[resolved]Old laptop MAC hardware faster than a new PC?

[karasu]

I'm not starting a flamewar. This is just an observation and I want to understand where the problem is.

I have a 2 years old Macbook Pro with Core Duo 2.16Ghz (first gen) 666Mhz FSB and 1x1Go DDR2 666Mhz.

I recently buy a Core2 duo 2.2Ghz (E4500) 800Mhz FSB and 2x1Go DDR2 666Mhz.

With handbrake 0.92, I encode a DVD, same source and settings for both

On the Macbook under OSX 10.4 I have 20FPS for the job.

On the PC under XP SP2 I have 12FPS.


With the latest MeGui and the same x264 option I have similar speed (~12fps)

This is very strange for a fastest CPU and bus speed. Can someone explain that?

[nm]

I doubt the encoding settings were exactly the same. To find out, use some H.264 stream analyzer (or strings -50 file.mp4) on the files to check the information about encoding parameters that x264 writes there. Or upload sample clips somewhere so that we can take a look.

[karasu]

Here's the results :

For The PC under XP with Handbrake (13FPS):

x264 - core 58 svn-736M - H.264/MPEG-4 AVC codec - Copyleft 2003-2008 - http://www.videolan.org/x264.html -
options: cabac=1 ref=5 deblock=1:-1:-1 analyse=0x3:0x133 me=hex subme=4 brdo=0 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 chroma_qp_offset=0 threads=3 nr=0 decimate=0 mbaff=0 bframes=3 b_pyramid=1 b_adapt=1 b_bias=0 direct=1 wpredb=1 bime=1 keyint=250 keyint_min=25 scenecut=40(pre) rc=crf crf=20.4 rceq='blurCplx^(1-qComp)' qcomp=0.60 qpmin=10 qpmax=51 qpstep=4 ip_ratio=1.40 pb_ratio=1.30


For the Macbookpro os10.4 Handbrake (23FPS):

x264 - core 56 - H.264/MPEG-4 AVC codec - Copyleft 2005 - http://www.videolan.org/x264.html -
options: cabac=1 ref=5 deblock=1:-1:-1 analyse=0x3:0x133 me=hex subme=4 brdo=0 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 chroma_qp_offset=0 threads=3 nr=0 decimate=0 mbaff=0 bframes=3 b_pyramid=1 b_adapt=1 b_bias=0 direct=1 wpredb=1 bime=1 keyint=500 keyint_min=25 scenecut=40(pre) rc=crf crf=19.4 rceq='blurCplx^(1-qComp)' qcomp=0.60 qpmin=10 qpmax=51 qpstep=4 ip_ratio=1.40 pb_ratio=1.30


I dont see differences that can explain the underperformance of my E4500...

[nm]

Ok, does x264 completely saturate both cores of the CPU on Windows? Maybe Handbrake's x264 wasn't compiled with pthreads.

[karasu]

Yes, both cores ar fully used.

Here's informations about the x264 encoder for windows :
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSE3 SSSE3 Cache64

and for OS10.4 :
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2

This is more strange, no sse3 optimisations for the MacbookPro...

[burfadel]

The x264 version used is different for a start!, so that would make it hard to compare...?

I don't use handbrake, although maybe other filters play a part? Although Mac users may not want to hear this, but the latests Mac's are essentially the same as a 'proper' ;) computer anyway, so there's no real reason why a properly optimised version of each programme should encode at a much different rate on the different platforms.

[Irakli]

The x264 version used is different for a start!, so that would make it hard to compare...?

Yes, it is very likely that speed differences are explained by different versions of x264. Version used in Handbrake on Mac is very old x264 core 56 (Copyleft 2005!), while Windows uses (almost) most recent x264 core 58 svn-736M.

Regards,
Irakli

[karasu]

The x264 version used is different for a start!, so that would make it hard to compare...?
I don't use handbrake, although maybe other filters play a part?

But the windows version is much newer and use more optimisations, but still slower...
I don't think that two recent different version of x264 can be so different in speed. The problem must be somewhere else.

I also try the latest version of MeGui with latest x264 build for windows with the same x264 setting and no filter, the encode speed is the same (13fps)

Although Mac users may not want to hear this, but the latests Mac's are essentially the same as a 'proper' ;) computer anyway, so there's no real reason why a properly optimised version of each programme should encode at a much different rate on the different platforms.
I totally agree, but I still see differences and can't find what is going wrong with my PC...

[karasu]

Yes, it is very likely that speed differences are explained by different versions of x264. Version used in Handbrake on Mac is very old x264 core 56 (Copyleft 2005!), while Windows uses (almost) most recent x264 core 58 svn-736M.

Regards,
Irakli

Is the x264 core 56 known to be so fast?

[nm]

Is the x264 core 56 known to be so fast?
No, and it's probably not that old anyway. Probably r6xx.

Perhaps there is something wrong with your hardware setup. Try some CPU/memory benchmark that allows you to compare results to those of other similar systems.

[karasu]

Perhaps there is something wrong with your hardware setup. Try some CPU/memory benchmark that allows you to compare results to those of other similar systems.

I tried this test as a benchmark : http://www.techarp.com/showarticle.aspx?artno=442&pgno=0#Results
It's a x264 test with 2 passes.

In the results, there is someone with a similar system (E4500 @2.2Ghz) with theses results : first pass: 56.10fps second:14.48fps

My results are: fisrt pass: 52fps, second : 14fps

So my machine doesn't seem to be so different.

If someone here also have a 2.2Ghz Core2 duo or similar, please post your results.

It's hard to find benchmark with the same hardware as me.

Data provided by CPU-Z are ok, no problem with my bus speed or memory. I don't have a hungry process, only x264 is at work during my test.

[nm]

Are you sure the FPS numbers are reported correctly by Handbrake--how about total encoding time? Could you post short sample clips of an encode done on both systems?

[karasu]

Ok, now the fog has gone!

Here's my mistake, I didn't play the results file after my last encoding tests. It appear that Handbrake for windows is unable to decrypt vob on the fly. So the files encoded under windows were scrambled. And encoding a semi-encrypted vob take more time than decrypted vob...

And for the megui result, I think the problem was the encode settings. I can't obtain exactly same setting when using the gui, Using custom command line give me similar results in speed.

Now here's my latest results :
The source is a chapter from a DVD; 3min31sec.
here's the query string :
-i "source.vob" -o "output.mp4" -e x264 -E ac3 --crop 72:72:0:0 -p -q 0.6 -x ref=5:mixed-refs=1:bframes=3:subq=4:weightb=1:brdo=1:bime=1:b-pyramid=1:me=hex:merange=16:analyse=all:8x8dct=1:deblock=-1,-1:trellis=1:no-fast-pskip=1:no-dct-decimate=1 -R 48 -D 1 -a none -v
Handbrake on the mac : 19.23FPS output file (http://gloomydream.net/bazar/encode/mac-chap18.mp4)
Handbrake on th PC: 22.44FPS output file (http://gloomydream.net/bazar/encode/pc-chap18.mp4)

So my new PC is 15% faster than my laptop, now everything look ok.

I'm confused about my mistake, sorry for the trouble. These days I'm trying every x264 options to observe the results in speed and quality. And after half a hundred of test I've lost my rigour when I begin to compare speed between my two machines.

Thank you for your perseverance nm!