Gene: In the good old days that Przemek alluded to, when US$10K-US$100K Unix-based workstations were being sold just because they could run even more expensive CAD/CAM software and there were only a few choices being offered in graphics hardware, the software driver situation was barely tolerable. Yes it took a lot of sweat equity to hew a solution out of the 'oak,' to use Przemek's metaphor, but it worked forever once done. That's the era in which I developed a true love/hate relationship with X-windows.
In the world of Linux on PC-based hardware the situation is totally intolerable. Neither the hardware nor the software costs anything, the technology turns over in 6 mo to 18 mo, and there simply is no money for driver development. Even in the Windows gaming and multimedia arenas, where all the profits appear to lie, the drivers are constantly being tinkered with because they are put together with baling wire and chewing gum to begin with. Every new application reveals yet another problem with the drivers. The emergence of LCD technology has screwed them up too. The VESA specification guys have tried to keep up but .... Unlike graphics card development, which can be done on a project-by-project basis, software driver development is continuous. I can't imagine product managers cheerfully paying for lots of software developers. Figure it costs a company US$100K to run one good, full-time software developer for one year (yes, I'm including overhead and benefits). It would just come out of the managers' annual bonuses. The reason "professional" graphics cards cost so much more than "consumer" cards is because of the software-driver development costs, not the hardware. Making OpenGL, rather than DX, run well on them is a particularly high-cost item. I haven't even mentioned designing for acceptable real-time performance, which is a non-issue for 99.9+ percent of the buyers and, hence, of the sellers. Bottom line---I don't think you should hold your breath waiting for better drivers in Linux that work well with EMC2. The market forces are all wrong. To add to your misery, every Linux distribution appears to go its own way on its X-server and graphics drivers, so you can't be sure that what worked in Ubuntu, say, will work in PCLOS, etc. "Herding cats" is the metaphor that comes to mind. At least we could start a spreadsheet on the wiki to aggregate information like card, m/b, driver, distribution, etc., along with some crude measure of performance. It could even be an expansion of the existing latency-test spreadsheet, although I believe we would be better served with a separate one. I expect it would be mostly a place to say "this works for me," "this doesn't work for me," and "watch out for these gotchas." Regards, Kent ------------------------------------------------------------------------------ uberSVN's rich system and user administration capabilities and model configuration take the hassle out of deploying and managing Subversion and the tools developers use with it. Learn more about uberSVN and get a free download at: http://p.sf.net/sfu/wandisco-dev2dev _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
