GTX980 hardware encoder support now the dev kit is out (giving 10x the speed)

[SE09]

Please describe what you would like to see added to HandBrake:

I would like to have GTX980 h265 hardware encode support in handbrake

Why would you like to see this added:

h265 1080p and 4k makes good videos but even on CPU's like the i7 it can take many hours to convert a video file to h265
waiting a few years for better CPU's brings about 2x the speed, but using the video cards hardware it gives us a speed up now of 10x the speed (i7 4790k vs GTX980 in my test)
(I have got a 2.2gb 2h.10m 720p movie down to 400mb but takes 7+ hrs)
What version of HandBrake are you currently running:

0.10.5.0 64bit

What operating system and version and you currently running (e.g. OSX 10.11, Windows 7, Ubuntu 14):
Windows 7 64bit 16gb ram 17 4790k GTX980

please consider this and other video card support as 4k is with us

[Woodstock]

Is the developer kit compatible with open source projects like Handbrake?

[s55]

Your probably not going to find much developer interest in this. 3rd party patches may be accepted if they are GPL compatible.

Also, 10x faster than what? If it's 10x faster than the reference encoder then it's marketing rubbish. If it's 10x faster than *equivalent* x265 settings and produces better or similar results, that's a bigger deal.

I've not seen any good comparisons of them yet.

[mduell]

10x faster than what is the important question.

Certainly not x265 at the same efficiency (quality per bit).
Probably not QSV on modern CPUs.

Just a garbage unsubstantiated claim.

[SE09]

Your probably not going to find much developer interest in this. 3rd party patches may be accepted if they are GPL compatible.

Also, 10x faster than what? If it's 10x faster than the reference encoder then it's marketing rubbish. If it's 10x faster than *equivalent* x265 settings and produces better or similar results, that's a bigger deal.

I've not seen any good comparisons of them yet.

The 10x is in my tests

it is just under 10x faster with h265 on my ref setup (i7 4790k 980)
just using the cpu vs using the 980, the 980 is 9.5 x faster (I'm testing some software for the devs at a video encoder maker that has just implemented the new hardware encoder using the free dev pack)

its always been about 10x faster in cpu vs video card as I did the same test years ago with a AMD 5870 vs i7 960

I don't know if I can post a video or screens showing it as I don't want to plug the software as handbrake is the one I like to use but it should be okay I guess as its the h265 that's the star not the software

[SE09]

Is the developer kit compatible with open source projects like Handbrake?

Sorry I didn't think of that but I think the dev pack is free I did see a post with a link in it saying give this to any software developers you can find but not sure if it is fully open source compatible
I just saw the quick sync option in handbrake and thought more hardware support would help close the speed gap we always get in the first few years of a game changing video coding standard like h265

[SE09]

10x faster than what is the important question.

Certainly not x265 at the same efficiency (quality per bit).
Probably not QSV on modern CPUs.

Just a garbage unsubstantiated claim.

Yes I can say it is just under 10x faster using a i7 4790k vs GTX980 using the same file same videos settings
since the beginning of videos cards being used to boost video conversion speeds hardware has always been about 10x over a cpu
unless the cpu was a top end monster and the video card was a low end card

i'll tell you what I will make two short output videos of cpu vs video card using the same software with the settings info and see if there is a loss in the video quality

[s55]

using the same file same videos settings

Well, right off the bat, that's wrong. NVEnc, QuickSync and x264, x265 all have different features and abilities. it's very difficult to get them to sync up giving they are not all equal to begin with.

Presets are not equivalent between the encoders. Even the RF value is not equivalent.

Doing fair testing across encoders is not trivial. It's complex and time consuming task.


Regardless, not the point of the thread.

The response is the same. There isn't any developer interest, but if someone wants to produce a patch, (preferably something that hooks the libav interface), we'll be willing to accept that if it's a good quality patch.

[SE09]

using the same file same videos settings

Well, right off the bat, that's wrong. NVEnc, QuickSync and x264, x265 all have different features and abilities. it's very difficult to get them to sync up giving they are not all equal to begin with.

Presets are not equivalent between the encoders. Even the RF value is not equivalent.

Doing fair testing across encoders is not trivial. It's complex and time consuming task.


Regardless, not the point of the thread.

The response is the same. There isn't any developer interest, but if someone wants to produce a patch, (preferably something that hooks the libav interface), we'll be willing to accept that if it's a good quality patch.

See for yourself
this is only to show h265 CPU vs h265 videocard and not what video software is best, we all know that handbrake is best

I still don't know if the dev kit is open but I will look as any speed up is a good thing

EDIT: Sorry in the video showing the recoding times there is a pause where nothing happens for 50 seconds about 3/4 just before the hardware recode part, I forgot I was recording so was listening to my Son beg me for Tom Clancy's Siege

Hi I've made a test n show rar for anyone who wants to see for themselves
its a rar file then link is http://www.filedropper.com/h265cpuvshardware

the password is handbrakeisbest
the rar is 50 meg and the SAH-1 is 06EA248FD5B2AB9718C0F8128515594E1249D3D8

in the folder you will find the video info for the two videos showing all the info you need like size and bitrate
one video is h265 using the CPU a i7 4790k and the other is the same but h265 but using a GTX980

the CPU one uses about 99% CPU and the GTX one uses about 20% CPU

i made a mistake in telling you that using a video card was 10x faster, it’s not, it’s 14x faster

recording the screen to show the speed of the conversion caused the CPU made video to stutter so i remade the same video after the screen record was turned off

i also set the converter to auto causing the video card made video to be bigger than the CPU made one so i redid that too
this time setting the bit rate to the same made it even faster, 14x faster


i didn't do a handbrake one as its not to see what is best its to show the same h265 software vs hardware
i was always told CPU made better videos over videos cards so i still use HB 99% of the time

problem as the software i’m testing is not done it is causing the video that i made using CPU only to stutter when being played in vlc, its the same bitrate and length and looks the same but it stutters so sorry about that but I can not do anything about that atm

also has anyone used the quick sync option in handbrake, how much faster is it, is it the same video quality?

thanks for your time and please remember this is not handbrake vs whatever its a request for video card support to help speed up h265 recoding
but as said the dev kit may not be right for handbrake

one thing handbrake could do with is a output file size option box
like set output to fit on a CD

thanks to the devs and helpers on this forum and in handbrake, its a must have

[mduell]

10x faster than what is the important question.

Certainly not x265 at the same efficiency (quality per bit).
Probably not QSV on modern CPUs.

Just a garbage unsubstantiated claim.

Yes I can say it is just under 10x faster using a i7 4790k vs GTX980 using the same file same videos settings

LOL, you have no idea what you're talking about. Rubbish thread.

[SE09]

10x faster than what is the important question.

Certainly not x265 at the same efficiency (quality per bit).
Probably not QSV on modern CPUs.

Just a garbage unsubstantiated claim.

Yes I can say it is just under 10x faster using a i7 4790k vs GTX980 using the same file same videos settings

LOL, you have no idea what you're talking about. Rubbish thread.

Wow did you mum tell you about social skills
I did a like for like test using the same settings just to show how fast gpu vs cpu on my system was
it was a polite request for a option so take your nasty attitude and grow up and act like a man
how old are you kiddy do you talk to all new posters like that?

it was a like for like test using the software that was available to me so tell me what was wrong with me testing and asking for a option

what an abusive person you are

[mduell]

I did a like for like test using the same settings just to show how fast gpu vs cpu on my system was

You used the same settings on two completely different video encoders?

Shocker no one believes you.

The ad hominem attacks are unwarranted.

[s55]

This thread has outlived it's usefulness ...


For people who can't be bothered with the noise. The official line is this:

There isn't any developer interest, but if someone wants to produce a patch, (preferably something that hooks the libav interface), we'll be willing to accept that if it's a good quality patch.