x264 Benchmark

[IMNOTDRPHIL]

Here are the results from my desktop:

---------- RUN1PASS1.LOG
encoded 1749 frames, 58.21 fps, 1850.89 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 58.12 fps, 1850.89 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 58.00 fps, 1850.89 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 58.09 fps, 1850.89 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 58.00 fps, 1850.89 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 14.14 fps, 1826.38 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 14.22 fps, 1826.38 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 14.19 fps, 1826.37 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 14.22 fps, 1826.38 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 14.23 fps, 1826.37 kb/s

System information:
CPU: AMD Athlon 64 X2 4200+
Code name: Manchester
Total number of cores: 2
CPU settings: 11x200 (stock)
Core Rate: 2.20 GHz (stock)
MB chipset: NVIDIA nForce 4 SLi
Timings/Frequency: 3-3-3-8 @ 400 MHz
OS: Windows XP SP2 32-bit
MeGUI Test Data FPS1: 58.14
MeGUI Test Data FPS2: 14.22

[JnZ]

Strange thing Graysky,
I run this bench manytimes, but every run gives different results.
1st pass gives 90-120fps, 2nd pass 31-35fps. I think, this isn't good benchmark...problem is maybe in x264 auto threading model, sometimes it detects more threads, sometimes not and this benchmark is too "light" for todays CPU's, 480p is too easy. You should thing over to make new HD benchmark, 720p is enough I think.

Second thing, you shoud give time results, no FPS, time of 1+2 pass.
For example:
CPU1 runs 1-pass 120fps, 2nd at 35fps. CPU2 runs 130fps and 34fps, but this mean CPU1 is better than CPU2 even if CPU2 have more 1st pass speed (coz 2nd pass takes more time and total time 1+2 pass for CPU1 is less)

Last thing, i'm not sure if is good to use loop in AVS. (I know, video size, etc.)

I must say with this test, my dual opteron is runnig about 70-90% cpu time, I gain more speed with HD encodings, which runs about almost 100% cpu usage.

This is only suggestions no critique
Best regards
John

[graysky]

Originally Posted by JnZ

Strange thing Graysky,
I run this bench manytimes, but every run gives different results.
1st pass gives 90-120fps, 2nd pass 31-35fps. I think, this isn't good benchmark...problem is maybe in x264 auto threading model, sometimes it detects more threads, sometimes not and this benchmark is too "light" for todays CPU's, 480p is too easy. You should thing over to make new HD benchmark, 720p is enough I think.

I'm not too sure what to make of that; in general, the data are pretty tight. I wonder if the results you're describing are a result of an unstable o/c (just guessing here, take no offense). Do you see this range when you run your machine @ stock levels? I just looked up your results in the table:

[code]124.65 124.51 124.38 124.24 123.55 34.96 34.84 35.09 34.7 34.78[/code] They seem pretty tight...??

I also noted that your machine is a dual, DUAL core machine (i.e. two chips x two cores per chip) and I have seen people w/ this setup before get <100 % CPU usage. I dunno about the --threads auto on your machine. You might try a test for yourself and change the auto setting to 6 which is what a 4 core chip gets in an auto mode. I do NOT think that the auto is getting confused and changing each time; it should be consistent in whatever it's selecting.

Also, I thought about doing the 1280x720 benchmark, but figured 720x480 would be good since the resulting fps values will be higher. Just FYI, in my own testing, the results are nearly linear between 480p and 720p. In other words, 2-2/3 times as slow for the 720p results.


[code]Second thing, you shoud give time results, no FPS, time of 1+2 pass.
For example:
CPU1 runs 1-pass 120fps, 2nd at 35fps. CPU2 runs 130fps and 34fps, but this mean CPU1 is better than CPU2 even if CPU2 have more 1st pass speed (coz 2nd pass takes more time and total time 1+2 pass for CPU1 is less)[/code]

Well, they aren't really sorted by anything but CPU model -- and they are sorted loosely at that. The problem becomes who to I keep multiple results from the same machine tied together if I sorted in the spreadsheet. At some point in the future, I hope someone in the TA user group can do some asp coding for the results to make them sortable/filterable within the html.

I might take your suggestion to compute a total encode time, but I think the table is too large as it is.

Last thing, i'm not sure if is good to use loop in AVS. (I know, video size, etc.)

Yeah, I dunno how else to minimize the filesize. I will say that the times on my machine for the looped clip match up with the times on my machine when I have it do an hour+ long 720x480 pretty closely.

I must say with this test, my dual opteron is runnig about 70-90% cpu time, I gain more speed with HD encodings, which runs about almost 100% cpu usage.

This is only suggestions no critique
Best regards
John

Again, I think this might have something do w/ your dual chip machine, but I really don't know for sure.

Thanks for taking the time to write up your suggestions by the way!

[graysky]

@IMNOTDRPHIL - thanks for the results!

[ditche]

I"m so noob...
During the bench, my P-M was ALWAYS @ 600 MHz...

It's slightly better @ 1,5 GHz.




---------- RUN1PASS1.LOG
encoded 1749 frames, 17.35 fps, 1854.10 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 17.67 fps, 1854.10 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 17.77 fps, 1854.10 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 18.03 fps, 1854.10 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 17.95 fps, 1854.10 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 4.42 fps, 1825.89 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 4.50 fps, 1825.89 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 4.52 fps, 1825.89 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 4.54 fps, 1825.89 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 4.55 fps, 1825.89 kb/s

[Emmett]

graysky..

Yes, 582 in bios..

Emmett

[Wwhat]

As to the OS I had the error (for mr. wong), it was WinXP

[PsYkHoTiK]

http://valid.x86-secret.com/show_oc.php?id=227164

---------- RUN1PASS1.LOG
encoded 1749 frames, 112.01 fps, 1850.89 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 111.45 fps, 1850.89 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 111.89 fps, 1850.89 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 111.89 fps, 1850.89 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 112.34 fps, 1850.89 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 27.96 fps, 1826.38 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 27.83 fps, 1826.38 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 27.83 fps, 1826.38 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 27.92 fps, 1826.38 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 27.95 fps, 1826.38 kb/s


I wonder how much difference Vista and XP makes.

[PsYkHoTiK]

marginally tightened ram timings. 5-4-4-12 (4 cas no go on 4-4-4-12) 1t. I ran SP32m on it (just under 13minutes) to check of its stable.

Lemme try some other stuff (just want to know if there is any impact).

---------- RUN1PASS1.LOG
encoded 1749 frames, 112.80 fps, 1850.89 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 111.22 fps, 1850.89 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 113.25 fps, 1850.89 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 109.27 fps, 1850.89 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 113.71 fps, 1850.89 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 27.68 fps, 1826.38 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 27.95 fps, 1826.38 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 28.00 fps, 1826.37 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 28.01 fps, 1826.38 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 27.99 fps, 1826.38 kb/s

[graysky]

@PsYkHoTiK - thanks for the results... what O/S?