Pandora 2 Might Have A Vp8 Hardware Decoder


darionapole

Member
Joined
Jul 10, 2008
Messages
221
well, do you remember that google open sourced the VP8 codec yesterday?
keep reading...

Chip makers back Google's open source codec
A handful of mobile chip makers including ARM, MIPS, Nvidia and Texas Instruments said they will support Google's move to establish VP8, a video codec it acquired with On2 Technologies in August.......

Full article:
TechRights, Ex Boycott Novell
 
Hey!

Craigix said that the thought of Pandora2 made him feel sick! We can't have him taking a sick day now!
 
Last edited by a moderator:
I'm not sure a hardware decoder would be necessary for most streaming video anyway.
Most services I know that even offer HD streaming also have SD versions, and there would be no difference on the Pandora.
Not that a better performing decoder is a bad thing of course. :)

I'm just much more interested in H.264. :p

EDIT: I missed the "2". The Pandora already has a general purpose DSP, so we probably wouldn't need to wait for new hardware. That's assuming they bother to produce codecs for older hardware of course.
 
Hahaha!
He still might be using the Read only access he had!
How could he not want to read the Vegemite thread!!
 
Aninhumer said:
I'm not sure a hardware decoder would be necessary for most streaming video anyway.
Most services I know that even offer HD streaming also have SD versions, and there would be no difference on the Pandora.
Not that a better performing decoder is a bad thing of course. :)

I'm just much more interested in H.264. :p

i said hardware because the article says the VP8 decoder is heavier than the H.264 one. If it's not hardware then we'll need a very fast cpu in order to decode something with more than 720x480.
 
Last edited by a moderator:
I don't think there is such a thin as a singular-codec specific hardware accelerator. It's just a DSP with an accelerated codec isn't it? Well we have a good DSP that's not being used yet. It just takes somebody to write one.
 
I'm a little disappointed that Google went straight to VP8. It would have been nice if they used Theora for SD, and VP8 for HD, since we know that current hardware will play 480p Theora. But so far Youtube is only doing VP8 for 720p and above.
 
Neko said:
I'm a little disappointed that Google went straight to VP8. It would have been nice if they used Theora for SD, and VP8 for HD, since we know that current hardware will play 480p Theora. But so far Youtube is only doing VP8 for 720p and above.

VP8 is easier to decode than H.264, and H.264 decodes on most relatively modern hardware.

We just need DSP-accelerated codecs to be written/ported to our platform.

Theora is okay - it's decent for an MPEG2 replacement - but I value image quality over ease of decoding.
 
Last edited by a moderator:
paulguy said:
Well, have fun with that. The VP8 spec is supposedly primarily a lot of obfuscated C code and nonsense x86 assembly.
Wasn't that said by an x264, who is obviously a conflict of interest. This dev spent years optimizing x264. It's not hard to understand why people are trash talking vp8. Lets not judge vp8 until we get our hands on it.
 
Last edited by a moderator:
maplesugarlover said:
paulguy said:
Well, have fun with that. The VP8 spec is supposedly primarily a lot of obfuscated C code and nonsense x86 assembly.
Wasn't that said by an x264, who is obviously a conflict of interest. This dev spent years optimizing x264. It's not hard to understand why people are trash talking vp8. Lets not judge vp8 until we get our hands on it.

This critique appears to be one of the only critiques on the internet. I don't know anything about it apart from I know Google aren't stupid. If it's worthless then they wouldn't have bought it and released it. Even if it is tripe, being able to access YouTube is going to be huge.
 
Last edited by a moderator:
maplesugarlover said:
paulguy said:
Well, have fun with that. The VP8 spec is supposedly primarily a lot of obfuscated C code and nonsense x86 assembly.
Wasn't that said by an x264, who is obviously a conflict of interest. This dev spent years optimizing x264. It's not hard to understand why people are trash talking vp8. Lets not judge vp8 until we get our hands on it.
It looks like a pretty good and in-depth analysis, interests or not. Looking at the spec, there is a fair bit of C code, but I wouldn't call it "obfuscated". There is absolutely no X86 assembly code in the spec, his comments on that was regarding the implementation. The reference implementation already includes ARM support, with NEON optimized assembly code.
Of course it wasn't going to be perfect, but the annoying thing is that it seems like many of the problems could have been fixed if they had waited a bit and gathered that kind of feedback before freezing the spec.
 
Last edited by a moderator:
Pandora 2?
Silly.. Why would they make a Pandora 2 when the Pandora is Vaporware, anyway? ;) ;)
 
Ah, guess I misread. Either way, I'm still disappointed in the quality that I saw.

People keep forgetting that h.264, theora or VP8, we WILL be able to use ANY of those on the Pandora!
 
Kramy said:
Theora is okay - it's decent for an MPEG2 replacement - but I value image quality over ease of decoding.
Yeah, but I was talking about Youtube. Youtube's H.264 encoder is notoriously bad, apparently favoring encoding speed over file size. Theora would actually be an improvement in image quality over what Youtube is currently using.

I think youtube doesn't care much about fixing their H.264 encoder since they're planning to move to VP8 anyway. I just hope VP8 works well on the Pandora.
 
Last edited by a moderator:
laharl said:
i said hardware because the article says the VP8 decoder is heavier than the H.264 one. If it's not hardware then we'll need a very fast cpu in order to decode something with more than 720x480.
My point was, why do we need to decode any more than 480p? It's not like the Pandora can show it.
The only reason it would be necessary is for playing existing HD downloaded video without transcoding, and that's going to be H.264 unless you do it yourself.

Kramy said:
VP8 is easier to decode than H.264, and H.264 decodes on most relatively modern hardware.
Currently the VP8 decoder is 13% slower than ffmpeg's H.264 decoder, and CoreAVC is even faster (not that it's relevant to the Pandora).
It's possible it will get faster, but apparently it's already pretty optimised.

maplesugarlover said:
paulguy said:
Well, have fun with that. The VP8 spec is supposedly primarily a lot of obfuscated C code and nonsense x86 assembly.
Wasn't that said by an x264, who is obviously a conflict of interest. This dev spent years optimizing x264. It's not hard to understand why people are trash talking vp8. Lets not judge vp8 until we get our hands on it.
Obviously DarkShikari is going to have some biases, but those comments were pretty objective statements.
The only way they could be biased is if he was flat out lying, and I think he values his reputation more than that.
The analysis is significantly more in depth than "lol it sux x264 r0x0rs" at any rate.
 
Last edited by a moderator:
Aninhumer said:
Currently the VP8 decoder is 13% slower than ffmpeg's H.264 decoder, and CoreAVC is even faster (not that it's relevant to the Pandora).
It's possible it will get faster, but apparently it's already pretty optimised.
Why isn't CoreAVC relevant? I thought they where getting into the ARM business?
Pandora has been so long it's hard to remember what was said, gossip, or fact.
Wait nearly over though :)
 
Last edited by a moderator:
Back
Top