Hardware vs. Software encoders: why take it for granted?

[PigOnWing]

Seems that every time someone says something about 'I got TMPGEnc to do only that much quality, and it doesn't compare to such-and-such commercial VCD that I've seen', the standard answer is 'Well, commercial VCDs are made using high-end hardware encoders, you can't really compare apples and oranges'.

Can't you?

I have to wonder, what's the magic stuff in Hardware encoders? After all, even hardware encoders are, eventually, running software. All that the 'hardware' bit does, is provide a dedicated platform, optimized for speed, to run a certain software. But the actual manipulation, is, eventually, software. Compressing a stream, be it audio, video, still picture, is, evnetually, 'playing with numbers' - that's software. A Graphic acceleration card is hardware which runs optimized software to take off 'burden' from the main CPU, regarding 3D parsing - but it is, at its core, software. A standalone player is dedicated hardware which runs MPG-parsing software. It's still software.

Granted, 'optimized' means lots and lots of tweaking. But why shouldn't it be _possible_ to achieve the same results, using non-dedicated hardware and (could be better!) software? Ok, so it may take even more time to encode. But it still should be able to do the same mathematical manipulation - and, not being attached to 'dedicated hardware', could actually continue to be improved (after all, what is all these 'firmware updates' for different kinds of hardware, if not 'updating the software bit that the hardware holds'?).

Wondering aloud.

-- Piggie

P.S. for moderators: I originally thought to post this at the 'advanced conversion' forum, but wasn't sure it's on-topic. If you think it is, by all means, move it to that - or to whatever on-topic forum you may think it belongs to - and I'll be happy as a bunny.

[omoses]

8) Well, there is a post, I alway wanted to write. The myth, that Hardware encoders are "so much better" can really only be told a myth.

Hardware-Encoders are faster (mostly), but there are just some basic mathematical rules that are applied to the pictures and here we have a real flaw of most hardware-encoders. Because they are made in hardware, and most of the logic on them is "fixed", they can never be as flexible as a software-encoder. I think most of todays hardware-encoders are really programmable transcoders, means they are really fast transforming yuv-pictures to cosine-transformation (which is fpu-intense), and a fast sub-processor is doing the rest, maybe assisted by other hardware-accelerators. what you really have in hardware-encoders is a fast processor (obviously not a x86, but a real signal-processor designed for multimedia-purpose, e.g. Texas Instruments have very powerful ones) with little helpers, that take some work that are cheaper when "externalized". And don't think hardware-encoders do not use software, they do! you just do not realize it.

The main reason, why hardware-encoders (the expensive ones) are better in some way is knowledge, which means, that is a great part of the whole process.
1. They use very good AD-converters, and mostly do not use video as a source, but 35mm-film-material. So most of the noise, that is generated with consumer-equipment (Video) is non-existant. And noise is THE major problem when you use low bandwidth (VCD, SVCD, DivX).
2. They use pre-filtering techniques, that help reducing bandwidth, and these filters are often manually controlled in creating the mpegs. They can be made stronger in high-motion-sequences, and are only used when needed.
3. To say in a few words: Making really well Mpegs is not only a metter of good equipment, you have to know what you do, and you need time.
[/u]

[homerpez]

I question whether hardware encoders are the real reason that commerical DVD's/VCD's are such good quality.

After all, and you can ask anyone with a Digital Cable box, a hardware-encoder doesn't always equal good quality. When I flip through my 100+ channels, I notice that some look absolutely astonishing. Others, well... don't.

There are blocks everywhere, some channels obviously opt for speed vs. bitrate. Some, an analog signal would even be better. I'm sure the same goes for Satellite.

It really is in the quality of the algorythms they use, and what they "tweak". Commercial VCD/DVD probably gets worked over scene-by-scene, perhaps even frame-by-frame, to make sure they have the best possible result.

And, as CCE users can tell you, the more passes that are calculated, the better it looks. I would not be shocked if some commercial DVD's are going through 10 or more passes to insure EVERY scene looks good. After all, they only have to do it once!

[resnullius]

one day when you get an incredible capture quality pass it through terran's media cleaner pro instead and then tell me about it