Handbrake and Haswell Quicksync

Archive of historical development discussions
Discussions / Development has moved to GitHub
Forum rules
*******************************
Please be aware we are now using GitHub for issue tracking and feature requests.
- This section of the forum is now closed to new topics.

*******************************
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Handbrake and Haswell Quicksync

Post by gmb »

For example, previous versions of Quick Sync exposed three pre-defined blends of performance and quality that Intel calls target usages. This time around, there are seven. Really, the intricacies deserve a story of their own. But at the highest-quality TU1 setting, HD Graphics 4600 is significantly better looking than 4000. Meanwhile, the fastest TU7 should be faster, higher-quality, and more battery-friendly for mobile devices on HD Graphics 4600 than 4000.
http://www.tomshardware.com/reviews/cor ... 521-2.html


Does anyone has a Haswell, does Handbrake support these higher quality options for Quicksync?
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

Working on it, but we don't have an updated build yet. Shouldn't be long now.
User avatar
s55
HandBrake Team
Posts: 10350
Joined: Sun Dec 24, 2006 1:05 pm

Re: Handbrake and Haswell Quicksync

Post by s55 »

The GUI was already updated for haswell, so the slider quality position will provide better quality when folks get access to their haswell chips.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Available for Haswell only? On Ivy Bridge I have 3 options with the current Handbrake QS version. I'm planning to buy Haswell in a week.
User avatar
s55
HandBrake Team
Posts: 10350
Joined: Sun Dec 24, 2006 1:05 pm

Re: Handbrake and Haswell Quicksync

Post by s55 »

You'll still have 3 options on Haswell. We are not exposing all 7 target usages as there isn't enough of a difference between them to really justify doing so.

When you choose the Quality option, you will get the quality target usage. Simple.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Not sure if I agree on this. x264 has 10 options and the higher options have small differences usually as well. I will test it out with QSTranscode and see what differences I get.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

In our opinion, the QuickSync results on HD4600 appear to be worse than what is obtained on the HD4000. With Haswell, Intel introduced seven levels of quality/performance settings that application developers can choose from. According to Intel, even the lowest quality Haswell QSV settings should be better than what we had with Ivy Bridge. In practice, this simply isn't the case. There's a widespread regression in image quality ranging from appreciably worse to equal at best with Haswell compared to Ivy Bridge.
http://www.anandtech.com/show/7007/inte ... spective/8

Looks like Haswell is not doing good in Handbrake. They call it broken.
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

gmb wrote:
In our opinion, the QuickSync results on HD4600 appear to be worse than what is obtained on the HD4000. With Haswell, Intel introduced seven levels of quality/performance settings that application developers can choose from. According to Intel, even the lowest quality Haswell QSV settings should be better than what we had with Ivy Bridge. In practice, this simply isn't the case. There's a widespread regression in image quality ranging from appreciably worse to equal at best with Haswell compared to Ivy Bridge.
http://www.anandtech.com/show/7007/inte ... spective/8

Looks like Haswell is not doing good in Handbrake. They call it broken.
It could well be. My own testing under Haswell didn't go really well (didn't check the results, but speed seemed hardly better than under Ivy Bridge, even accounting for the difference in the target usage meaning between IVB and HSW). It's on my TO-DO list. Hopefully I'll finally be able to get some QSV-related work done this week.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

In all other reviews I saw Haswell has faster Quicksync, no surprise really. Something is wrong in this anandtech review (or Handbrake itself).
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

gmb wrote:In all other reviews I saw Haswell has faster Quicksync, no surprise really. Something is wrong in this anandtech review (or Handbrake itself).
Probably the latter, although I'm not 100% sure.
User avatar
s55
HandBrake Team
Posts: 10350
Joined: Sun Dec 24, 2006 1:05 pm

Re: Handbrake and Haswell Quicksync

Post by s55 »

It's pretty bad reporting by Anand not to follow up on that but it seems they are too anxious to get the article out.

Their x264 screenshots look really quite awful. I have no idea what the heck they are up to, but the QuickSync images are maintain a lot more grain and don't look nearly so soft. Makes me think they've tagged the images wrong.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Anand did not write itself, it was Ganesh. Especially this is a lame statement:
The biggest disappointment with Haswell in the media department is the regression in QuickSync video transcode quality. The salt in the wound is really Intel's claims before launch of significant increases in QS video quality. Ivy Bridge definitely produces better quality QSV accelerated video transcodes. Combine that with a lack of significant progress on the software support side until recently (hooray for Handbrake, boo for no substantial OS X deployment) and you'd almost get the impression that Intel was trying its best to ruin one of the most promising features of its Core microprocessors.

There is no reason why Haswell should have worse QS than Ivy Bridge. Based on one QS encoder and just one test video is far from enough and far too early for such a statement.

Hardware.fr tested with MediaEspresso but since mp4 output only converts videos in baseline profile they encountered a bad quality as expected for mp4 Mediaespresso. They should have tried m2ts output which converts the video in high profile. The difference can be huge.
Intel claimed to offer more options in the MediaSDK to improve quality. We wanted to check this practice. Unfortunately, MediaEspresso still retains the same limitations that two years ago on our first test as follows:
- Baseline profile only
- Fixed GOP
- No B frames
http://translate.googleusercontent.com/ ... ld3zLxyu2w
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

I think the fixed GOP size (as in, no scenecut detection) is a limitation of QSV/MSDK? Could be wrong though.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

I can say (valid for Handbrake and QSTranscode) Haswell with quality preset is way slower than IVB. Haswell with balanced preset has comparable speed to IVB quality preset. On Haswell quality preset didn't look better than balanced preset but 2x-3x slower? Quality preset doesn't make sense then. I only tested 2-3 videos though. Handbrake has quality issues compared to QSTranscode by the way. In two of my test videos quality was extremely good with QSTranscode, much better than Ivy Bridge.
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

gmb wrote:Handbrake has quality issues compared to QSTranscode by the way. In two of my test videos quality was extremely good with QSTranscode, much better than Ivy Bridge.
Anandtech found similar issues re: HandBrake/Haswell and quality. So, we're aware of it, but not quite sure what's going on yet.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

The Quick Sync engine in 4th Generation Intel Core Processors includes these new H.264 encoding features:

Per-MB bit rate control
Trellis quantization
Multi-level hierarchical motion estimation
Multi-reference
Multi-predictor
B-pyramid
Lookahead
http://hothardware.com/Reviews/Intel-Co ... d/?page=17


Is this really new with Haswell?
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

I don't know, I'd have to check with someone at Intel. Pretty sure Haswell does have B-pyramid, haven't checked frame types on IVB yet. I was under the impression IVB already supported multiple reference frames, but maybe I'm wrong.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Two videos tested. Encoding time Video 1/Video 2
Testbed: i5-4670-HD4600 9.18.10.3186 Windows 7x64

MediaEspresso 6.7.3521 faster 1:25/0:22 (m2ts output for high profile)
MediaEspresso 6.7.3521 better 2:07/0:56 (m2ts output for high profile)

Because I don't really see a quality improvements with quality preset over balanced on Haswell, I used the balanced preset:

QSTranscode 1008= 0:56/0:28
Handbrake= 2:09 (60 fps encoding)/0:38


Ivy Bridge times with quality preset: viewtopic.php?f=12&t=27251#p125457

(Video 1 and Video 4)

With QSTranscode Haswell Balanced slighty faster than Ivy Bridge Quality whereas Handbrake is slower in both videos with Haswell than Ivy Bridge Quality preset. Screenshots following. I can't do 1080 screenshots till next week on my monitor. I did 720p screenshots. Not ideal but better than nothing.
Deleted User 11865

Re: Handbrake and Haswell Quicksync

Post by Deleted User 11865 »

gmb wrote:Handbrake has quality issues compared to QSTranscode by the way. In two of my test videos quality was extremely good with QSTranscode, much better than Ivy Bridge.
gmb wrote:With QSTranscode Haswell Balanced slighty faster than Ivy Bridge Quality whereas Handbrake is slower in both videos with Haswell than Ivy Bridge Quality preset. Screenshots following. I can't do 1080 screenshots till next week on my monitor. I did 720p screenshots. Not ideal but better than nothing.
We really need to narrow down this issue. Questions:

1) you are running QSTrancode and HandBrake on the same hardware/BIOS/software/driver combination, correct?

2) can you elaborate on the settings used (for both QSTranscode and HandBrake)? source video codec (H.264 or other), target usage (QSTranscode accepts a number, IIRC), H.264 profile and level, resolution, framerate, any other options used in QSTranscode/HandBrake…
maxim_d33
Bright Spark User
Posts: 175
Joined: Mon Mar 21, 2011 12:02 pm

Re: Handbrake and Haswell Quicksync

Post by maxim_d33 »

if we would like to go in more details, we can start to use Intel® Media Software Development Kit Tracer,
this application allows to do a snapshot of application and MediaSDK activity.

Practically, it created a log file with many useful details written inside.

This will help to see difference(s) in settings etc
just by comparison of these logs.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

MediaEspresso Faster
(Video 1) http://img5.imageshack.us/img5/8103/1me ... term2t.png
(Video 2) http://img853.imageshack.us/img853/4571 ... faster.png

MediEaspresso Better
(Video 1) http://img191.imageshack.us/img191/1095 ... term2t.png
(Video 2) http://img849.imageshack.us/img849/5594 ... better.png

QSTranscode
(Video 1) http://img404.imageshack.us/img404/4103 ... nscode.png
(Video 1) http://img818.imageshack.us/img818/8995 ... lanced.png
(Video 2) http://img38.imageshack.us/img38/2277/4qstranscode.png

Handbrake
(Video 1) http://img33.imageshack.us/img33/2127/1handbrake.png
(Video 1) http://img203.imageshack.us/img203/117/ ... dhaswe.png
(Video 2) http://img189.imageshack.us/img189/7927/4handbrake.png


Ivy Bridge Handbrake Quality preset
(Video 1) http://img543.imageshack.us/img543/4842 ... ivybri.png
(Video 1) http://img32.imageshack.us/img32/2797/8 ... ivybri.png
(Video 2) http://img441.imageshack.us/img441/9127 ... ivybri.png

X264 Medium high profile
(Video 1) http://img855.imageshack.us/img855/3657/1x264medium.png
(Video 1) http://img40.imageshack.us/img40/1435/8x264medium.png

X264 Super Fast high profile
(Video 2) http://img811.imageshack.us/img811/6734 ... erfast.png


There is nothing wrong in video 2 with Handbrake, comparable to QSTranscode. Compared to Ivy Bridge way better as you can see. QSTranscode does the best job in Video 1, even better than x264 Medium encoding and of course a nice improvement over Ivy Bridge as well. Only 2 videos, so it is too early for a summary. But it shows that Haswell has big potential to improve over Ivy Bridge.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Rodeo wrote:
gmb wrote:Handbrake has quality issues compared to QSTranscode by the way. In two of my test videos quality was extremely good with QSTranscode, much better than Ivy Bridge.
gmb wrote:With QSTranscode Haswell Balanced slighty faster than Ivy Bridge Quality whereas Handbrake is slower in both videos with Haswell than Ivy Bridge Quality preset. Screenshots following. I can't do 1080 screenshots till next week on my monitor. I did 720p screenshots. Not ideal but better than nothing.
We really need to narrow down this issue. Questions:

1) you are running QSTrancode and HandBrake on the same hardware/BIOS/software/driver combination, correct?

2) can you elaborate on the settings used (for both QSTranscode and HandBrake)? source video codec (H.264 or other), target usage (QSTranscode accepts a number, IIRC), H.264 profile and level, resolution, framerate, any other options used in QSTranscode/HandBrake…

Yes all on the same system. I use VBR ~6 Mbps in Full HD for all videos. For Video 1 here is a Link: http://dlsecondary.gamersyde.com/secdl/ ... utpost.mp4

Source Video Codec AVC 36.8 Mbps. Here my command line:

qstranscode.exe -i c:\Far.mp4 -o c:\Faroutput.mp4 -h264 -w 1920 -h 1080 -f 30 -aac -b 6315 -u 4 -rc 2
qstranscode.exe -i c:\avatar.m2ts -o c:\avataroutput.mp4 -h264 -w 1920 -h 1080 -aac -b 6760 -u 4 -rc 2

Bitrate can differ because I have to match the bitrate from other videos.

Video 1 is a 60 fps video, so I used -f 30. For my other test videos I don't use -f command. Encoding Profile and Level is automatically adjusted from the SDK. Same for Handbrake (auto).
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

Overview from Intel: http://newsroom.intel.com/servlet/JiveS ... s-4200.pdf

With TU1-TU7 comparisons.
gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

gmb
Bright Spark User
Posts: 350
Joined: Thu Mar 28, 2013 12:49 pm

Re: Handbrake and Haswell Quicksync

Post by gmb »

New HSW vs IVB comparison. Video Bitrate on all the same. VBR encoding used. I have used QSTranscode for Haswell. I may test Handbrake later but for my limited time now I didn't because Handbrake possibly has some issues with Haswell currently.

Video 1
QSTranscode Ivy Bridge Quality 0:58= http://imageshack.us/a/img109/4103/1qstranscode.png
Handbrake Ivy Bridge Quality 1:52= http://imageshack.us/a/img801/2127/1handbrake.png
QSTranscode Haswell Balanced 0:56= http://imageshack.us/a/img837/1386/10m2.png

Video 2
QSTranscode Ivy Bridge Quality 0:09= http://imageshack.us/a/img849/5572/2qstranscode.png
Handbrake Ivy Bridge Quality 0:11= http://imageshack.us/a/img259/1391/2handbrake.png
QSTranscode Haswell Balanced 0:07= http://imageshack.us/a/img812/3678/q1n.png

Video 3
QSTranscode Ivy Bridge Quality 1:21= http://imageshack.us/a/img248/3800/3qstranscode.png
Handbrake Ivy Bridge Quality 1:28= http://imageshack.us/a/img10/3553/3handbrake.png
QSTranscode Haswell Balanced 1:19= http://imageshack.us/a/img546/6812/qwt.png

Video 4
QSTranscode Ivy Bridge Quality 0:30= http://imageshack.us/a/img543/796/83l.png
Handbrake Ivy Bridge Quality 0:35= http://imageshack.us/a/img199/8830/evfs.png
QSTranscode Haswell Balanced 0:25= http://imageshack.us/a/img248/543/vtt.png

Video 5
QSTranscode Ivy Bridge Quality= failed (fixed in 1009)
Handbrake Ivy Bridge Quality 0:36= http://imageshack.us/a/img842/7936/5handbrake.png
QSTranscode Haswell Balanced 0:16= http://imageshack.us/a/img153/3505/84a.png

Video 6
QSTranscode Ivy Bridge Quality 0:55= http://imageshack.us/a/img27/867/6qstranscode.png
Handbrake Ivy Bridge Quality 0:59= http://imageshack.us/a/img191/8673/6handbrakeqs.png
QSTranscode Haswell Balanced 0:49= http://imageshack.us/a/img404/4497/luw.png


Video 7
QSTranscode Ivy Bridge Quality 1:08= http://imageshack.us/a/img600/9541/7qstranscode.png
Handbrake IVy Bridge Quality 1:18= http://imageshack.us/a/img195/2650/7handbrakeqs.png
QSTranscode HSW Balanced 0:56= http://imageshack.us/a/img580/6189/mba.png

Video 7 with QP encoding instead VBR (video Bitrate 300-400 kbps less than VBR):

Handbrake Ivy Bridge Quality QP26 1:09= http://imageshack.us/a/img51/2881/7handbrakeqp26.png
QSTranscode Haswell Balanced QP26 0:48= http://imageshack.us/a/img545/676/qok.png


IVB used: i5-3570k
HSW used: i5-4670
Post Reply