speeding up tmpgenc

[vandakeg]

Ok, here's what I've done so far:
1. Captured video via firewire
2. Used Adobe After Effects 5.5 (pro bundle) to enhance color and add a
few minor effects.
3. Rendered video from AE as uncompressed .avi
4. Imported to Premiere, editied and added titles etc..
5. I am currently frameserving from Premiere 6.5 to tmpgenc via
avisynth.. the problem is it is taking 35 hours to encode!!

This just got me wondering, is it possible to speed up tmpgenc encodes? I'm assuming that the speed of the encodes is directly related to processor speed and amount of ram. I am running a P4 1.4ghz w/ 1024mb PC800 memory.
Could it be the uncompressed avi that is slowing it down so much? Also, I'm encoding at 8000kb/s CBR at the highest quality (slowest).
I'm thinking about building another system, could I expect a dual Xeon 2.8ghz sytem encode this in nearly realtime? Or is that asking too much? TIA for all the help!

[Craig Tucker]

Adding filters increases the encode times considerably (if you are using any), and you are working with high bitrates, so it will take some time to encode to start with. Processor speed is the biggest factor in determining encoding speed, if you double your processor speed you should expect to almost half your encoding time. TMPGEnc can utilise dual processors, how this effects encoding times I am not sure. But I think you are asking too much to be able to encode 8mbit cbr slowest setting(assuming mpeg2 DVD res)

[vandakeg]

Thanks for the reply, it's very good to know that tmpeg utilizes dual procs, just another reason for me to upgrade to duallies.

[vhelp]

vandakeg,

>> Or is that asking too much?
Yes!

However, you didn't say what version of TMPG you are using. I heard
or read somewheres here that the PRO version allows you to modify the
CACHE for 2pass encodes. I can't remember the exactness of the procedure
but I still can't see TMPG being speed up (in no-PRO versions) in any
sort of tweaking.

But,the best that anyone could advice for you, since you ARE encoding
already, at high bitrate ie, 8000... is to use CQ instead, and use
8000 as your max. You'd be surprise at the quality, AND, you'd be
cutting yourencoding time by haldf already. But, that's not exactly
called SPEEDing up TMPG. Actaully, it's called be smart or something
like that.

But, ceriously, try my idea above, and see for yourself
what quality you get. Again, you may be surpised at the quality you
obtain, AND lesser time to encode (though not exactly SPEEDing up TMPG
in any way)

Filters, as craigtucker said, adds greatly to your encoding
time. If you're using TMPG's "noise", then... the higher the values,
the longer it will take.

Mostly, I've ben searching the reason why a *.d2v file encodes much
faster than an captured *.avi file or Firewired DV avi file, and, about
the only answer I could come up are that:
A, the captured or firewired transfer source is in compressed format,
B, and, "keys" are not placed the same way in *.avi files are they
... may be, in DVD MPEG-2 files, or something like that.
But, even though a source could be from a non-compressed capture,
(not DV, since DV IS compressed, and IS transfered and not captured)
the speed of encoding from such sources is still the same - slow, or
slower than DVD MPEG-2 sources ie, *.d2v
So, my guess is that it's something to do w/ "key's" or "indexes"
or something in that direction.
And, your above method/process of feeding into Adobe/Premier make only
add to the indexing or what-have-you of the *.avi source.

Hmmm..., in reference to above, a good test of this theory would be to
do a series of divX encodes, but w/ various "key" settings... And,
see if TMPG encodes those series at the same speed or not. hmmm.

-vhelp

[vandakeg]

Yeah, I'll give CQ a try, thanks for your input!

[tuco1979]

If you are using tmpgenc filters, why don't you use avisynth filters instead? That would quicken the process a decent amount and should be just as good (maybe better) than tmpgenc filters. I don't if premiere is slowing things down or not since I don't use that program. To figure out what filter/program that is giving the most speed hit, try encoding small sections of video without filters or framesevering, with framsevering no filters and with everything. Then fool around with avisynth filters and see if that helps.

[Morheus]

I hope nobody blasts me for this ... but I cant see any difference past 5000 bitrate
I found big differences between the same file encoded at 2500 and then 3500 but no real difference going from 5000 to 7000 ...

my AMD 700 processor with 512 ram gives me a fully compliant mpeg2 dvd at 5000 bitrate in about 22 hours ...
I think staying away from filters and bringing the bitrate down could cut your encoding time by 20-30%


Tony

[Craig Tucker]

my AMD 700 processor with 512 ram gives me a fully compliant mpeg2 dvd at 5000 bitrate in about 22 hours ...
Tony

Well, that could be really fast, or extremely slow, depending on the length of your video file. :P

[Crystal Symo]

Why don't you try Cinema Craft Encoder instead? It's faster than TMPGENC and gives you a very good quality.

[THE RIPPER]

With a dual setup you can expect a 33% increase on encode time with tmpge. For reference download the sample movie clip from the tmpge site, encode it with standard vcd template with default settings, then switch to dual for your increase. My system encodes the 1 minute clip with one cpu in 39 sec, with dual 18 sec .

[ralexander41]

Guys, I purchased an ASUS A7V333 motherboard with raid, an Athlon XP 1600 processor $54, 256M PC2100 memory and a Western Digital 120G, 8 megabyte buffer drive (very fast). I encoded using TNPGENC a 22G AVI file, at 9500 CBR and it took 14 hours. That's quite good for a $54 processor. Roger

[Craig Tucker]

Why don't you try Cinema Craft Encoder instead? It's faster than TMPGENC and gives you a very good quality.

And a lot more expensive

[fingernailX]

For even more expense use a Targa 3000 and Pinnacle MpegWorks. We have a dual Xeon and 360 gb raid setup but we can export uncpmpressed avi to DVD template Mpeg 2 in real time!!!!
http://www.pinnaclesys.com/ProductPage.asp?Product_ID=91&Langue_ID=7

[vandakeg]

Thanks for the help... I must not have been very clear before, but the effects I added were in AE, not TMPGenc.
I am very interested in hardware solutions for faster encodes, but not at the targa 3000 price tag. If I'm going to spend that kind of money, I'll put it towards a completely new NLE system.... which I'm doing anyway!
That brings me to my next question... I bleed Intel through and through, and never really considered building a dual AMD system... however, the price difference is almost too inviting to ignore. Would I regret going w/ AMD duallies?? I've heard they heat up and have a tendency to be unstable with processor intensive tasks such as encoding.
Like I said, I am Intel to my very core, and am very unwilling to make the change. I know being an Intel fan is not so much something to be proud of, but I am, it's me... it just makes me feel good knowing that I've got Intel inside.... that being said, can I be happy w/ a dual AMD system? The price would make me happy, but how about the performance?

[THE RIPPER]

Well i've been very happy with my amd dualie, my temps never get above 42 at full load. Check out www.2cpu.com

[MrKGB]

I read some posts here recently which were saying that Athlon, especially XP is much faster than same Intel.
Which may be not too surprising if imagine that developer of TMPG has XP system. Of course he would optimize his soft for Athlon's 3d Now Pro, instead of Intel's SSEII.
One more suggestion - don't use the slowest motion search, use the slow, you will never see difference, especially at 8000.

[Peka]

Well, it's been proven in numerous tests with CPU's of the same speed (GHz) - AMD are faster than Intel for the same GHz. There may be plenty of reasons to still get Intel, but if you want bang for ya buck, AMD is the way to go.

Yes they can get hot, but only if you don't keep them cool - there's plenty of sites with info on keeping AMD's cool, as well as overclocking them.