x264 Benchmark

[PsYkHoTiK]

Originally Posted by graysky

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

Vista Ultimate 32bit.

New validation.

[graysky]

Originally Posted by PsYkHoTiK

Vista Ultimate 32bit.

New validation.

killer man, thanks (added to the table and will update with the hour)

[PsYkHoTiK]

Hey no problem man. I'll keep pushing my stuff and update as needed.

You know what would be real cool? If we could make that table like a pivot table / add filters to the columns so people could come in and interactively check stuff out (i.e: filter out all other CPUs and just look at Kentsfield results). Also, I wonder if you could add a checksum (sorta like superpi) to circumvent cheatzors.

If its do-able, it'd be really awesome.

[graysky]

@PsYkHoTiK - yeah a sortible html would be pretty nice... I think Adrian was going to ask around, I'll have to pm him. About the MD5 idea. I did think of it, but so long as this is batch file driven, that won't matter because someone could easily modify it to generate an md5 for whatever they would want.

I think keeping people's names out of the table might be some what of an incentive not to cheat, but you never can tell.

[PsYkHoTiK]

Hmm good point. Well cpuz validation is another way too I guess. People by large generally aren't lame enough to cheat but like you said, can never tell I guess. By having a list like that, any apparant cheating can be spotted out (it would be an outlier compared to similar spec'd results).



Off topic: Go Colts (grew up in Indianapolis)! Go Seahawks (live up here in Washington)!

[JnZ]

Originally Posted by graysky

I'm not too sure what to make of that; in general, the data are pretty tight......

Thanks for long reply. First of all, I'm pretty sure, that my machine is 100% stable at oc state (1\it runs many hours on Prime95. 2\This machine runs many days without shutdown with HD encodings. 3\2.8GHz oc isn't so high, it's working in windows even at 3GHz, but Prime95 says that isn't stable (mayby litlle increase of CPU voltage helps). )
Second thing. Thing about different FPS, I'm talking about this: I do many bench and every run give different FPS.

Example: (rounded fps)
1) 120/121/119/120/121 - 34/34/34/34/34
2) 104/105/105/104/105 - 33/33/33/33/33
3) 122/123/121/122/122 - 34/34/34/34/34
Then reboot:
4) 98/100/ 96/104/101 - 31/31/31/30/31
5) 115/115/113/116/117 - 33/32/32/33/32
6) 119/120/121/120/120 - 33/33/33/33/32
Then run and close some apps:
7) 109/107/110/108/108 - 34/33/33/34/33
8) 119/120/121/121/120 - 34/34/34/34/34

You see - every run different results.

Then testing number of threads:

+------+--------+-------+--------+-------+--------+-------++--------+--------+
| thds |  1p-1  |  2p-1 |  1p-2  |  2p-2 |  1p-3  |  1p-3 || 1p-avg | 2p-avg |
+------+--------+-------+--------+-------+--------+-------++--------+--------+
| AUTO | 116,84 | 33,99 | 116,72 | 33,65 | 121,15 | 33,74 || 118,24 | 33,79  |
|   1  |  42,38 |  9,27 |  41,89 |  9,58 |  42,14 |  9,53 ||  42,14 |  9,46  |
|   2  |  77,63 | 16,64 |  77,90 | 16,71 |  77,95 | 16,75 ||  77,83 | 16,70  |
|   3  |  86,30 | 20,72 |  85,51 | 20,42 |  85,12 | 20,56 ||  85,64 | 20,57  |
|   4  | 115,76 | 23,45 | 116,97 | 23,64 | 116,00 | 23,44 || 116,24 | 23,51  |
|   5  | 111,82 | 32,05 | 105,80 | 32,30 | 107,84 | 32,49 || 108,49 | 32,28  |
|   6  | 119,85 | 33,49 | 120,49 | 34,19 | 120,75 | 33,68 || 120,36 | 33,79  |
|   7  | 106,91 | 33,36 | 106,61 | 33,73 | 107,94 | 34,16 || 107,15 | 33,75  |
|   8  | 120,62 | 34,95 | 121,67 | 35,07 | 120,75 | 35,20 || 121,01 | 35,07  |
|   9  | 106,30 | 35,36 | 106,30 | 35,34 | 107,02 | 35,61 || 106,54 | 35,44  |
|  10  | 119,98 | 36,07 | 119,38 | 36,06 | 119,98 | 36,21 || 119,78 | 36,11  |
|  11  | 105,80 | 35,43 | 107,22 | 35,81 | 106,41 | 35,64 || 106,48 | 35,63  |
|  12  | 120,88 | 36,24 | 121,53 | 36,18 | 121,67 | 36,20 || 121,36 | 36,21  |
|  13  | 106,50 | 35,90 | 107,43 | 35,81 | 108,04 | 35,93 || 107,32 | 35,88  |
|  14  | 120,88 | 36,13 | 120,49 | 36,32 | 120,49 | 36,25 || 120,62 | 36,23  |
|  15  | 107,53 | 35,76 | 107,94 | 35,85 | 108,67 | 35,91 || 108,05 | 35,84  |
|  16  | 121,53 | 36,17 | 121,67 | 36,23 | 120,88 | 36,19 || 121,36 | 36,20  |
|  17  | 105,90 | 35,81 | 107,94 | 35,88 | 108,05 | 35,89 || 107,30 | 35,86  |
|  18  | 120,23 | 36,16 | 120,23 | 36,21 | 119,98 | 36,25 || 120,15 | 36,21  |
|  19  | 106,40 | 35,81 | 107,22 | 35,84 | 108,57 | 35,88 || 107,40 | 35,84  |
|  20  | 121,27 | 36,16 | 120,23 | 36,27 | 120,88 | 36,24 || 120,79 | 36,22  |
+------+--------+-------+--------+-------+--------+-------++--------+--------+

It seems, that AUTO threading model use 5-6 threads, but for my configuration 12 threads seems best choice. Another remark: Even number of threads give more FPS.

When I encoding HD, CPU work almost at 100% (using AUTO threads) with this bench only 70-90%. The solution is to use 12 or more threads.
Anyway great benchmark, thanks
J.

[rocketmen2k5]

I could never get the damn thing to work so I uninstalled it.

[Team Scream]

Here is my new V8 using SysTool to get a slight OC, 4 gigs of Crucial 667 FB (4x1GB)

Intel Xeon X5355 (x2) Rev B3
Multiplier: x6.0
Bus Speed: 383.1 Mhz
Rated FSB: 1532.2 Mhz
Core Speed 3065.0 Mhz
Chipset: Intel 5000x (Tyan S2696)
Memory Timings 5-5-5-15 @ 333 Mhz
Windows XP Pro SP2

P.S. I also have a dual Opteron 285 System sitting right here that I will run the tests on as soon as my video card gets here, should be Wednesday at the latest, it is running completely stock speeds with 4 gigs of slightly slower ram than the V8 monster listed above.
I also have a single QX6800 based system that I will run the tests on now, it is water cooled and running at 3.4Ghz....results to follow...
Please let me know where I stack up in the list with this V8, it was a pricey bastard to build, and video editing/encoding is what it was built for so I am REALLY interested in knowing what the results actually mean.

---------- RUN1PASS1.LOG
encoded 1749 frames, 150.24 fps, 1849.61 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 151.47 fps, 1849.61 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 150.66 fps, 1849.61 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 151.06 fps, 1849.61 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 151.27 fps, 1849.61 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 75.58 fps, 1834.85 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 75.74 fps, 1834.86 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 75.38 fps, 1834.86 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 75.18 fps, 1834.86 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 74.87 fps, 1834.86 kb/s

[Team Scream]

And here is the QX6800 System, it is quite a bit slower on the 2nd pass, but faster on the first pass? why is that?

Intel QX6800 (Kentsfield)
Multiplier: x13.0
Bus Speed: 266.7 Mhz
Rated FSB: 1066.7 Mhz
Core Speed 3467.2 Mhz
Chipset: Intel i975x (Intel D975XBX2) (Bad Axe 2)
Memory Timings 4-4-4-13 @ 400 Mhz
Windows XP Pro SP2
4 Gigs of Crucial DDR2-800



---------- RUN1PASS1.LOG
encoded 1749 frames, 173.55 fps, 1850.94 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 172.21 fps, 1850.94 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 174.08 fps, 1850.94 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 172.74 fps, 1850.94 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 173.01 fps, 1850.94 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 47.88 fps, 1829.06 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 47.80 fps, 1829.22 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 48.00 fps, 1829.43 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 47.57 fps, 1829.27 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 47.90 fps, 1829.50 kb/s

[graysky]

Originally Posted by Team Scream

Here is my new V8 using SysTool to get a slight OC, 4 gigs of Crucial 667 FB (4x1GB)

Intel Xeon X5355 (x2) Rev B3
Multiplier: x6.0
Bus Speed: 383.1 Mhz
Rated FSB: 1532.2 Mhz
Core Speed 3065.0 Mhz
Chipset: Intel 5000x (Tyan S2696)
Memory Timings 5-5-5-15 @ 333 Mhz
Windows XP Pro SP2

Thanks for the octa results. Also cool to see them.

Are you sure about the 6x383? That's only 2.3 GHz. If you meant 9x383 then you're 3.45 GHz. Also, are you sure about your mem timing @ 333 (667 since DDR2). That ratio seems kinda odd to me 1.74:1.

Originally Posted by Team Scream

And here is the QX6800 System, it is quite a bit slower on the 2nd pass, but faster on the first pass? why is that?

It's a trend with 4x2 = 8 vs. 4x1 = 4 throughout the table. Seems like the 2nd pass does a better job multithreading the workload (perhaps since it's more CPU-intensive than the 1st pass). I honestly don't know