Slow encoding times with DivX 5.1

[TotalChaos]

I'm not sure if I even know what I am talking about, but I have been ripping and compressing DVD videos into DivX since v4.11. Through out this time I've only used the latest gordian knot and the acompanying codec pack or latest divx codec. As any who compress full movies knows the proccess takes a long time. As a being of NO patience I created and finaly finished my latest toy. I used a P4C 3.0GHz with HT and over clocked it a tad to 3.25GHz 868FSB. It would seem to me sufficient to kill these long compression times. With the DivX 5.05 and gordian knot I could compress a full length movie doing two passes at 900Kbps in about 1-hour 45-minutes. Now with this new POS (Divx 5.1) My encode times are like 6+ hours. I'm useing the EXACT same set up as I always have, and YES I did use the slowest setting for DivX 5.05 and am now using the standard setting for this new DivX 5.1. Can anyone tell me WHY this thing is so slow. I have ZERO patience for this!!!

[Tuning]

Hi TotalChaos,

This has been several times topic of different threads. If you use :search: , you will get the reason why DivX 5.1 is slow. If you want to get some speed improvements use DivX 5.1.1 as, it is a updated vesion of previous DivX and has some speed improvements.

From now on try to search first before posting.
Also you may want to look at forum rules :[http://forum.doom9.org/forum-rules.htm]

Happy Encoding!:)

[jggimi]

Welcome to Doom9's forum, TotalChaos!

The reason Tuning suggests reviewing the rules is because the encoding speed differences between 5.1 and 5.0.x releases (and 5.1 / 5.1.1) has been discussed at great length, and you've posted the same questin in two different forums.

As you no doubt understand, Rule 1 exists to reduce repeat requests for the same information over and over. Rule 8 exists to keep from splitting a discussion into multiple forums.

But don't worry, you'll figure the rules out fairly soon, I'm sure. And if any rule confuses you, just ask.

[TotalChaos]

Your right in saying that I probably could have looked further into the forum for more info. The threads I did look at didn't seem to give me the answers I was looking for though. You mentioned a difference between 5.1 and 5.1.1. The threads I followed didn;t mention any difference there only some things like virtualdub being the culprit. Sorry to break the rules?! Thanx any ways though.

[jggimi]

I'm sorry you didn't see any of the many posts on this issue. For example, in DigitAl56K's thread (http://forum.doom9.org/showthread.php?s=&threadid=65334) where you asked the same question, JohnMK referred to the first post in thread. In that first post, DigitaAl56K refers to this thread at DivX Networks (http://forums.divx.com/viewtopic.php?topic=55597&forum=5), which states: Performance enhancement that sees version 5.1.1 up to 112% faster than version 5.1 and up to 30% faster than version 5.0.5A search for "5.1" and "slow" will find 57 posts just in the DivX Encoding forum, including several in DigitAl56K's thread you posted in, such as:

http://forum.doom9.org/showthread.php?s=&postid=401541&highlight=slow+5.1#post401541
http://forum.doom9.org/showthread.php?s=&postid=401579&highlight=slow+5.1#post401579
http://forum.doom9.org/showthread.php?s=&postid=401678&highlight=slow+5.1#post401678
http://forum.doom9.org/showthread.php?s=&postid=401700&highlight=slow+5.1#post401700

Sorry you missed them all. It can take time to learn to use the forum search tool efficiently and effectively.

[TotalChaos]

I've searched a few key words and came up with mostly the same threads. I reviewed these threads and others, and still I find that even using the latest divx 5.1.1 my encode times have NOT changed. As one thread suggested I deleted my default divx settings files from the GK folder. I've even uninstalled the entire codec and deleted EVERY reg entry I could find pertaining to the divx codec. Upon reinstallation and resetting the codec's setttings (again to standard speed and without using profiles) the speed just will NOT increase. I have about 23 movies pending for archiveing, so I don't want to continue encoding unless I can get the speeds back up to around 70-100FPS that I was getting with the earlier DivX 5.05 (getting like 10-20FPS MAX right now) PS if I just am too ignorant to find the correct thread myself could anyone help out a fellow encoder with the links to the info I need. (should change my handle from 'totalchaos' to 'ZERO patients') thanx in advance!!!

[colordog]

10-20 FPS is certainly what I've been getting with 5.1.1, more towards the 20 FPS (on the "Slower" setting with Fast PV). (With P4 2.4 GHz HT, 800MHz FSB, 1 GB of 1GHz RAM)

What kind of codec / codec settings / hardware / encoding properties did you use to get 70-100FPS? That sounds unbelieveable to me! Prior to the DivZX 5.1 series, the 5.0 series would give me about ~35 FPS on the same hardware, if I wasn't using deinterlace or IVTC. Even with your overclocked system, I'd guess you wouldn't be able to get above 40-50 with the older 5.0 series codec - wow.

But if I'm getting ~16FPS, you should be getting 20-25 with that clockspeed. I actually did test the encoding times between 5.1 and 5.1.1 and it actually did increase by 50% for me. What's your screen size, PV setting, and are you using standard, slower, or slowest?

[TotalChaos]

I'm not trying to brag or look good here. I do wish I had an actual multi CPU system with some crazy overclocking; however, with this OCed HT 3.0Ghz using the Divx 5.05 codec I had a 60FPS normal with momentary peaks of 70-100FPS usualy the lower or closer to 70 the peak FPS was the more frequently it would peak at that FPS. IE peaking more at 70 FPS then at the vastly more rare 100 FPS. When using my older 2100+ athlon NOT OCed with older Divx 4 I was getting like 25-35FPS then barely able to get 30FPS when I moved to DivX 5.05. This computer in almost every benchmark doubles any results from the 2100+. If that one can get 30FPS why can;t this one get 60FPS now, or rather when I was using the DivX 5.05. All said and done I think 10-20FPS sucks a _ _ with a system of this caliber. On a better note, I found that enabling Qpel did kill my FPS alot, but even with this feature off I can only get 40-50FPS now. Hope I am not as crazy as I sound here...thanx for all this info. (bad or good)

[TotalChaos]

One thing I found that will help your encode times is to disable 'quarter pixels' or 'Qpel.' That helped me out a lot, still not like I had hoped of course, but deffinity better then before! Also according to some that feature didn't realy help as much compared to how much it slowed down encode times. If you like try it without Qpel enabled and see if you like the results.

[fabecool]

Hi,

I'm experiencing the same problem. I don't think I've ever reached 70 to 100 fps when encoding with divx 5.0.5, but it was much faster than the 5.1.1 (currently between 2 and 4 fps with peaks up to 6 fps), and I was able to backup two DVDs overnight.

Now I'm trying to backup my "Pirates of the Caribbean" DVD, and it's been running for almost 24 hours, but it is still not done (I think I'm in the second pass right now).

My computer is not a brand new one, but still, it's an Athlon XP 2000 with 1 Gb DDR (266). I had started the encoding with my old computer (P3 @550 mhz, 256 mB sdram), but fortunately, I realized it would take forever, so I just swiched to the other machine and started the whole process over again.

Of course, I must confess I chose the slowest setting to get the best quality, I'll try with the standard one next time, but it seems strange that it is taking so long.

Also I found this on another thread (post by SeriousWorm (http://forum.doom9.org/showthread.php?s=&threadid=73039&highlight=slow+%7E2200)):


- you said backuping to DivX lasts 6 hours... I wonder how come you got so slow of a PC and you still claim you have a DVD burner? I'd rather upgrade to a, say ~2200+ Athlon XP and shorten the process to ~3 hours (2 passes).

3 hours, 2 passes? Really? What settings? What's the quality like???

[TotalChaos]

In my resent 'studies,' three things changed my encode times the most: Quarter pixels-QPEL(on = slower | off = faster), Global Motion Conpensation-GMC(on = slower | off = faster), final resolution after cropping and resizing(larger sizes = slower | Smaller = Faster)
I realize now that the high frame rates I was getting were mainly the result of compressing animated content at lower resolutions. Also too try and point out as many factors the affect encode times, I must include the settings of either standard, slow, slowest:(Standard = fastest | slowest = slowest). One last thing is the fact that you are using an athlon system to encode. While I have both AMD and Intel systems Intel is king when it comes to encoding.(Be it only for now I hope....come on AMD!!!) It appears encoding is clock speed friendly.

(my system for encoding OCed 3.0GHz@3.25GHz 868MHz FSB 1GB PC3700 dual 10,000RPM western digital HD)

[Sharktooth]

Avisynth filters makes encoding a really slow job.
Remove any avisynth filters.

[scmccarthy]

@Sharktooth,

Avisynth has gotten a lot faster since the days of DivX 4.11 when the original poster started using Divx.

It is not kosher to restart an old thread to add something completely irrevelant to the thread.

By now DivX has released DivX 5.1.1 promising that it has the speed of 5.0 and the quality of 5.1. So even the people at DivX admit that 5.1 was slow. That makes this thread moot.

So it is obvious you only replied to this thread in order to rip on AviSynth.

That is a pretty ridiculous thing to do in the GordianKnot forum, since GordianKnot uses AviSynth.

Stephen

[jggimi]

@scmccarthy:

It was fabecool who restarted the thread, not Sharktooth.

While I can understand your ire at Sharktooth's comment, I find it necessary to make the following request. This is not just to you, but to all:

Please, keep an appropriate level of decorum in this forum.4) Be nice to each other ....

[scmccarthy]

People are afraid to say anything of any substance on this forum.

I can appreciate my ire too.

I replied 'appropriately' not 'nicely'. They are not synonyms.

Decorum would be nice. Rational discourse would also be nice. Neither is possible until Jesus comes again. We live in a fallen world.

There is no other way to respond to a gross over-generalization.

Stephen

[Sharktooth]

i wonder why ppl keep bashing again and again on users who try to help.
as jggimi said, it was not me who restarted the thread.
'nuff said. at your polite reply here's my polite retaliation. Read my lips and look at my finger: F**K U!

@the rest of the ppl: I'm sorry to use this language but sometimes it becomes necessary. Really. If mods want to strike or ban me for violating the rules i can understand it.
Thanks.

[len0x]

Any further insulting replies will result in more strikes immediately.

[manono]

scmccarthy-

Your whole first post just about, was based on the mistaken impression that Sharktooth had restarted the thread. That made much of your post nonsensical. So did you apologize in your second post? No, you bashed him some more. Personally, I think you had it coming, when Sharktooth finally replied.

I had no trouble understanding what he was saying, and didn't find that he was badmouthing AviSynth at all. It's a fact that certain AviSynth filters will slow the encoding considerably. One of the earlier posters wondered why the encoding was so fast earlier, and then slower later on, and then "discovered" that the resolution used just might have had something to do with it. With that lack of knowledge displayed, Sharktooth thought it worth pointing out that the filters used might also have played a part, and I agree.

You also have the huge advantage of having English as your first language (and the proven ability to use it well), and Sharktooth doesn't. Maybe Sharktooth should have said, "Certain Avisynth filters makes encoding a really slow job." He didn't, but I still had no trouble making out what he was saying. You shouldn't have, either.

[TotalChaos]

@manono

I must admit being somtimes confused with the diffirent options provided by the latest gordian knot/divx/virtualdubmod...ect and the older versions I was "slighty" more familiar with. You mentioned a poster "discovering" that the final resolution made a difference in encoding speed. You then say this user displayed a lack of knowledge. I assume this comment is directed at me. Offend?? Well if your wrong your wrong. Guess I can't get offended if I in truth do not know what I am talking about. I will admit finding I know less and less about video encoding then originaly thought; as I am finding video encoding to be a much larger subject. Since this thread is already way off it's original topic may I ask, "what effect(s) if any the resolution (after crop and resize that I call the final res) has on the encoding time?"

[manono]

Hi TotalChaos-

Yes, I used you as an example purposely, to help point out how Sharktooth's post might have had some relevance to the encoding speed issue.

It has to do with the number of pixels being encoded. To take an obvious example; 640x480 has 4 times the number of pixels as 320x240, and will therefore require much more time to encode. Will it take 4 times longer? Not necessarily, I don't think, as there are many other factors involved. But it's easy enough for you to test it out, by encoding a sample at the 2 different resolutions.

Here's a less obvious example; fullscreen 512x384 seems to be a much lower resolution than widescreen 640x272. But because it consists of more pixels, it'll take longer to encode, all other factors being equal, than will 640x272.

I knew when you mentioned 100fps encoding that you were probably using a real low resolution. But you have to understand that such claims are pretty much meaningless unless you detail your hardware, all of the codec settings, provide the .avs, and even the name of the movie. If you check out Doom9's Codec Comparison (http://www.doom9.org/codecs-203-1.htm#setup), you'll see how careful he was to list everything that might affect the encoding, so that those people that might want to confirm his findings can replicate his tests.

[scmccarthy]

@manono

I always assume people mean what they exactly what they say.

Critisism is good when you offer an alternate. If GordianKnot is not the best, what is the alternative. Obviously, for instance, someone could knock GKnot and say AutoKnot is better or someone else could say GKnot is better than AutoKnot and no one would be offended. That is why we have choices.

Slow is a relative term, if someone asserts that AviSynth is slower than another application or 'some' of the filters are slower than others and which ones might be so, that is good and useful info, obviously (it would be obvious to me anyway) given in good faith. Even is someone were offended by that, I would stickup for him, free speech and all that.

Manono, since he did *not* say 'certain AviSynth filters' are slower than others, I have a real hard time understanding what Sharktooth meant to say, if he did not mean to knock AviSynth.

Personally, I don't think AviSynth is slow for what it does or that there are not alot of speedy filters. That is just my opinion and I don't mind if people disagree with it. I do mind when someone knocks AviSynth's speed in a thread that was discussing how to optimize DivX' speed without qualifying the comment at all.

Because Sharktooth posted after 5.1.1 was released, it seemed to me he *had* an ax to grind. (To be fair, maybe he would not say the same thing now if I had not made an issue of it, I am not holding a grudge here.) I don't even mind when someone *has* an ax to grind as long as he is honest about it and says as much.

I hope no one thinks I am still upset about it. This is offered as an explanation of what I think 'honest discourse' would look like.

By the way, I did not read Sharktooth's response because I read Len0x' first and I do not want to get into it.

Stephen

[scmccarthy]

Look, for the record, I am sorry if I hurt anyones feelings.

Stephen