TmpgEnc Plus 2.54a faster?

[mguru]

Anyone tested 2.54a yet? I am seeing 2xVBR having almost a 50% increase in performance (using High Quality). The version history doesn't mention optimization, so I guess this is a behind-the-scenes kind of thing.

mGuru

[Erik_Osterholm]

I'm not seeing much of a speed increase, maybe you've forgotton some of your normal settings/filters? Or optimizations for your processor in the older version?

Erik

[mguru]

No I didn't forget anything. Just upgraded to the newer version. Nothing changed with system either.

I was doing 120 min 2200avg 2xPass on HighQ in about 63-65 hours with 2.53. Now with 2.54a I am currently doing a 131 min movie with 2350avg 2xPass on HighQ at ~ 54 hours (its 85% complete).

Don't get me wrong, I'm not complaining about it at all, just wondered if anyone else had noticed any increases as well...

I'll doo somemore testing after this encode to check the HighestQ setting.

mGuru

[Uli]

The speed difference depends more on source material. Maybe the movie, you're actually encoding, is much easier to compress because it's very clear or dark or is extremly wide screen (much bigger black borders).

greetz, Uli

[mguru]

doh! Haha, I usually just check the tmpgenc.net version history which doesn't contain that line. Thanks Bach for being so attentive.

Anyways, I like the improvement so far. With my crap computer, its nice not to have to wait 3/4 of a week anymore to encode something. Now if we can get it from 1/2 a week to a day, I'll be rocking.

mGuru

[ninja.rogue]

I have found this dramatical increase in speed.
I agree with Mguru: I am experiencing from 30 to 40% less encoding time, sometimes even 50%.
My pc is a P4 1.9 with 512 rambus Asus th7ii
The first time I thought I had set it to CBR, as it took more or less half the expected time; then, I could verify that it was this option enabled which gave me so much speed.
I am very surprised and amazed, but can confirm this too.
This software probably lacks some further sort of SSE2 optimization, but now it has a comparable speed as CCE - almost.
It would be very interesting if anybody could make a small test - with and without cache enabled.

[mguru]

Well I went back and checked...my speed improvements were *NOT* because of that new cache feature, because it is disabled by default. I am now encoding the SAME movie with it enabled to see if there is an overall reduction in time WITH the new cache feature enabled. I am 65% done so far and it is taking longer with the cache feature enabled than without (about 3 hours more so far). But still 25-30% below pre-2.54a performance levels.

I will post my results tomorrow after this encode finishes. Thanks for responding ninja.rogue.

mGuru

[mguru]

WOW!!!

OK here's the comparison with 2.54a. Clip is Progressive FILM, 131 minutes long encoded SVCD at 2350 2xPass-VBR with High Quality & Pulldown. No Audio.

Pre-2.54a (2.53):

- Approx. 68 hours (didn't encode whole way, only to 10% mark, but from experience this is pretty accurate to within a few hours by this time).

2.54a (Without the new cache feature enabled):

- 53 hours (28.3% increase)

2.54a (With new cache feature set at 4096):

- 46 hours (47.8% increase over pre-2.54a and 15.2% increase over non-cache-enabled 2.54a).


Thus I see 3 interesting points to make out of this:

1. 2.54a, even without the new cache feature, has significant improvements in speed when using the 2xPass-VBR and HQ (I haven't tried HighestQ or non-FILM).

2. The new cache function ONLY seems to start improving performance after the 1st pass; which of course is common sense with what its function is. At 65% on the enabled test, I was still showing 56 hours for total completion. However, by the end it actually ended with only 46 hours.

3. I would attribute a higher initial total completion time with the cache enabled in the 1st pass phase OVER the 1st phase without the cache enabled due to extra processing time (I/O) to write the cache to disk. In my tests, 53 hours was showing in the non-enabled test most of the way through the encode. However, it was showing a range of 54-58 hours total completion time (which is higher) even through 70%.

Please feel free to post your own tests & settings. I would like to know if anyone else can squeeze better results out of 2.54a.

All the best,
mGuru

[destarded]

I can confirm that 2.54 plus is faster than the earlier versions.

I got a decrease from 22 hours to 12! Same vid.

Pretty amazing imo :)

[zaa]

does anyone use the ctrl+alt+delete tip when using tmpgenc under xp?

exit all other programs/processes start tmpgenc, ctrl+alt+del, and set the priority to high and it shaves off at least an hour total.

sorry if this has been posted before.

darren