Elecard HD vs x264

[Dark Shikari]

Elecard/Mainconcept's latest encoder, Converter Studio HD, has a new feature that b0bor dubbed "AT"*.

See it for yourself here (http://mirror05.x264.nl/Dark/x264vsElecard/index.html).

Mouseover = x264
Mouse not over = Elecard






























*AT stands for "adaptive trashcan", which adaptively throws away perfectly good bits in the name of lower quality.

[qyqgpower]

lol
yesterday I just tried MainConcept Reference Encoder 1.1, it also failed to limit the bitrate in that (http://forum.doom9.org/showthread.php?p=1119575#post1119575) source. there are still some spikes over 50mbps.

It even failed to achieve the target bitrate in 2pass mode, gave me 5381kbps instead of 6000kbps.

target 6000kbps
SSIM
MC@5381kbps:AVG: 0.99168
x264@5858kbps:AVG: 0.99247

Although SSIM indicates that MC is not far from x264 (or may be better if taking 500kbps bitrate gap into account), but my eyes seem to be more pleased with x264 result.

[Sagittaire]

Well I can certainely obtain very better result with Elecard/mainconcept encoder:

1) Where is the source
2) Parkrun source is not available
3) post your command line ... please

[Dark Shikari]

Well I can certainely obtain very better result with Elecard/mainconcept encoder:

1) Where is the source
2) Parkrun source is not available
3) post your command line ... pleaseThe source is on my hard disk. I can make a lossless encode later, but its extremely large. The FRAPS is over 3 gigabytes. I could provide a sample of part of it.

Commandline: x264 --bframes 3 --b-rdo --subme 7 --b-pyramid --weightb --bime --me tesa --direct auto --crf 30 --ref 16 --8x8dct --partitions all --mixed-refs --threads auto --progress --trellis 2 --ipratio 2.5 --keyint 600 --aq-strength 1.3

Keyint 600 was used because the source is 60 FPS (keyint = 10*framerate). I turned on complexity masking on the default strength in Elecard, and put all options to what appeared to be their slowest options. I likely played with the x264 settings a bit more than Elecard simply because I know more about it (so in this regard this test may be slightly unfair; I suspect someone more versed in Elecard could get slightly better results). In Elecard I maxed reference frames, GOP size, B-frames, RDO, etc.

I don't actually have the full Elecard output; the encoder crashed halfway through the second pass, but the output was fully correct up to that point and had the correct bitrate, so I figured I could still continue the comparison.

[Sagittaire]

Keyint 600 was used because the source is 60 FPS (keyint = 10*framerate). I turned on complexity masking on the default strength in Elecard, and put all options to what appeared to be their slowest options. I likely played with the x264 settings a bit more than Elecard simply because I know more about it.

Strenght at 0 desactive in practice AQ. Negative value mean more bits for flat area with complexity mask.

[Dark Shikari]

Strenght at 0 desactive in practice AQ. Negative value mean more bits for flat area with complexity mask.That's the oddest interface I've heard of: lower values mean stronger AQ? :p

By that method perhaps --subme 1 would have the best quality and --subme 7 the worst? ;)

I may try encoding with that later (if perhaps Elecard doesn't crash...)

[Sagittaire]

That's the oddest interface I've heard of: lower values mean stronger AQ? :p

By that method perhaps --subme 1 would have the best quality and --subme 7 the worst? ;)

I may try encoding with that later (if perhaps Elecard doesn't crash...)

Well as you know strength for psy masking is by definition subjective. With recent elecard build you can make mask combinaison with complexity, extreme luma, contrast. For example more bit for dark area, more bit for low contrast area and more bit for high complexity area. Less bit for high complexity area is not always the best way (for grain/noise retention for example).

[audyovydeo]

Hello DarkS

somewhat off-topic, but your cmdline intrigues me :


--bframes 3


I thought you'd established months ago that bframes should be maxed to 16, and play with b-bias or let the encode decide (dxva thread notwithstanding ?)


--b-rdo --subme 7 --me tesa --ref 16 --trellis 2


you max out every setting, for a CRF of 30 ? To me that's unusual. I tend to use stronger settings for sources where I seek higher quality (ie crf 18-22). Am I completely off-track ?

puzzled,
audyovydeo

[Sagittaire]

I thought you'd established months ago that bframes should be maxed to 16, and play with b-bias or let the encode decide (dxva thread notwithstanding ?)

It's comparison between x264 and Elecard. You can't use 16 bframes with elecard. Moreover 16 bframes is in practice useless setting.


you max out every setting, for a CRF of 30 ? To me that's unusual. I tend to use stronger settings for sources where I seek higher quality (ie crf 18-22). Am I completely off-track ?

If you have 10% gain between good-insane setting at crf 18-22 then there are gain too at crf 30. Gain at crf 30 will be even very higher with insane setting. In fact I think that for high quality quantisation insane setting are useless and usefull for low quality quantisation.

[Dark Shikari]

If you have 10% gain between good-insane setting at crf 18-22 then there are gain too at crf 30. Gain at crf 30 will be even very higher with insane setting. In fact I think that for high quality quantisation insane setting are useless and usefull for low quality quantisation.I find this to be generally true also. The benefit of all options is magnified at higher QPs, especially RDO.

Edit: Source for this thread's comparison (not lossless, but pretty close (http://www.filedropper.com/higherqualitysource).

[Sagittaire]

Edit: Source for this thread's comparison (not lossless, but pretty close (http://www.filedropper.com/higherqualitysource).

don't work ...

[Didée]

Seems to work for me. Klicking the button in the middle "Download this file" opens a dialog window: "Enter password for this file", with the default password "undefined". Just acknowledging that window with "OK", the download starts.
Perhaps some security setting of you browser is interferring?

[Ranguvar]

Wow... all you x264 devs need to pat yourselves on the back.

[DeathTheSheep]

Wow, this ridiculous game again! No, if I downloaded that source, I'd be watching it, not encoding it.

Video game footage, huh? That's quite a high-motion, flashy, fun-filled sample if I've ever seen one. I wonder if Elecard even remotely considered the possibility of encoding video footage like...this. Though on the other hand, Euclid, maybe. :)