RobRedbeard

Politely as possible -- you are completely wrong.

These types of formats don't work that way because there is processing that must be done to the output image. Lossful formats do not produce bit-for-bit replicas. This is why two different MP3 decoders can have very different quality, why DivX and Xvid will produce differences in images when decoding the same stream, etc. Finally, this is why FFDShow's decoder doesn't, at all, in any way, produce an identical image to CoreAVC's. Please independently verify this.

Here is a place to begin: http://www.avsforum.com/avs-vb/showthread.php?t=920229
Rob

__________________
CPU-Z Validation | GPU-Z Validation | Full System Specs | Photo Gallary | Endor IIa: Adventures in Cable Management |
Endor II 3dMark06: 14697 | Endor III 3dMark06: Coming Soon!

JoeUbi

I'm using the regular MPC.

I also encode much of my own video, so my set-up is riddled with many different codecs.

__________________

Coma

Okay, sorry for being rude

But I really don't think so. Different encoders of the same standard can compress more efficiently, so they don't produce identical streams.

On the other hand, DivX and XviD decoding the same stream should produce identical output. I'll actually test for this - I'll use GraphEdit with ffdshow and CoreAVC to decode a H.264 stream and dump it to file, but... if both decoders comply to the standard, then they should have the same output, bit by bit. That's what standards are for.

Joe: Then it's probably that. I'm guessing the extra memory usage is because of DXVA support.

edit: Okay, Rob - you win :P I asked on doom9, and even tested for myself, and nope, they weren't identical. They shouldn't be visually distinguishable, though... which means you *did* have something doing postprocessing, or you just dug out some very incomplete decoder :E

You could judge this from neuron2's reply, too. Like, the difference would be in HOW the stream is... (can't find a better word) "contained" - that is, as frames or fields, etc... but the actual contents (what you see) shouldn't be different (since the specification defines how to decode, you just can't do it any other way).

So basically the difference is in how it is "contained", similar to the difference between MKV vs MP4, which can contain the same encoded streams but each file will be different because the container is different.

To prove my point, (unless I made a horrible mistake), in Avisynth, I used AVCSource() and directshowsource() on the Mirror's Edge animated trailer, and interleave()d the two clips.

I saved PNG screenshots of frame 372 (avcsource, libavcodec (ffdshow)) and frame 373 (directshowsource, coreavc), and they both produced the same hash.

edit9000: neuron2 missed the "decoder" part in my thread :P
Apparently, decoders are supposed to produce identical output, the only exception being MPEG-2 decoders... but as for H.264, it should always be identical.

RobRedbeard

To be honest, I've done video and audio work for a long time. I used used to own an audio/video production studio. This is the same reason that different DVD players can have different visual quality. The decoding process is scalable. In addition, during decoding a lot of filtering is applied as part of the standard process, but that filtering quality and what patented technologies it uses can contain a lot of variation. This is why ATi and Nvidia have been duking it out over whose DVD and BR output looks the best, and why there are expensive test DVD's made for that purpose.

I hope this gives you a greater appreciation for what the fine people at CoreCodec have created. It's not only fast as hell and cheap, but also boasts really impressive quality. I only use Cyberlink because I figure I've got the GPU designed to do it, I might as well use it.

Cheers,
Rob

__________________
CPU-Z Validation | GPU-Z Validation | Full System Specs | Photo Gallary | Endor IIa: Adventures in Cable Management |
Endor II 3dMark06: 14697 | Endor III 3dMark06: Coming Soon!

Coma

Well, actually I forgot to edit that part... but it turns out you didn't win at the end :P
While each stream is not identical (because the frames can be contained in different ways), the individual frames are. XD
Check the new replies at doom9 >:O

Input -> libavcodec -> dump and Input -> CoreAVC -> dump are different, but Input -> libavcodec (mpc) -> BMP and Input -> CoreAVC (mpc) -> BMP are identical, bit for bit.

NVIDIA and ATI have been battling over who has the best postprocessing or which is the best way to display the decoded video, but not the actual quality.
For example, ATI cards convert the levels to PC levels, while NVIDIA cards leave them as is... so ATI color and brightness looks much better, but sometimes they lose details in dark areas of the image.

Also, NVIDIA cards give everything a red tint.

P.J

Just use Haali + PDVD 8 Ultra (w/o DxVA, cuz you've 8800GTX)
Don't forget to rename *.mkv to *.avi

__________________
(Vegnagun666's TANK)

●█████▄▄▄▄▄▄▄▄ -=Pwn'd
▄▅███████▅▄▃▂
█V█O█S█T█R█O█████► - - - = MacBook Air = - - -
◥☼▲⊙▲⊙▲⊙▲⊙▲⊙▲☼◤

Coma

no >_>

MKV is socket 775 and AVI is 478... you don't put a socket 775 CPU in 478, do you?

RobRedbeard

The video as it is displayed on screen and perceivable by the view is where quality is measured, not at the level of an internal frame-buffer between the decoding and filtering stages of the codec. When you use different codecs you can SEE sizable differences in the final result. The reason why isn't as important as the reality of the result.

Also, Nvidia's system does more than slightly red-tint the output.

Rob

__________________
CPU-Z Validation | GPU-Z Validation | Full System Specs | Photo Gallary | Endor IIa: Adventures in Cable Management |
Endor II 3dMark06: 14697 | Endor III 3dMark06: Coming Soon!

P.J

Just try to see

__________________
(Vegnagun666's TANK)

●█████▄▄▄▄▄▄▄▄ -=Pwn'd
▄▅███████▅▄▃▂
█V█O█S█T█R█O█████► - - - = MacBook Air = - - -
◥☼▲⊙▲⊙▲⊙▲⊙▲⊙▲☼◤

Coma

If the frames being output are identical then you're seeing different quality because:
a) postprocessing, or a different filter somewhere else
b) imagination
c) not looking at the same frame

So there SHOULD be no difference between decoders d; Save two screenshots, one with ffdshow and one with CoreAVC, and hash them... you'll get the same hash UNLESS one of the decoders (probably ffdshow) is doing postprocessing.

Rename a RAR to ZIP and it's still a RAR, it just has a different extension.
If it works then it's because your player went into compatibility mode (it's ignoring the extension and using the contents to determine type), but solving whatever problem you're having with MKVs is a much better idea.