x264 video encoding benchmark

[graysky]

I put together a self-contained x264 video encoding benchmark. Techarp kindly agreed to host the file and results at this URL.

Basically, you run the test encode and it will report back frames-per-second values for your machine @ it's clock/overclock level. You can run it at your stock settings and at your overclock settings to see how your machine compares to others in the database.

The database is small right now (as of 08-sep), but as you guys report in results, I will populate it. My goal is to have a representative set of data for many different chips and chipsets. Hopefully, we'll get some Penryn and Phenom data when they become available. Also, if anyone out here has some of the high end AMD chips, please contribute. Instructions and the file are at that url.

Also, please report your results here in this thread. I will keep the data at that url to keep things simple.

Thanks all.

[dunn2953]

cool idea ill have to try it out on my media rig

[Hachi_Roku]

shouldn't this be in benchmark section? especially if you want Phenom results

[graysky]

Dammit, I actually thought I posted to that section; maybe a mod can move this for me?

[graysky]

As of 20-Sep-2007, we have data on over 100 Intel-based systems and on over 40 AMD-based systems. There are a few trends I picked-up on while browsing through the database. I put them into a single table and color coded them to make them easier to see. If you see a trend I missed, lemme know and I'll add it to the table.

Request: we don't have a single example of a machine that has both WinXP and WinVista on it. If you have a dual-boot setup, it would be cool to see the difference the O/S makes. Another missing trend is a 32-bit O/S vs. the same O/S that's 64-bit.

On to the table:



Yellow: Nearly 1:1 increase by adding an additional processor to a dual-chip MB; Comparison of a Q9650 ES Yorkfield (45 nm quad) core chip to a Kentsfield (65 nm quad) core chip using the same RAM, MB, O/S etc. shows about 6 % boost and that's without using the SSE4 instructions. Two overclocked chips (e6600/q6600) both @ the same clock rate, you see a nearly 2x speed up out of the quad.
Orange: Some operating systems seem to handle x264 more efficiently than others.
Red: Insignificant or slight gains realized by upping the DRAM speed by 50 %.
Blue: For the most part, these chips scale in a pretty linear fashion.
Green: Tighter/looser memory timings have a pretty slight affect on performance.
Purple: Keeping the same over-all clock speed using a different combo of multiplier and FSB can give pretty insignificant gains.

Again, I only gave this a once-over look; please point out any trends you see that I missed and also don't forgot about the O/S request!

Thanks again to all who contributed!

[Ace-a-Rue]

here are my results from XP Pro SP2


---------- RUN1PASS1.LOG
encoded 1749 frames, 192.98 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 193.67 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 194.33 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 194.01 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 193.99 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 54.44 fps, 1829.10 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 54.31 fps, 1829.07 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 54.21 fps, 1829.42 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 54.36 fps, 1829.27 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 54.39 fps, 1829.14 kb/s

[njkid32]

Here are my 2..

Running on XP Pro SP2

---------- RUN1PASS1.LOG
encoded 1749 frames, 174.10 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 174.36 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 45.78 fps, 1829.44 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 45.74 fps, 1829.13 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 45.86 fps, 1829.18 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 45.84 fps, 1829.00 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 45.82 fps, 1829.32 kb/s





~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Running on Vista x64 SP1

---------- RUN1PASS1.LOG
encoded 1749 frames, 177.96 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 178.80 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 177.11 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 177.11 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 177.96 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 49.09 fps, 1829.48 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 48.90 fps, 1829.06 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 48.92 fps, 1829.30 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 49.01 fps, 1829.48 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 48.90 fps, 1829.10 kb/s

[Vapor]

A nice 7.9-9.4% gain for Yorkfield over Kentsfield at 3GHz

[Ace-a-Rue]

Here are my 2..

Running on XP Pro SP2

---------- RUN1PASS1.LOG
encoded 1749 frames, 174.10 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 174.36 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 174.34 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 45.78 fps, 1829.44 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 45.74 fps, 1829.13 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 45.86 fps, 1829.18 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 45.84 fps, 1829.00 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 45.82 fps, 1829.32 kb/s





~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

i thought this would be a good comparison between same OS, at the same cpu speed and memory speed to show the real time differences between Kentsfield and Yorkfield runs.

11.5% difference between pass 1...

7.2% difference between pass 2...

significant results, i would say for video encoding!

GO YORKFIELD!

[Hachi_Roku]

Q6600 433x9 3900
1:1 4-3-3-10
Vista x64

---------- RUN1PASS1.LOG
encoded 1749 frames, 192.64 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 190.03 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 193.64 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 193.64 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 193.99 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 54.16 fps, 1829.43 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 54.56 fps, 1829.11 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 54.90 fps, 1829.53 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 54.82 fps, 1829.40 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 54.66 fps, 1829.28 kb/s

[graysky]

Thanks for the results, guys. Will update the table soon.

@ace-a-rue: any idea what the chip name is on the ES Yorkfield? Also, any chance you can do the head-to-head comparison with it and a Kentfield (i.e. same board, same clock settings, same physical memory/settings, same O/S, etc.)?

[Ace-a-Rue]

that is not my yorkfield; it is Njkid32..i just borrowed his image to show the difference at the same cpu speed and memory...he probably got the yorkfield from intel to evaluate and thus, engineering sample (ES).

[njkid32]

[Ace-a-Rue]

can't hear the tune...LOL

[graysky]

that is not my yorkfield; it is Njkid32..i just borrowed his image to show the difference at the same cpu speed and memory...he probably got the yorkfield from intel to evaluate and thus, engineering sample (ES).

My mistake.

@njkid32: are you able to do the head-to-head w/ the two chips?

[njkid32]

Yes, I'll test both the Q6600 and the York at the same speed and multi's on the same machine I'll just swap cpu's.

[njkid32]

Same board, memory and OS..

Q6600-

---------- RUN1PASS1.LOG
encoded 1749 frames, 160.84 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 164.40 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 163.67 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 161.56 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 164.40 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 42.96 fps, 1829.22 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 42.99 fps, 1829.41 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 42.78 fps, 1829.23 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 42.50 fps, 1829.46 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 42.76 fps, 1829.42 kb/s



Yorkfield-

---------- RUN1PASS1.LOG
encoded 1749 frames, 171.42 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 171.71 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 171.96 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 171.96 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 169.61 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 45.65 fps, 1829.18 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 45.78 fps, 1829.17 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 45.78 fps, 1829.51 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 45.76 fps, 1829.21 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 45.72 fps, 1829.12 kb/s

[Ace-a-Rue]

so...it is lower than what i came up with (board difference)...your test figures with the same board, memory and at the same speed, i come up, after averaging the numbers and then dividing:

PASS 1: 5.1%

PASS 2: 6.8%

not as good as i first figured but still a nice increase without taking advantage of the SSE4 set of registers.

[njkid32]

Ace, I'm wondering if Vista x64 is a result of slower results

[Ace-a-Rue]

Ace, I'm wondering if Vista x64 is a result of slower results

vista is like a wrench in the spokes! ...most likely the reason...i installed 32 bit vista and 64 bit and i swear 32 bit feels far more peppier...have you experienced that?

[njkid32]

The only reason I'm using x64 is for crunching it does quite well. I have not tried x86 vista yet though.

[Ace-a-Rue]

it just seems quicker in the 32 bit version although benches would be about the same in results.

[STEvil]

2x Woodcrest 5120 (266mhz stock, BSEL modded to 333mhz) @ 7x333 (2335mhz)
4x 1GB FB-DIMM's @ 333mhz 5-5-5-15 (stock)
Supermicro X7DAL-E


---------- RUN1PASS1.LOG
encoded 1749 frames, 117.33 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 117.09 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 117.34 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 117.08 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 116.96 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 32.38 fps, 1829.41 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 32.38 fps, 1829.53 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 32.43 fps, 1829.18 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 32.31 fps, 1829.00 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 32.45 fps, 1829.34 kb/s


2x Woodcrest 5120 (266mhz stock, BSEL modded to 333mhz) @ 7x386 (2706mhz)
4x 1GB FB-DIMM's @ 386mhz 5-5-5-15
Supermicro X7DAL-E



---------- RUN1PASS1.LOG
encoded 1749 frames, 134.86 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 135.03 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 135.51 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 135.51 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 135.68 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 37.46 fps, 1829.49 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 37.55 fps, 1829.41 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 37.54 fps, 1829.54 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 37.47 fps, 1829.32 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 37.57 fps, 1829.09 kb/s

[graysky]

@njkid32 - very cool man. thanks for taking the time to do this. Looks like a change of 6.3 % when you compare the total encoding time of the two. I'll update the tables shortly.

@STE - thanks for the data.. which o/s?

[STEvil]

XP-64, sorry