?

Latest version 10.2.4 has removed Codec Pack and after installation, it downloads the necessary HEVC codec on demand.
Furthermore, doesn't expose this DS filter (?) to 3rd party apps.

After DivX 10.2.4 installation, it blocked every x86 codec installed (!) and couldn't load anything in DXVA Checker.

If someone could tell me how could I use DivX 10 HEVC codec, I would definitely test it :)

[Yups]

Interesting for Haswell and Broadwell.

- Improved video playback through partial hardware acceleration support for the VP9 video format.
- GPU accelerated decode of HEVC video file format including both 8-bit and 10-bit support. This will provide improved video playback capabilities on the platforms.
https://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=24596

[NikosD]

I tried that driver and doesn't look to have a DXVA decoder device for VP9.
Unless it's not named as VP9 or maybe DXVA Checker doesn't recognize it.
http://s15.postimg.org/71k4hap0b/DXVA_Checker.png

I also tried 10bit HEVC clips with latest LAV Video 0.63.41 HEVC DXVA decoder and I only get a black screen with no decoding.

[Yups]

What was your previous driver?

[NikosD]

v. 3960.

Why ?

Are you looking for comparisons between drivers ?

[Yups]

Because I noticed some differences from my previous driver (4029).

This video: http://demo-uhd3d.com/files/uhd4k/Demo_Samsung_2014_-_Lovely_Swiss.zip

With driver build 4029 GPU usage was low and CPU usage 80-100%. With driver 4080 I get 40-50% CPU usage and 90-100% GPU usage in Potplayer (colour issues unfortunately), in MPC Home Cinema there is only a black screen.

This video: bbb-3840x2160-cfg02.mkv

GPU usage is higher than before, 99-100% usage now. Previous usage was ~80%.

[wanezhiling]

driver 10.18.14.4080

Just like a slide show...
:o:o
http://i3.tietuku.com/21a83a47c35363e0.png

[JohnLai]

10 bit HEVC dxva decoding by intel? I not sure if intel Haswell HD IGPU has the hardware to decode it in first place.

This one probably refers to upcoming Broadwell.

^.^ If this 10bit dxva for haswell igpu really work, somebody please inform NVIDIA that their competitor already has 10bit dxva working. AMD too.

[nevcairiel]

From what I've seen its still kind of pointless, as it only supports 8-bit output, so while it may be able to decode it partially, it then probably rounds the 10-bit output to 8-bit, which degrades the quality.
I need to check in more detail if its actually possible to get it to output 10-bit.

[NikosD]

10 bit HEVC dxva decoding by intel? I not sure if intel Haswell HD IGPU has the hardware to decode it in first place.

This one probably refers to upcoming Broadwell.

^.^ If this 10bit dxva for haswell igpu really work, somebody please inform NVIDIA that their competitor already has 10bit dxva working. AMD too.

We are not talking about fixed-function HW decoder, like QuickSync, but for partial GPU assisted (via shaders) HW decoding.

So, it's possible to decode 10bit HEVC using a hybrid decoder.

From what I've seen its still kind of pointless, as it only supports 8-bit output, so while it may be able to decode it partially, it then probably rounds the 10-bit output to 8-bit, which degrades the quality.
I need to check in more detail if its actually possible to get it to output 10-bit.

But why the black image and no decoding without fall-back to SW using LAV Video ?

If all modifications are inside drivers, I think it could or should work instantly, without modifications of LAV video.

Or at least fallback to SW mode.

[nevcairiel]

They probably broke something else.

[NikosD]

With driver build 4029 GPU usage was low and CPU usage 80-100%. With driver 4080 I get 40-50% CPU usage and 90-100% GPU usage in Potplayer (colour issues unfortunately), in MPC Home Cinema there is only a black screen.

GPU usage is higher than before, 99-100% usage now. Previous usage was ~80%.

They probably broke something else.

PotPlayer seems to use DXVA for 10bit clips, but with artifacts and colour issues.
But it uses DXVA.

Also, PotPlayer plays fine in DXVA those Samsung NX1 clips posted above http://www.imaging-resource.com/news/2014/11/19/samsung-nx1-4k-video-samples-and-new-full-res-frame-grabs-available-for-do, when LAV DXVA native crashes immediately.

Also, LAV DXVA copy-back doesn't crash, but falls back to SW decoding using the above NX1 clips.

[nevcairiel]

PotPlayer seems to use DXVA for 10bit clips, but with artifacts and colour issues.
But it uses DXVA

I deliberately disabled 10-bit from even attempting to use DXVA2 because broken decoding is worse than no decoding. LAV will never knowingly produce wrong output.

HEVC 10-bit has a special second decoder device called HEVC_VLD_Main10 (or {107AF0E0-EF1A-4D19-ABA8-67A163073D13}). If any GPU supports proper 10-bit decoding, I expect it to offer this device. Its the only way to really know if it supports it or not, without hard-coding a list of GPUs, which I would be very reluctant to do.
In fact, the DXVA2 HEVC specification clearly states the the HEVC_VLD_Main device shall only be used for the "Main" profile, and HEVC_VLD_Main10 for the "Main10" profile.

So far I have not seen this new device on any GPU, even with Intels new driver which presumably claims 10-bit support.

[NikosD]

OK, but Samsung NX1 clips are not 10bit, they are 8bit.

[wizziwig]

So far I have not seen this new device on any GPU, even with Intels new driver which presumably claims 10-bit support.

Only on Broadwell. Haswell is limited to 8-bit only.

https://communities.intel.com/message/273816#273816


10bit hybrid HEVC decode is enabled on BDW U and H sku. Not available on HSW.

8bit hybrid HEVC decode is available on both HSW U sku and above and BDWU and H sku.

Thanks
Robert

[Yups]

Ok that makes sense. No wonder there are playback issues on my HSW with 10 bit. Looks to me that the playback improvements refer to Broadwell.

[nevcairiel]

Thats unfortunate, as I'm unlikely to get a Broadwell-U CPU. Maybe once a desktop Broadwell comes out, or maybe the rumors are true and NVIDIA has 10-bit support in the 960, and working on it there may work on Intel as well.

[NikosD]

Only on Broadwell. Haswell is limited to 8-bit only.

https://communities.intel.com/message/273816#273816

Ok that makes sense. No wonder there are playback issues on my HSW with 10 bit. Looks to me that the playback improvements refer to Broadwell.

Thanks wizziwig.

I really like the speed of posting in Intel forums by Yups [emoji2]

The Broadwell thing crossed my mind, but I didn't want to think that, that was the reason.

Probably VP9 HW acceleration refers to Broadwell only as well.


Thats unfortunate, as I'm unlikely to get a Broadwell-U CPU. Maybe once a desktop Broadwell comes out, or maybe the rumors are true and NVIDIA has 10-bit support in the 960, and working on it there may work on Intel as well.

Nobody with a decent Haswell is going to get a Broadwell-U.

Personally, I'll move directly to Skylake, skipping even desktop Broadwell, even if Skylake is not going to be released along with desktop Broadwell as Intel says.

In my opinion, Skylake will come after desktop Broadwell due to the well known Intel's delays, but that's only a speculation.

[Yups]

Broadwell-K might be interesting to see how GT3e performs but Skylake-S with its Gen9 has a newer Gfx architecture, it should support HEVC decode in hardware finally. This is much more appealing.


In my opinion, Skylake will come after desktop Broadwell due to the well known Intel's delays, but that's only a speculation.


Broadwell-K and Skylake-S are both scheduled for H2 2015. Broadwell-K might come 2-3 months earlier though.

[JohnLai]

Only on Broadwell. Haswell is limited to 8-bit only.

https://communities.intel.com/message/273816#273816

>.< Deep inside my heart, I really hope for Haswell 10bit support.....I guess this is it...I can only wait for either AMD or Nvidia gpu product with fixed hardware acceleration to come out.

BTW, nobody wanna report the mis-exposure of 10bit HEVC DXVA interface for Haswell IGPU at Intel forum?

[nevcairiel]

OK, but Samsung NX1 clips are not 10bit, they are 8bit.

I fixed the Samsung clips, they work fine now.

For 10bit..
Intel broke their drivers, similar to the problem that NVIDIA had quite recently.

For some reason, EVR tries to accept a 10-bit/16-bit input (P016 in this case), but it fails at it, resulting in overall playback failure.

NVIDIA had the same problem a few weeks back, when a driver update caused P010 to be accepted by EVR, but EVR did not handle it properly and failed.

Someone should bring this to Intels attention.
You can work around this problem by disabling P010 and P016 output in LAV if you use EVR.