AMD X2 / Pentium D & AutoGK DivX Encoding

[Technonut]

I am curious about how well the X2 and Pentium D is doing with Auto Gordian Knot 1.96.... Download Link EDIT: 1-29-2006: Starting now, please download and run the latest AGK version 2.26 from the link above.

I tried to choose a DVD that many would have, and decided upon Lord Of The Rings: The Fellowship Of The Ring.

To have consistant results, I am using the AutoGK setup used Here.

It would be great if any of you X2 and Pentium D users out there would run an AutoGK 2.26 / DivX 6.1.1 VFW codec (codec only) encode, and post your log file here to compare times. AutoGK can also be considered stability testing... I know that the actual encoding really works my processor even with VapoChill cooling. (But then again, I have a Prescott. )

Anyhow, if you happen to have the LOTR DVD, and want to post results, please use the 2 CD's (1400Mb) setting, and the rest found in the above linked guide.

Here are my results (P4 640@4.4GHz) for comparison: (Edited for space)

[8/20/2005 5:10:39 PM] Target size: 1400Mb
[8/20/2005 5:10:39 PM] Started encoding.
[8/20/2005 5:10:39 PM] Demuxing and indexing.
[8/20/2005 5:13:12 PM] Running compressibility test.
[8/20/2005 5:13:12 PM] Writing the following script to E:\lotr fow divx\agk_tmp\lotr for_comptest.avs

[8/20/2005 5:17:09 PM] Duration was: 3 minutes 57 seconds
[8/20/2005 5:17:09 PM] Speed was: 54.04 fps.
[8/20/2005 5:17:09 PM] Compressibility percentage is: 32.53
[8/20/2005 5:17:09 PM] Using softer resizer.
[8/20/2005 5:17:09 PM] Chosen resolution is: 528x224 ( AR: 2.36 )
[8/20/2005 5:17:09 PM] Predicted comptest value is: 63.15
[8/20/2005 5:17:09 PM] Running first pass.
[8/20/2005 5:17:09 PM] Writing the following script to E:\lotr fow divx\agk_tmp\lotr for_movie.avs

[8/20/2005 5:57:49 PM] Duration was: 40 minutes 39 seconds
[8/20/2005 5:57:49 PM] Speed was: 105.16 fps.
[8/20/2005 5:57:49 PM] Running second pass.
[8/20/2005 6:36:29 PM] Duration was: 38 minutes 40 seconds
[8/20/2005 6:36:29 PM] Speed was: 110.58 fps.
[8/20/2005 6:36:30 PM] Splitting video into: E:\lotr fow divx\lotr for.cd1.avi
[8/20/2005 6:37:27 PM] Splitting video into: E:\lotr fow divx\lotr for.cd2.avi
[8/20/2005 6:38:19 PM] Job finished. Total time: 1 hour, 27 minutes 40 seconds



EDIT: 8/25/05: Some Pentium D results would be nice to see also...

 

[RichUK]

this test will only stress one of the cores, and will be the same as any equivalent Venice or San Diego core.. except a little faster because it will have a dedicated core to one task, unlike a single core having run the OS and its services etc ..

 

[IeraseU]

You will get the major benefit of having your computer accessible to do other tasks while encoding. That is what I hated about single core encoding.....not the time it takes (although I do think a DC will be faster in sheer time also), but how long I would be stuck with my computer 'locked up' unable to do anything other then browse/email while the process took place. Some of my more exotic encoding jobs have taken up to 12hrs, so if you do this type of thing frequently, certainly a dual core would help with the frustration level immensely.

 

[Technonut]

Originally posted by: RichUK
this test will only stress one of the cores, and will be the same as any equivalent Venice or San Diego core.. except a little faster because it will have a dedicated core to one task, unlike a single core having run the OS and its services etc ..

Really? It seems to make use of Hyper-Threading during the actual video encoding... My processor usage was pretty much 100% during the DivX video encoding compared to around 50-75% for the audio encoding portion....


EDIT: Exuse me... I noticed the drop in processor usage during audio encoding using AutoGK with XviD, not DivX... Still, AutoGK is pretty much @ 100% processor usage during DivX encoding, so it appears that it uses Hyper-Threading. Would it not use both cores in the X2?

 

[PetNorth]

Meanwhile some guy do the test, you can calculate some numbers taking anandtech benches with AGK 1.60 @ DivX 5.2.1:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2388&p=9

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=6

Pentium D 840 3.2GHz: 73.2 fps
Pentium 4 EE 3.73GHz: 57.9 fps
Athlon 64 X2 4800+ 2.4GHz: 84 fps
Athlon 64 X2 4200+ 2.2Hz: 75.2 fps

Your oced P4 4.4GHz, would be about 17% faster than P4 EE 3.73. In anand bench it would obtain about 67.74 fps


So, A64 X2 4800+ would be about 24% faster than your oced P4 4.4ghz
Or X2 4200+ about 11% faster.

(If I'm not in some mistake )

 

[coomar]

i thought amd encodes xvid faster than intel, the opposite case for divx

 

[RichUK]

 

[Technonut]

Originally posted by: PetNorth
Meanwhile some guy do the test, you can calculate some numbers taking anandtech benches with AGK 1.60 @ DivX 5.2.1:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2388&p=9

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=6

Pentium D 840 3.2GHz: 73.2 fps
Pentium 4 EE 3.73GHz: 57.9 fps
Athlon 64 X2 4800+ 2.4GHz: 84 fps

So, A64 X2 4800+ is about 45% faster than P4 EE 3.73

Your oced P4 4.4GHz, would be about 17% faster than P4 EE 3.73

So, A64 X2 4800+ or 4600+ 2.4ghz would be about 28% faster than your oced P4 4.4ghz; X2 4400+ or 4200+ 2.2GHz would be about 20% faster, or X2 3800+ 2.0ghz would be about 11% faster than your oced P4 (if I'm not in some mistake )

Thanks for the links.... I would still like to see the results of a complete LOTR DVD-DivX encode using AutoGK.... From the X2 review that you linked:

Armed with the DivX 5.2.1 and the AutoGK, we took all of the processors to task at encoding a chapter out of "Pirates of the Caribbean". We set AutoGK to give us 75% quality of the original DVD rip and did not encode audio.

 

[PetNorth]

I've edited my previous post. I've made some more number and I think now is more accurate

 

[n7]

I'm too lazy to want to try this on my Venice, though i'd be happy to see someone else's results

 

[Technonut]

Originally posted by: PetNorth
I've edited my previous post. I've made some more number and I think now is more accurate

Pentium D 840 3.2GHz: 73.2 fps
Pentium 4 EE 3.73GHz: 57.9 fps
Athlon 64 X2 4800+ 2.4GHz: 84 fps
Athlon 64 X2 4200+ 2.2Hz: 75.2 fps

Your oced P4 4.4GHz, would be about 17% faster than P4 EE 3.73. In anand bench it would obtain about 67.74 fps

So, A64 X2 4800+ would be about 24% faster than your oced P4 4.4ghz
Or X2 4200+ about 11% faster.

Hmmm... My FPS in the complete AutoGK-DivX LOTR encoding came out to:

First Pass: 105.16 fps

Second Pass: 110.58 fps

Like I said, I'm just curious.... Hopefully someone will posts results..... Perhaps Duvie ?

 

[PetNorth]

Of course, it's only a reference.

Different AGK version, different settings, perhaps different DivX version. BTW what DivX version have you used? this is important to obtain accurate results if someone do the test.

 

[Technonut]

BTW what DivX version have you used?

5.2.1

 

[Duvie]

I dont have that movie!!! I have the other to LOTR movies though....

I dont have that version or that older codec....I would I instll the older codec when I have the newer one short of I guess removing all the divx codecs....

 

[Technonut]

Originally posted by: Duvie
I dont have that movie!!! I have the other to LOTR movies though....

I dont have that version or that older codec....I would I instll the older codec when I have the newer one short of I guess removing all the divx codecs....

I have been doing some research, and gather that the earlier DivX codecs do not support dual core or dual processors. I ran across This Info regarding the DivX Helium codec.....

DivX Helium codec demonstrates our latest research into processor optimization, achieving encoding times up to twice as fast as DivX 6 in the highest quality (Insane) mode when run on dual core, HT and SMP CPUs.

Please note that this release takes advantage of multi-thread capable systems, like SMP multi-processors, HyperThreaded and Dual Core processors. You will only see marginal improvement with single-thread systems.

I will have to look into this further....

 

[Markfw]

Well, I tried, and I have the movie, but I couldn't get the software to work, kept getting errors (probably user error, but....) Last it said was completed in 3 minutes 15 seconds, and that can't be rght.

 

[stevty2889]

I've got the movie as well, I'll give it a try when I get the chance.

 

[Bona Fide]

Originally posted by: Technonut

Originally posted by: RichUK
this test will only stress one of the cores, and will be the same as any equivalent Venice or San Diego core.. except a little faster because it will have a dedicated core to one task, unlike a single core having run the OS and its services etc ..

Really? It seems to make use of Hyper-Threading during the actual video encoding... My processor usage was pretty much 100% during the DivX video encoding compared to around 50-75% for the audio encoding portion....


EDIT: Exuse me... I noticed the drop in processor usage during audio encoding using AutoGK with XviD, not DivX... Still, AutoGK is pretty much @ 100% processor usage during DivX encoding, so it appears that it uses Hyper-Threading. Would it not use both cores in the X2?

1. Hyper-threading is the Intel "pseudo" dual-core effect. It makes the OS believe that there is a second core, but it is only a virtual piece of a single physical core.

2. Use Task Manager to see if BOTH cores are at 100% usage. My guess is that CPU 0 (Core 1) is at full stress and the second one is idling.

 

[Technonut]

1. Hyper-threading is the Intel "pseudo" dual-core effect. It makes the OS believe that there is a second core, but it is only a virtual piece of a single physical core.

2. Use Task Manager to see if BOTH cores are at 100% usage. My guess is that CPU 0 (Core 1) is at full stress and the second one is idling.

1. I know how Hyper-Threading works....

2. I did use Task Manager to determine that both cores were at 100% usage in my testing.....

It is the codec that determines whether HT, SMP or dual core is utilized, not AutoGK. From the AutoGK FAQ:

5.1 Do you support Hyperthreaded/dual CPUs ?

yes, AutoGK has no problems running on those CPUs (although encoding is NOT done by AutoGK directly, so the whole process may not run faster on such CPUs - codecs have to support that natively).

It does appear that DivX 5.0 and up is optimized for dual processors... Link

The combination of DivX 5.0 software and the newest modern CPUs makes Double Real Time encoding speeds at full-screen (4CIF) resolutions a reality. We've not only met the holy grail of real-time encoding, we've blown right past it. On properly configured dual-processor machines, it's even possible to achieve up to Triple Real Time (70+ fps) encoding speeds for full-screen video. (And no, we're not kidding.)

Markfw900: Thanks for the effort... I wonder what the issue is? Did you use the setup guide that I linked?

stevty2889: Thanks... I look forward to your results.

I will go through my DVDs and list some other popular titles that others may have to test with when I have the time...


EDIT: I ran through it again just to to be sure, and the results were consistant....

[8/20/2005 11:28:44 PM] Target bitrate is: 643kbps
[8/20/2005 11:28:44 PM] Running compressibility test.
[8/20/2005 11:32:41 PM] Duration was: 3 minutes 57 seconds
[8/20/2005 11:32:41 PM] Speed was: 54.09 fps.
[8/20/2005 11:32:41 PM] Compressibility percentage is: 32.57
[8/20/2005 11:32:41 PM] Using softer resizer.
[8/20/2005 11:32:41 PM] Chosen resolution is: 528x224 ( AR: 2.36 )
[8/20/2005 11:32:41 PM] Predicted comptest value is: 63.24
[8/20/2005 11:32:41 PM] Running first pass.
[8/21/2005 12:13:19 AM] Duration was: 40 minutes 37 seconds
[8/21/2005 12:13:19 AM] Speed was: 105.24 fps.
[8/21/2005 12:13:20 AM] Running second pass.
[8/21/2005 12:51:53 AM] Duration was: 38 minutes 33 seconds
[8/21/2005 12:51:53 AM] Speed was: 110.89 fps.
[8/21/2005 12:51:54 AM] Splitting video into: E:\lotr fow divx\lotr for.cd1.avi
[8/21/2005 12:52:50 AM] Splitting video into: E:\lotr fow divx\lotr for.cd2.avi
[8/21/2005 12:53:43 AM] Job finished. Total time: 1 hour, 27 minutes 31 seconds

 

[Markfw]

I can verify that it uses both cores. When I was running it, it got up to 80-90% usage. 50% in task manager is only one core at 50%, so it really does use them both.
Technonut, I looked over the install guide, and thought I had it, but when I went to open the source, it wouldn;t let me select a disk or a folder, only one file, so my 3 minutes was only one file of the movie.

 

[stevty2889]

Defintaly uses both cores. Here's what I got.

[8/21/2005 11:39:49 AM] Audio: English
[8/21/2005 11:39:49 AM] Subtitles: none
[8/21/2005 11:39:49 AM] Codec: DivX
[8/21/2005 11:39:49 AM] Target size: 1400Mb
[8/21/2005 11:39:49 AM] Started encoding.
[8/21/2005 11:39:49 AM] Demuxing and indexing.
[8/21/2005 11:42:10 AM] Target bitrate is: 643kbps
[8/21/2005 11:42:10 AM] Running compressibility test.
[8/21/2005 11:46:35 AM] Duration was: 4 minutes 24 seconds
[8/21/2005 11:46:35 AM] Speed was: 48.53 fps.
[8/21/2005 11:46:35 AM] Compressibility percentage is: 102.19
[8/21/2005 11:46:35 AM] Chosen resolution is: 720x320 ( AR: 2.25 )
[8/21/2005 11:46:35 AM] Predicted comptest value is: 94.92
[8/21/2005 11:46:35 AM] Running first pass
[8/21/2005 1:14:31 PM] Duration was: 1 hour, 27 minutes 56 seconds
[8/21/2005 1:14:31 PM] Speed was: 48.63 fps.
[8/21/2005 1:14:31 PM] Running second pass.
[8/21/2005 1:38:42 PM] Duration was: 24 minutes 10 seconds
[8/21/2005 1:38:42 PM] Speed was: 176.94 fps.
[8/21/2005 1:38:42 PM] Error splitting file
[8/21/2005 1:38:42 PM] Job finished. Total time: 1 hour, 58 minutes 52 seconds

 

[PetNorth]

Something wrong there, stevty2889

1 hour 27min first past (a lot) and very few fps average; and only 24 min 10 secs, second pass with a lot of fps average.. non sense

 

[stevty2889]

Yeah, I thought something looked wrong myself. Re-running it now, first pass finished at 42m 8s this time, 101.48FPS, second pass is almost done. Looks like I was compressing at a differant resolution than him the first time as well(720x320, rather than 528x224). This is my first time running this program, so I probably didn't set something up correctly.

 

[stevty2889]

This is what I got the second time I ran it. Something still seems wrong, this time the second pass was a lot longer than the first. This is the first time I've used this program so I could be doing something wrong.

[8/21/2005 4:25:03 PM] AutoGK 1.96
[8/21/2005 4:25:03 PM] Job started.
[8/21/2005 4:25:03 PM] Input dir: C:\FELLOWSHIP\VIDEO_TS
[8/21/2005 4:25:03 PM] Output file: C:\FELLOWSHIP\VIDEO_TS.avi
[8/21/2005 4:25:03 PM] Audio: English
[8/21/2005 4:25:03 PM] Subtitles: none
[8/21/2005 4:25:03 PM] Codec: DivX
[8/21/2005 4:25:03 PM] Target size: 1400Mb
[8/21/2005 4:25:03 PM] Started encoding.
[8/21/2005 4:25:03 PM] Demuxing and indexing.
[8/21/2005 4:27:18 PM] Processing file:
[8/21/2005 4:27:18 PM] Target bitrate is: 643kbps
[8/21/2005 4:27:18 PM] Running compressibility test.
[8/21/2005 4:30:11 PM] Duration was: 2 minutes 52 seconds
[8/21/2005 4:30:11 PM] Speed was: 74.24 fps.
[8/21/2005 4:30:11 PM] Compressibility percentage is: 99.44
[8/21/2005 4:30:11 PM] Switching b-frames off
[8/21/2005 4:30:11 PM] Chosen resolution is: 528x224 ( AR: 2.36 )
[8/21/2005 4:30:11 PM] Predicted comptest value is: 84.32
[8/21/2005 4:30:11 PM] Running first pass.
[8/21/2005 5:12:20 PM] Duration was: 42 minutes 8 seconds
[8/21/2005 5:12:20 PM] Speed was: 101.48 fps.
[8/21/2005 5:12:20 PM] Target bitrate is: 834kbps
[8/21/2005 5:12:20 PM] Running second pass.
[8/21/2005 6:07:33 PM] Duration was: 55 minutes 12 seconds
[8/21/2005 6:07:33 PM] Speed was: 59.88 fps.
[8/21/2005 6:07:34 PM] Splitting video into: C:\FELLOWSHIP\VIDEO_TS.cd1.avi
[8/21/2005 6:08:15 PM] Splitting video into: C:\FELLOWSHIP\VIDEO_TS.cd2.avi
[8/21/2005 6:08:55 PM] Splitting video into: C:\FELLOWSHIP\VIDEO_TS.cd3.avi
[8/21/2005 6:09:05 PM] Job finished. Total time: 1 hour, 44 minutes 2 seconds

 

[Duvie]

From what I may have noticed looking at some sites about this the first compressibility test may not be using both cores but the actual encoding which is done in the 2nd pass is where it will use it the most...So perhaps not a big increase in speed initially as seen by Techno's 4.4ghz being faster in the first pass....I think you will be faster in the 2nd pass and overall....