While I find the Phoronix survey's pretty interesting, it's important to keep in mind that they're highly biased (because the respondents are self selected) and not scientific. They represent the small minority of folks who actually read Phoronix and care enough to respond, not the much larger majority that doesn't worry about the minute details of what's going on in the graphics world.
I would imagine that's one explanation for why binary blobs might be oversampled - that the people who are reading Phoronix tend to care more about 3D performance and getting good hardware than the average user. Or maybe not, the point is we really can't know.
Wayland, Gallium3D, OpenGL 2.1/3.0...cool concepts, but with progress like this we will never see useful graphics or good productivity done on Linux.
You mean the development is too slow? I think its just getting momentum. With wayland we are really close. Now it just needs to be taken seriously by the rest of the community and the industry. Otherwise it looks quite down the hill...
Last edited by misiu_mp; 11-16-2009 at 03:36 PM.
Reason: Clarified what is too slow
I hate having to get pedantic but there's a pattern going on here. One person talks about open source drivers being "good enough", then another person replaces "good enough" with "sucks" or "slow" (ie words with completely different meanings) then disagrees with the original statement. Unless everyone just enjoys arguing we're going to have to be very careful with choice (and substitution) of words.
I'll toss out a sample definition for discussion :
If a driver plays the games a user normally wants to play, at the native resolution of their display, at the refresh rate of their display, a lot of people consider that "good enough". The fact that another driver can play the same game at 377 FPS doesn't make much difference because the display can't show the additional frames anyways.
Within a generation of GPUs there is typically a 3:1 performance jump between GPUs in the same family, or ~9:1 from top to bottom. That difference swamps any difference between drivers, ie a midrange card running open source drivers will normally be quite a bit faster than an entry level card running binary drivers.
What's missing right now is support for some GL features which either prevent games from running or force them to slower code paths - and adding support for those features is pretty much the top priority for the devs working on 3D code today.
I completely agree with the post above. I was a bit worried when I thought of how the opensource drivers may reach 75% of the binary's performance at best, but to be honest, I rather buy a new VGA (and maybe get even more performance) than to keep my old one and continue on using proprietary drivers. It will be very interesting to see how games like ETQW will run when the Gallium3D drivers are in a good shape in late 2010/2011.
I hate having to get pedantic but there's a pattern going on here. One person talks about open source drivers being "good enough", then another person replaces "good enough" with "sucks" or "slow" (ie words with completely different meanings) then disagrees with the original statement.
Arguably there might be some right in doing so. After all 'good enough' feels less good than 'good'. 'Good enough' for some people, is more likely to be not good enough for more demanding people than a thing that is 'just good' or 'very good' (think normal distribution). The key is to achieve a 'good' value so that the displeased are a small enough percentile.
Besides 'good enough' does sound like something you really are not pleased with, or expect not to be pleased with in the future.
"I would really like to drive dad's Koenigsegg, but he only lets me drive the VW. Well I guess thats 'good enough'. After all it will take me from A to B and I cant drive faster than 90 anyway."
Maybe we should use the term 'satisfying'. It seems less prone to erosion.
Originally Posted by bridgman
a midrange card running open source drivers will normally be quite a bit faster than an entry level card running binary drivers.
Hmm, sounds like "good enough" has bad connotations for you. That was not my intent when using the term - I meant it literally, ie that it meets your needs. For most people "good" is a point on the continuum between sucky and spectacular whereas "good enough" is relative to their requirements.
Two generations of advertising have twisted us into believing that "good" is for losers and the only things worth having are "excellent" or better, and it may take another couple of generations to undo the damage. That's another reason for using "good enough" rather than "good"
The car analogy is problematic, since there are other reasons than speed to drive the "nice car". A more accurate analogy might be two identical cars, one with a governor set to 125% of the highest speed limit in your area and another with a governor set to 150%. In normal use you would neither see a difference nor care about it.
Anyways, that's what "good enough" means to the open source developers. "Satisfying" would probably work as well.
When I hear "good enough that users will keep the default", I immediately think of IE. And that's not something you want to bring to mind.
I think there's a difference here, though. IE just sucks at everything compared to it's competition, while the OSS drivers should have a significant edge over the binary ones in at least a few areas. So with the drivers it's more of a tradeoff where you have to pick out what matters the most to you.