Tmpgenc 2.0 --- some observations

[dsa44]

Well, I did a real quick test between 12 beta (original) and 2.0 I encoded the same 1000 frame (42 second) clip with same settings and this is what I found. (P4 1.

12 beta 5:18 min
2.0 sse on 4:40 min
2.0 sse off 4:57 min

Also the file size was much smaller with 2.0 (6539 kb vs 7077 kb) That means I could bump up my bitrate a bit!
And, for the most important part.....the quality looked quite similar. I could not tell the clips apart! In the past, I had done a quality test for myself which showed 12 beta looked better than the later versions. (I think it was e or f)
Well, just my $.02. I'll be using 2.0 for now on.
Cheers.

[Yeshi]

did you have IDENTICAL settings? because i did a test with version 12a and 2.0 and the file size was identical, make sure EVERYTHING is the same, including quantize matrix, ESPECIALLY gop structure, and motion search precision, it all makes a difference.

[Yeshi]

I also didnt notice any quality difference. 2.0 however was definetaly faster

[dsa44]

All of the settings were identical...as far as I can tell. The bitrate, quality, audio, motion, etc. were exactly the same. I didn't even look at the gop structure and used whatever defaults pop up. I am assumming these are the same between the programs. BTW, I did not use 12a...it was 12 beta original.

[Vejita-sama]

Hmm? The file size should be 100% based on the bitrate. Did you encode CBR for both? Cause if you used VBR you can't really compare/predict file sizes.

[dsa44]

Interesting. The bitrate was identical, as was the audio encode. Both CQ max bitrate 1950, quality 65, audio 128.

I had done an earlier compare between 12beta original and some of the 12x versions, and found that with the exact same encodes, 12beta orig had the smaller file size. Now I am finding this true with 2.0. Maybe it has to do with how efficient the versions are with the mpeg2 encode, or how it is implemented.

[vitualis]

That explains the size difference... You are using a VBR encoding method. Numbers such as "quality 65" are simply arbitrary and doesn't actually mean anything other than a guide for the user. There is no way to tell (without looking at the source code or doing extensive testing) that "quality 65" is the same or gives the same quality across all versions of TMPGEnc.

Regards.

[Gordian]

I have had very interesting results.
I have tried to encode a 54' AVI only video file using CQ 80, max bitrate 8500, min 4500, normal quality for m.e.
I got these results - using PIV 1.9:
SSE-2 enabled: 3 hours 20'
SSE-2 disabled: 4 hours 40'

For a first release, I would consider this amazing.
I am going to test quality, but have no reason to doubt about it being the same.
Paolo

[Yeshi]

hmm im going to do some testing of my own using version 12a, 12j and 2.0

[Yeshi]

OK here are my PC Specs

celeron 400
192 ram
win 98se

i encoded a file of length 100 frames with 1150CBR, here are the results with time and size.

VERSION 2.0 - 736kb/18sec
VERSION 12.a - 711kb/14sec
VERSION 12.j - 736kb/14sec

i suppose you cannot really go by these results because the movies were so small.

[Horrorking]

<TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font size=-1>Quote:</font><HR size=1 color=black></TD></TR><TR><TD><FONT SIZE=-1><BLOCKQUOTE>
On 2001-10-16 05:25:59, Yeshi wrote:
i encoded a file of length 100 frames with 1150CBR, here are the results with time and size.

VERSION 2.0 - 736kb/18sec
VERSION 12.a - 711kb/14sec
VERSION 12.j - 736kb/14sec

i suppose you cannot really go by these results because the movies were so small.
</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR size=1 color=black></TD></TR></TABLE>

I agree that with just 100 frames it is too small to really go by, but I would like to see a similar test done with a 2 or 3 minute source file.

Heck I have a Celeron 500, Win ME, 7200rpm HDD & 256MB ram, so I might do a test when my system is finished the current conversion of The Mummy Returns...

I may also do the same on an AMD K6-2-400 as well...

David
http://www.horrorking.com