Benchmark your computer with Handbrake

[Deders]

Hmm, Makaveli pipped me on the line with that sweet Mushkin Ram - they are my favorite RAM maker (awesome customer service), but I cheaped out going from 6 G to 12 GB (which annoyingly doesn't want to run @ 1T).

That setting us usually to help the memory controller cope with more sticks of ram, can you say you've noticed any difference in performance or responsiveness?

 

[t-ray]

Will some one that has a quicksync-enabled machine that also dual boots between Windows and Linux try this out in both environments? I'd like to see the performance difference.

 

[BallaTheFeared]

QS is slower for me than the avx2 version in this.

 

[SPBHM]

QS is slower for me than the avx2 version in this.

quick sync?

the Android profile is not going to use quick sync,

 

[Makaveli]

That setting us usually to help the memory controller cope with more sticks of ram, can you say you've noticed any difference in performance or responsiveness?

I think it was more noticeable before cpu's had the memory controllers on die.

However it doesn't mean there isn't a performance difference regardless of how small it is. Would it be enough for you to buy more expensive memory is totally up to the buyer and their budget.

Ajay and I are using the same motherboard same cpu and same overclock yet I'm getting more FPS and its because of the memory timings no doubt.

 

[BallaTheFeared]

quick sync?

the Android profile is not going to use quick sync,

Which would be why the QS version is slower then, it doesn't contain the FMA3 and AVX2 code which boosts the performance of Haswell with that version.

 

[ChronoReverse]

This test really should be using the AppleTV 3 profile instead. The workload for the Android preset is too easy for modern CPU's.

I'll still contribute though:

Core 2 Quad Q9450 @ 3.4GHz
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.1 Cache64
work: average encoding speed for job is 110.889351 fps






As a sidenote, the available OpenCL encoders aren't very good and the CUDA ones are terrible. Not always faster and always much worse quality.

Quick Sync is the fastest but quality is still not very good. x264 (the software encoder Handbrake uses) at its lowest quality setting is as fast as the slowest Quick Sync setting for similar quality. Except you have the option of actually getting good to best-in-class quality with x264.

 

[mavere]

This was (probably) mentioned at some point in the previous 10 pages, but the x264 version used here was outdated when the current Handbrake was released and is *quite* outdated by now, as the developers have been on an optimization spree the past several months.

There's nothing necessarily wrong with comparing numbers on an outdated backend, but hopefully everyone is on the same page and recognizes the overly transient nature of these results.

 

[Hitman928]

Just an fyi, I benched this with quick sync enabled on my IB laptop and went from ~95 fps to:

[19:40:18] work: average encoding speed for job is 427.752808 fps
[19:40:19] sync: got 14315 frames, 14315 expected

The output doesn't seem to play quite right though, skips a bunch and looks quite washed out. It could be a driver thing or a setting thing, didn't really have time to spend messing with it though.

 

[cmdrdredd]

Just an fyi, I benched this with quick sync enabled on my IB laptop and went from ~95 fps to:



The output doesn't seem to play quite right though, skips a bunch and looks quite washed out. It could be a driver thing or a setting thing, didn't really have time to spend messing with it though.

I remember reading that while quick sync is pretty fast, the quality is very poor compared to just letting the CPU do it all in software. Cuda is not as fast (for apps that can use it) but the quality is a bit better (more accurate might be the proper term here).

The best way for quality is just let the CPU do it. A modern CPU is pretty fast for this task anymore.

 

[tential]

I remember reading that while quick sync is pretty fast, the quality is very poor compared to just letting the CPU do it all in software. Cuda is not as fast (for apps that can use it) but the quality is a bit better (more accurate might be the proper term here).

The best way for quality is just let the CPU do it. A modern CPU is pretty fast for this task anymore.

Wasn't that in the Anandtech review on handbrake?

 

[Essence_of_War]

On my work computer:

HandBrake 0.9.9.5530 - 64bit Version
CPU: Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.1 Cache64
work: average encoding speed for job is 48.225414 fps

 

[YBS1]

For kicks and giggles I cranked up the clock speed on my old Xeon W3520 to 4.3GHz (it's only running a simple Corsair H60 now as opposed to the Swiftech setup it used to be on so it can't quite get the old clocks) and it actually held it's own pretty well on this. It managed to break the 10,000 points barrier in 3DMark FireStrike physics test as well which suprised me.

HandBrake svn5649 (Nightly Build) - 64bit Version
OS: Microsoft Windows NT 6.2.9200.0
CPU: Intel(R) Xeon(R) CPU W3520 @ 2.67GHz
Ram: 6135 MB, Screen: 1920x1080

[11:01:01] work: average encoding speed for job is 227.670135 fps

 

[bbhaag]

AMD Turion X2 64-2.1ghz

[11:41:59] starting job
[11:52:24] work: average encoding speed for job is 22.968678 fps

x264 [info]: using cpu capabilities: MMX2 SSE2Slow SlowCTZ

 

[Essence_of_War]

My home comp:
Standard build - HandBrake 0.9.9.5530 - 64bit Version
CPU: Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX
work: average encoding speed for job is 213.928909 fps

 

[Makaveli]

My home comp:

CPU: Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 FastShuffle SSE4.2 AVX
work: average encoding speed for job is 427.729340 fps

Please use the andriod preset these numbers are way off what you should be getting on that chip.

And you need to state which build you are using Standard or nightly.

 

[sm625]

The output doesn't seem to play quite right though, skips a bunch and looks quite washed out. It could be a driver thing or a setting thing, didn't really have time to spend messing with it though.

After two+ years, that seems to be the best way to sum up quicksync. Intel has successfully managed to etch vaporware onto a cpu die.

 

[Essence_of_War]

Please use the andriod preset these numbers are way off what you should be getting on that chip.

And you need to state which build you are using Standard or nightly.

Fixed. My bad, misclicked presets

 

[bbhaag]

Intel Core Duo @ 1.66ghz Yonah

[16:12:06] starting job
[16:29:14] work: average encoding speed for job is 13.964025 fps

x264 [info]: using cpu capabilities: MMX2 SSE Cache64

 

[ChronoReverse]

After two+ years, that seems to be the best way to sum up quicksync. Intel has successfully managed to etch vaporware onto a cpu die.

It's not really vapourware since it at least works, but quality leaves something to be desired as with every single "accelerated" encoder has so far.

At least Quicksync's quality in its "best/slowest" setting is about as good as x264 in its "worst/fastest" setting which is more than could be said about OpenCL encoders (which are simply bad) or, *shudder*, CUDA encoders (which output results worse than bad).

 

[NostaSeronx]

Old

HandBrake svn5614 (Nightly Build) - 64bit Version
OS: Microsoft Windows NT 6.2.9200.0
CPU: AMD FX(tm)-8320 Eight-Core Processor
Ram: 16336 MB, Screen: 1920x1080
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX XOP FMA4 FMA3 SSEMisalign LZCNT BMI1
work: average encoding speed for job is 200.373764 fps

New => New Build and Centurion AGESA

HandBrake svn5675 (Nightly Build) - 64bit Version
OS: Microsoft Windows NT 6.2.9200.0
CPU: AMD FX(tm)-8320 Eight-Core Processor
Ram: 16336 MB, Screen: 1920x1080
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX XOP FMA4 FMA3 SSEMisalign LZCNT BMI1
work: average encoding speed for job is 207.006699 fps

Nothing to brag about I guess.
---
http://sourceforge.net/projects/handbrake/files/Betas/QSV/ <-- QuickSync
Simply select the "H.264 (Intel QSV)" encoder option from the "Video Encoder" dropdown menu on the "Video" tab.
We highly recommend running Intel's iGFX driver 15.31.3071 or later for best results.

 

[monstercameron]

amd phenom ii 965BE @4GHz, 8gb 1333 ddr3, ubuntu 64-bit

x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSEMisalign LZCNT
x264 [info]: profile Main, level 3.0
[16:08:15] reader: done. 1 scr changes
[16:08:16] work: average encoding speed for job is 120.524353 fps

amd phenom ii 965BE @4.1GHz, 4gb 1600 ddr3, ubuntu 64-bit

x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSEMisalign LZCNT
x264 [info]: profile Main, level 3.0
[21:37:18] reader: done. 1 scr changes
[21:37:18] work: average encoding speed for job is 123.208076 fps

slight bump for 3fps...

 

[YBS1]

Well, I did get my temps under control well enough to push the Xeon W3520 up to 4458MHz. I had it running at 4522 but it wouldn't complete the encode correctly (Was hoping to push past that 3770K Ivy Bridge lol). Not too shabby for a chip that debuted in early 2009.

HandBrake svn5675 (Nightly Build) - 64bit Version
OS: Microsoft Windows NT 6.2.9200.0
CPU: Intel(R) Xeon(R) CPU W3520 @ 2.67GHz (4.458Ghz)
Ram: 6135 MB, Screen: 1920x1080
[23:21:33] work: average encoding speed for job is 237.178284 fps

 

[soccerballtux]

amd phenom ii 965BE @4GHz, 8gb 1333 ddr3, ubuntu 64-bit

x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSEMisalign LZCNT
x264 [info]: profile Main, level 3.0
[16:08:15] reader: done. 1 scr changes
[16:08:16] work: average encoding speed for job is 120.524353 fps

amd phenom ii 965BE @4.1GHz, 4gb 1600 ddr3, ubuntu 64-bit

x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSEMisalign LZCNT
x264 [info]: profile Main, level 3.0
[21:37:18] reader: done. 1 scr changes
[21:37:18] work: average encoding speed for job is 123.208076 fps

slight bump for 3fps...

interesting, I had 118fps (see sig for specs). What's your CPU-NB set at?

 

[Edgemeal]

amd phenom ii 965BE @4GHz, 8gb 1333 ddr3, ubuntu 64-bit

slight bump for 3fps...

What about the nightly build? I have you on the handbreak score chart for 127.271416 FPS from your nightly build run here.