HC vs CCE vs AutoQmatenc vs Quenc

[gizzin]

I used DVD rebuilder v1.25.0. The source was progressive. It was Sopranos S1 D1 (runtime was 3hrs 26min). The average bitrate was 2,398. The TargetSectors=2260000 was the same for every encoder, I like to get closer to 4.37 (4,700,000,000). I almost always get really close to that mark with every encoder.

I will start off by saying I was lazy and had lots of time on my hands. Also, that this objective as to what I think is best. I remember saying awhile back that Quenc was really slow it actually wasn't my computer was messed up, the timing on the ram was wrong. I got new ddr2 dual channel 800mhz so its fixed now. I also have to point out I have a amd x2 4200+ dual core processor. HC was 107min (v.21 best). CCE SP was 124min encoding time (v2.70.02.12 trial 4pass). Quenc was 206min(v.72 2pass, trellis, high quality mode, scene detection). AutoQmatenc was 248min (v0.33.0.7, QMatOpt 4pass). In the speed department HC was the fastest and AutoQmatenc was the slowest. For quality, I think HC was the best. Then I would say AutoQmatenc. Followed by CCE. And In last Quenc. I do this by looking at the mosquito noise and blockiness the encoding artifacts. I have sony lcd 60'' HDTV as my viewing device. I watched the first episode once for each encoder. HC had little blockiness except for the fade in and out scenes. It wasn't noticeable noisy at all either. You had to look closer to see it. To me, AutoQmatenc and CCE SP were the same in the amount of blockiness especially bad in the flat areas. The noise for CCE was bad very noticeable annoying almost. AutoQmatenc noise seemed different to me I can't remember why I thought I think it was because it was in different areas then other encoders but I don't really remember. Most of it was around the hair region, around the ears, edges of the hands, etc. Quenc was the most blocky. Really bad on flat areas and it was in other places then the other encoders. Were as the other encoders at no signs of it in those areas. Less noise then AutoQmatenc and CCE SP however. The only reason I rated it the lowest is because of the blockiness and to me I'd rather have more noise than blockiness. In my opinion the only way to really see the encoding quality is by actually watching it.

[ricardo.santos]

....Quenc was 206min(v.72 2pass, trellis, high quality mode, scene detection)....AutoQmatenc was 248min (v0.33.0.7, QMatOpt 4pass)....

you enabled al the whistles and bells no wonder it takes that much time, what were the settings in HC?

...HC was 107min (v.21 best)...

there lots of posts here saying trellis actually is not a good choice, deactivated and you'll sure get a speed boost

ive done that test before, with dvd sources and compressed ones and Quenc came on top not just on speed

[techreactor]

HC was 107min (v.21 best). CCE SP was 124min encoding time (v2.70.02.12 trial 4pass). Quenc was 206min(v.72 2pass, trellis, high quality mode, scene detection). AutoQmatenc was 248min (v0.33.0.7, QMatOpt 4pass). In the speed department HC was the fastest and AutoQmatenc was the slowest.
Enabling trellis/ QmatOP increases the time considerably with these respective encoders although 4 pass is good enough with AQMenc since it will use only as many passes as required irrespective of the max passes choosen.

Were you using multiple instances option ??

For quality, I think HC was the best. Then I would say AutoQmatenc. Followed by CCE. And In last Quenc.

There is not much difference left under normal BR amongst these encoders for quality but the real test is under very low bitrates and finally its a matter of personal prefrence.

[gizzin]

you enabled al the whistles and bells no wonder it takes that much time, what were the settings in HC?

there lots of posts here saying trellis actually is not a good choice, deactivated and you'll sure get a speed boost

ive done that test before, with dvd sources and compressed ones and Quenc came on top not just on speed

Best quality setting HC

why is that not a good choice it is suppose to raise quality? Any setting that raises quality is
good. If a encoder took 600min and it was the best quality compared to the others ones I used. I'd use that one all the time then. Time is not a factor for me.

I don't think I ever used Quenc without trellis, I'll try that today and post the difference. I'll defrag HD, close all apps etc and encode Trellis and non Trellis and see if I can notice any difference. Thanks for the idea.

[ricardo.santos]

i never use trellis, usually i use a bitrate beetween 3500 and 4000 and im quite happy with the output, also one of the reasons i use quenc is that i can encode video+audio using the same program, if i want to convert an avi i just quickly build a avs feed it to quenc, select video and audio bitrate and press encode, with HC i cant do that last time i tried AutoQmatenc i didnt, CCE only encodes to MPG2 not Ac3 like Quenc.

[dragongodz]

why is that not a good choice it is suppose to raise quality?

it is NOT a good choice if you are comparing speed. trellis is unoptimised and slow as hell. it is only of use for very low bitrates IMHO. even then its more minor improvement rather than maximum.

Any setting that raises quality is
good. If a encoder took 600min and it was the best quality compared to the others ones I used. I'd use that one all the time then. Time is not a factor for me.

then you would be better off using the extreme setting instead of trellis. that is even slower but should provide more quality increase. however i only ever reccomend using it for harder material or encodes. for example there was a thread on here where a guy had a piece of footage that had lots of chroma issues. QuEnc with extreme gave great results compared to other encoders, according to those that tested it.

everyone please remember that when you test different encoders they may have different settings. these settings do not mean the same thing, ie. HC best is NOT the same as turning on trellis or extreme in QuEnc, the HQ setting in QuEnc is closer to HC's "best" setting, even then its still not exactly the same thing.

infact gizzin you started a thread about speed comparisons and hank315 showed exactly how to do it right.
http://forum.doom9.org/showthread.php?p=970400#post970400

[gizzin]

Everytime I post something you come in Quenc defence. That's your baby I would come in its defence also. The way I do my comparisons is using the best quality settings provided by the encoder. I know Quenc and HC are 2 different encoders and the settings do different things, of course. Speed Vs quality that's always been a big debate, for me quality for some its speed. How would I use extreme & slow in dvd rebuilder? I looked for the hidden setting but I couldn't find it. As far as hank encoder tests, I would just say there more comprehensive. Using all the different settings and comparing them. I would do that but I don't have the time or processing power to do that. I should of bought a pentium heh. I was really comparing quality not speed just thought I'd point out the difference. Like I said before if there was more options in HC that increase quality I'd use them I don't care about time. And in Quenc case they greatly decrease speed as you said.

When refering to chroma what exatly do you mean

[dragongodz]

Everytime I post something you come in Quenc defence.

not at all. i just would rather see if people are going to post something they include relevant information. such as the reason QuEnc was so much slower is the trellis setting which i do not reccomend.

The way I do my comparisons is using the best quality settings provided by the encoder.

then the extreme setting should have been what you used and not trellis. though as i said i do not even reccomend that except for extreme cases. oh and its even slower than trellis of course.

How would I use extreme & slow in dvd rebuilder?

run QuEnc. tick "extreme". quit QuEnc. when next you run DVD-RB it should still be turned on, so long as DVD-RB doesnt turn it off, which i dont think it does. you can check by running QuEnc again after you have finished DVD-RB and see if it is still on. if its not then DVD-RB must have turned it off.

As far as hank encoder tests, I would just say there more comprehensive.
understatement. thats like looking at a piece of an apples skin and saying thats just as good as cutting it in to pieces and looking at every bit of it. :)

Using all the different settings and comparing them. I would do that but I don't have the time or processing power to do that.

well frankly i dont expect you to. all i ask is if you use a setting that you know is as slow as hell then mention that that setting is why its so much slower.

When refering to chroma what exatly do you mean

heres the thread i was talking about
http://forum.doom9.org/showthread.php?t=110992

chroma being shorthand for chrominance. use search on this and other video related sites and you will generally see chroma.

http://en.wikipedia.org/wiki/MPEG-2

go to "video coding(simplified)"
Each picture element (a pixel) is then represented by one luminance number and two chrominance numbers. These describe the brightness and the color of the pixel (see YCbCr).

[gizzin]

My thread got hijacked. PSNR I don't think it even should be mentioned when it comes to quality of the video. As far as Hank's encoding test were, it was just for the different speeds so in another words a speed test am I right? or am I wrong there?. For sure hank's test are better he used every single feature of every encoder and showed the difference. He never ever said what looked the best to him, if he even looked :) I still do think my post are very relevant. My post should be called "The best supposed settings of every encoder". I'm sure most people get the point that trellis increases quality very little and should only be used in hard to encode scenes same with extreme and slow setting. In your Humble opinion if using a lower average bitrate say around 2000kb. Does using trellis and or extreme setting increase quality in the tests that you've done? Should I mention that doing 3 or 9 passes in CCE is waste of time? Everytime I do a comparison? For me it is to squash every little bit of quality out of that 4.7gigs of data. For an average user looking at what encoder to use. I would say not everyone has a beefy machine and therefore speed would be a factor. All I have been trying to say the whole time is you put those features in for a reason because they increase quality. And thats what my tests are quality not speed. quality...

It's *not* emulator output, but analog Amiga RGB run through a broadcast scan converter to YUV 4:2:2 and captured through a DeckLink SP via component cables -- pretty nice capture, eh?

Then its original is from a analog source so isn't it going to have luminace problems to begin with? Also the bitrate was 8500kb CBR. That's not a realistic figure when it comes to encoding DVD to have 2 or 3hrs of video. But I guess that wasn't his purpose. Probably what most people use encoders for. It's to see how a encoder performs at a lower bitrate than higher. I find at 4000 avg with almost any encoder theres no encoding artifacts or even be able to dintiquish them from the original.

[dragongodz]

As far as Hank's encoding test were, it was just for the different speeds so in another words a speed test am I right?
yes basically speed.

And thats what my tests are quality not speed.
however you do say
In the speed department HC was the fastest and AutoQmatenc was the slowest.
and say how long each was. any newb reading that may not understand WHY QuEnc was so slow in your test. and yes i think if you did 9 passes with CCE and it was the slowest because of it i would think it reasonable to say thats why considering most agree more than 3 passes will gain you next to nothing.

In your Humble opinion if using a lower average bitrate say around 2000kb. Does using trellis and or extreme setting increase quality in the tests that you've done?
trellis very little if anything, you would have to be looking at still images to see the differences for a lot of footage.
extreme would gain you more but again depending on the footage it could be hard to see in normal playback. i would definatly try extreme over trellis if you were not happy with the quality.

Then its original is from a analog source so isn't it going to have luminace problems to begin with? Also the bitrate was 8500kb CBR.
the point was that settings like extreme can help difficult footage. for general purposes trellis and extreme will gain you very little if anything but slow encoding down greatly.

also lets look at your encoding again.
Sopranos S1 D1 (runtime was 3hrs 26min). The average bitrate was 2,398.
thats a getting low bitrate. i would have recommended using the QLB matrix with QuEnc. CCE and HC default matrices are already aimed at mid to lower bitrates.

[gizzin]

and say how long each was. any newb reading that may not understand WHY QuEnc was so slow in your test. and yes i think if you did 9 passes with CCE and it was the slowest because of it i would think it reasonable to say thats why considering most agree more than 3 passes will gain you next to nothing.

I love this response. heh

I will try using QLB matrix see how it turns out. I'm still gonna use trellis and extreme and slow. Speed wasn't my main concern is what I mean. But they should still know what it was. From now on I will add "trellis and extreme and slow settings are meant for hard to encode scenes and are good for low bitrates but the downfall is it greatly reduces speed." To any of my quenc related encoding comparisons. Good job on the encoder by any means its good.

[manolito]

@gizzin
I remember some time ago there was a discussion in the QuEnc thread if it was better to remove the Trellis opption from the interface altogether. The reasoning behind this was that users who did not take the time to dive in a little deeper might just select Trellis and expect an improvement in quality.

The sad truth is that Trellis not only does very little to improve quality, it even might do quite the opposite in some cases. There have been numerous reports for FFMPEG that Trellis caused all kinds of problems.

So please for any future tests which include QuEnc, please just ignore the Trellis option. Use HQ or even Extreme and Slow to your liking, just DO NOT use Trellis.

Cheers
manolito

[Sagittaire]

1) As always if you want compare speed you must try to use exactly the same quality level. The only way for make that is to compare speed with the same objective quality. Use a subjective methode to measure quality here is impossible.

2) As always use the max quality for the codec is not an obligation. QEnc use the core codec libavcodec library and with max quality (RDO at max, trellis, SADT, diamond search at max ... ect ect) the speed can be really really really slow. With libacodec you can use multipass, RDO for best adaptative bframe placement and other really slow setting. QEnc "high quality" profil is not in fact the best possible profil for libavcodec and if dragongodz want then he can choose very faster setting for the "high quality" profil. "highest quality" profil is a really subjective subject for a video codec.

[linx05]

If a encoder took 600min and it was the best quality compared to the others ones I used. I'd use that one all the time then. Time is not a factor for me.

As far as hank encoder tests, I would just say there more comprehensive. Using all the different settings and comparing them. I would do that but I don't have the time or processing power to do that. I should of bought a pentium heh.
I thought you said time wasn't a factor? :P

In all seriousness, and I'm not trying to take a jab at you, have a look at Hanks test and do it that way.

[gizzin]

In all seriousness, These are 2 totally different things. And I was refering to the encodes I do for backup purposes not for testing every single setting a encoder has. Like I said more comprehensive but not my purpose.

Oh yea, I did the sopranos disc again. This time adding extreme and slow to the arsenal Quenc provides. The encoding this time was 468min. It provided better quality compared to the other encode I did which had all the other features except extreme and slow. It just seemed to reduce the macroblocks or blockiness whatever you wish to call it. The noise wasn't very bad along with the other encode I did. No difference there as far as I could tell. Trellis and extreme and slow settings are meant for hard to encode scenes and are good for low bitrates but the downfall is it greatly reduces speed

[gizzin]

1) As always if you want compare speed you must try to use exactly the same quality level. The only way for make that is to compare speed with the same objective quality. Use a subjective methode to measure quality here is impossible.

And the same objective quality would be what? And by subjective you mean like psnr?

I forgot to mention at recommended settings using Quenc (HQ, 2pass) it took 112min

[dragongodz]

This time adding extreme and slow to the arsenal Quenc provides. The encoding this time was 468min. It provided better quality compared to the other encode I did

i take it that was with trellis still enabled ? you could have turned that off and saved quite a bit of time for practically the same quality increase IMHO.

did you also try using QLB matrix , even without extreme and trellis, to see if that helped aswell ?

[gizzin]

Ya, that was with trellis as well. I did a encode with QLB matrix and one without I just haven't looked at them yet, No extra settings just 2pass, HQ. I plan to do it tonight.

[gizzin]

I tried really hard to notice any difference between the two. I mean there might have been a slight improvement in using the qlb matrix but I couldn't really tell.