Slashdot Mirror
New VP8 Codec SDK Release Improves Performance
An anonymous reader writes "Google released a new version of the VP8 codec SDK on Thursday. They note a number of performance improvements over the launch release including 20-40% (average 28%) improvement in libvpx decoder speed, an over 7% overall PSNR improvement (6.3% SSIM) in VP8 'best' quality encoding mode, and up to 60% improvement on very noisy, still or slow moving source video. In other WebM news, Texas Instruments has a demo of 1080p WebM video playing on their new TI OMAP 4 processor, in both Android and Ubuntu."
So you'll notice that WebM is getting built in to hardware, just like AVC. Means soon portable/embedded devices will be able to decode it too. Ok so just another format right? Well sort of. You have to pay per decoder (up to a maximum) for AVC and VC-1 and so on. You don't for WebM. So a company is developing really cheap devices, they don't want to pay that royalty. It adds unit cost. Maybe they decide not to, and instead use WebM because it doesn't cost anything. Sure it saves only a few bucks per unit in licensing but that can add up to $5-10 when you are talking sale price and that can be a big deal in cheap devices. Maybe they sell streaming kiosk/info devices that are $40 where the best a competitor does with AVC is $50 or $60.
There is no doubt AVC is here to stay. It has good quality, never mind the massive installed base and standards behind it. Professional (and consumer) cameras are using it for shooting video in the form of AVCHD and AVC-Intra. Blu-Rays are by and large encoded in it these days (you have a choice of MPEG-2, VC-1 or AVC) and so on. It isn't going to die. However WebM may become preferable when cost is key. No encoder, decoder, format, stream, or any costs of any kind ever for any application. That's worth something.
If I ran a video website, I'd seriously think of looking at moving to it once browsers got support. It would ensure that I don't get fucked with fees at some point in the future.
Google follows a really interesting pattern. As far as I can tell, all their software is reactive, rather than proactive.
It is the result of saying "Everyone's using X, but it sucks. We can do it better." They then take a very methodical, PhD-oriented approach to solving the problem. A few parts innovation, many parts simple engineering.
Now they have 10000 employees, but the basic formula hasn't changed. Is there software that Google has made that hasn't been a direct response to an existing product?
That said, I think there's definitely a case to be made that Google is the software industry's first adult. Software's awkward adolescent foibles are on their way out. No more 90s, no millions and millions of VC dollars being spent on Pets.com, no more Netscape and Microsoft working furiously on really terrible codebases adding incompatible nonstandard crap to the internet. No more Myspace, no more Geocities. No more paperclips bouncing around asking me if I'm writing a letter; I'm using Google Docs now.
Google approaches software the way a civil engineering firm would approach a skyscraper: they are actual engineers. They collaborate with academia. They write papers. They sit on the W3C and help create standards. They have architects, PMs, devs, testers, and even lawyers to support their projects.
In a way, this is a sad thing. It was a magical time, when a university student in Finland could just sit down, write a simple OS for x86, and watch half the internet run on it a few years later. When a kid from Texas could create a whole new genre of games in a few thousand lines of C. Sometimes I worry that I was born a couple years too late.
Halfway through my CS degree, I hope that the era of cowboy coders isn't entirely done. It would be a terrible shame if CS became just another engineering specialization. At the same time, Google's professionalism is a breath of fresh air.