MPEG2 Encoding Time Benchmarks (Comparing PCs)

[Soopafresh]

Source = 1080i Mpeg - 4664 Frames, 400MB Size- 03:06 Running Time

Destination = 720x480 vbr 7500-8500 MPEG2 Video Only

Encoder Used = QuEnc 0.56 Alpha (wanted a command line encoder so I could script the test)

----------AVS File Contents--------------
LoadPlugin("c:\test\DGDecode.dll")

mpeg2source("foo.d2v")

LanczosResize(720, 480)
---------------------------------------------

---------Batch File Contents--------------

quenc.exe -i "q.avs" -o "q.m2v" -b 7500 -maxbitrate 8500 -dc 9 -mpeg2 -nohq -vbr -noscene -notrell -nocgop -interlaced -noextreme -gopsize 15 -maxbframes 2 -aspectratio 16:9 -auto -close

pause

---------------------------------------------


RESULTS

-------------------------------------
2.4Ghz P4 533FSB 512 MB RAM Cost = $350

650 sec to complete = 7.09 FPS
-------------------------------------
Dual 3.0Ghz Xeon 800FSB 4GB RAM Cost = $2500

327 sec to complete = 14.26 FPS
-------------------------------------
Athlon 64 3200 2GB RAM Cost = $ 750

321 sec to complete = 14.52 FPS
-------------------------------------


I know QuEnc isn't Multi Proc Aware, so the numbers would be higher on the Dual Xeon box had I used something like TMPGenc to encode.

Still, the AMD 64 is a great processor for the dough.

[alsyed]

can you run a tmpgenc test on that too? and spreading some tests over, such as 2pass vbr (with various motion search precision settings)?

cheers

and happy new year from NZ!! me already 1.5 hours into 2005!

[Soopafresh]

Yes, I'll run some TMPGenc tests, but how do I accurately measure the time of the encode ? Does it create an encode log ?

[alsyed]

it has a timer built it just above the preview window, and if you make a batch of files, at the end it will spit out an encode log window, with total time taken, and the individual times taken.

cheers!

[BJ_M]

a single P4 3.0 w/ 2GB of ram would have cost the same or less than the A64 and yield exactly the same result as the dual xeon .. maybe faster ..

[thecoalman]

a single P4 3.0 w/ 2GB of ram would have cost the same or less than the A64 and yield exactly the same result as the dual xeon .. maybe faster ..

Am I missing something here? I can encode from AVI>mpeg2 720x480 6000 VBR in about the same length as the video.... using MSP.

p4 3ghz
1 gb ram
7200 8mb Sata Drive

[Soopafresh]

Okey Dokey. I'll post results next week.

[BJ_M]

a single P4 3.0 w/ 2GB of ram would have cost the same or less than the A64 and yield exactly the same result as the dual xeon .. maybe faster ..

Am I missing something here? I can encode from AVI>mpeg2 720x480 6000 VBR in about the same length as the video.... using MSP.

p4 3ghz
1 gb ram
7200 8mb Sata Drive

he was doing resizing and the source was mpeg - i am just saying to compare apples to apples .. not apples to oranges

[thecoalman]

a single P4 3.0 w/ 2GB of ram would have cost the same or less than the A64 and yield exactly the same result as the dual xeon .. maybe faster ..

Am I missing something here? I can encode from AVI>mpeg2 720x480 6000 VBR in about the same length as the video.... using MSP.

p4 3ghz
1 gb ram
7200 8mb Sata Drive

he was doing resizing and the source was mpeg - i am just saying to compare apples to apples .. not apples to oranges

I'll do the apple to the apple ..... Do you have a link to your source file soopafresh?

[Soopafresh]

Let's use this as our source media.

http://trailers.divx.com/Universal/MeetTheFockers_HD.zip

It's not TOO big, and it's a crisp looking clip.


How about the following as the encode settings - ( I modified the NTSC DVD template for VBR 2000 min 7000 ave 8000 max and 10 bit precision)



BTW, here are the first set of results -

Dual Xeon 3Ghz w 4GB RAM TMPGEnc 2.524.63 w Hyperthreading Boxes Checked
-------------------------------------------------------------
C:\test\MeetTheFockers.mpg
Format: DVD NTSC (MPEG-2 720x480 29.97fps VBR 7000kbps)
Video: C:\test\MeetTheFockers.avi
Audio: C:\test\MeetTheFockers.avi

Result: Succeeded
Dealing time: 00:05:11
-------------------------------------------------------------

Ran It Twice, Second Time Score

Result: Succeeded
Dealing time: 00:05:07

[racer-x]

73 MB isn't too big?

Let's be realistic here, not everyone is on broadband. I guess I'll just read everybody else's results.

Unless everybody uses the same encoder with the exact same settings, the results would be virtually useless.

[Soopafresh]

RacerX - Well, we need some source file that will take more than 3 minutes to encode, otherwise the results aren't going to be very accurate. The original media I used for the first series of tests at the top of this thread was 400MB.

[rallynavvie]

If you're going to do the mark then save your TMPGEnc settings file and post it so we all use the same encoder settings. Recommend a good VBR setting too and high motion search.

[thecoalman]

I'm in the middle of being drunk I'll look a it tommroorrrrrrr :P

Let's be realistic here, not everyone is on broadband. I guess I'll just read everybody else's results.
.

It's video ..... it's going to br big.....shall we do a 1 seconcd clip..... BTW I hated dial up....

[BJ_M]

here are two 3 sec. clips i made up from my files....

I shot these on HDCAM and downsized to 720x480 and converted to DiVX archive 4meg/s - or if anyone wants the hd version ...



part1.avi

part2.avi

[rallynavvie]

Nice linking n00b, you still nursing a beer?

[Capmaster]

Nice linking n00b, you still nursing a beer?

Awww, cut 'im some slack. This is the silly season, after all

[BJ_M]

Nice linking n00b, you still nursing a beer?

sorry - i don't get piss ass drunk like 50% of the population (at least not anymore) .

the server didnt like the underscore and didnt link properly -- sometimes underscores work - sometimes not ....

fixing now ....


u nOOb :P

[BJ_M]

in fact i drank nothing yesterday come to think of it ...

[Soopafresh]

Athlon 64 3200 2 GB RAM

08:11 to encode with the above file and settings


P4 2.4 533 FSB 768MB RAM

13:43 to encode with the above file and settings

[rallynavvie]

I'm going to start another thread here shortly with a more standardized benchmark that everyone here should be able to perform easy enough. Not everyone will want to register TMPGEnc in order to use its MPEG2 function (or clean their registry and reinstall it which goes against its license agreement for the MPEG2 functionality).