My horrible experience with ATI
I have a rather old system with AGP graphics (ASUS P4P-800) & nvidia 7600GT. I only boot into Windows once in a while for games and i wanted to replace my nvidia card with something more powerful, more current. So i bought the AGP version of ATI's HD 3850 (RV670).
Now, not only did it not work with kernel 2.6.27 out of the box (Catalyst 8.10) but it also gives me bad performance and unstable FPS. The only improvement Catalyst 8.11 brought to the table was, that at least it compiled against 18.104.22.168 without having to patch anything besides the "tainted" license.
I won't even go into the details of what i went through to get OpenGL and direct rendering to work. The "installation process" of ATI's drivers is atrocious compared to nvidia. I spend hours on end to find documentation, HOWtos that weren't outdated etc. and still can't switch to a text console, switch from a full-screen OpenGL app to any other process without screen corruption AND get bad FPS.
Frankly, i'm p***** off beyond words. I'm hoping to buy a new rig early next year and I will sure has hell never ever touch anything ATI, not with a 10-foot pole. Meanwhile my old graphics card does a better job, gives me at least decent, constant FPS in OpenGL games and installation of new drivers is easy as pie.
Well the install of ATI drivers can be a bit challengeing, even for experience users, when some things break from one release to the next one. I have got no idea how many people test those drivers, but at least they test different things than I do and my tests often show failures of the driver of it's installer. For a new box Nvidia is certainly a better choice. That's why I would _never_ recommand buying a AGP upgrade card, much better is to replace the rest too. Thats 200 € about more expensive but you can choose a much better card.
Try to use 8.9 version of driver. I am not sure about 2.6.27 (Is that stable Fedora release?) but 8.10 and 8.11 are useless for gaming
Originally Posted by Fafnir
for me using my low end X1250, at least.
Much as I appreciate Bridgeman's presence here, ATI fglrx drivers will in my opinion continue to be broken by design until the two following consistent problems are fixed:
1. the monthly release needs to be dropped in favor of a more rational release model, ie: release it when it works. See nVidia stable and beta release method, coupled with timely, well written patches, for an example of a method that works, is proven, and fits how most of the professional software world works.
2. stop releasing broken installer scripts in the package. If ATI cannot test an installer script and verify that it works for the distro, it should not be allowed to be in the package in the first place. That anyone even has to say something this basic and obvious is really astounding to me. I cannot understand how AMD/ATI expects anyone to take them seriously when the code they release in their packages, even if they do not write it, doesn't work. Why is this so hard to understand?
I said about 1 year ago that I'd wait to see how ATI was doing last summer before drawing any final conclusions, but the problems all persist, ATI continues to release code that does not work, and rather than either fix it or remove it, they make excuses.
Nobody is interested in hearing more ATI excuses I'm sad to say.
On the bright side, for any very low end requirement, ie, server/office machine video card, etc, the cheapest ATI card you can find, plus the free radeonhd drivers, are now an acceptable option, but you are in my opinion throwing your money away if you spend more than $40 on an ATI card for Linux, not until the first two points are addressed would I consider recommending ati cards to anyone, except as basic desktop display if the mobo has no built in graphics, or if the xorg driver does an adequate job for your needs.
After watching ATI for years now, I feel fairly comfortable narrowing the problems down to the top two points, both of which to me indicate ongoing procedural errors in how ATI is developing their drivers. Excuses are not interesting to hear any longer.
Last edited by gfxdrone; 11-19-2008 at 03:21 PM.
+1 ATI needs a more robust release strategy for its driver
Originally Posted by gfxdrone
Actually there is a fairly large beta test group which does try out the drivers before you see them.
Originally Posted by bridgman
Attacking the beta testers for being incompetent in 3... 2.. 1...
Nope, not at all. They do find problems and we usually get the problems fixed before release.
The reality here is that we will probably follow the same path as all the other Linux driver vendors -- gradually accumulate functionality and compatibility over time (we've come a long way in the last year) then lock in that behaviour and keep it good going forward. Our testing focus is still relatively heavily biased towards CAD workstation users although we are increasing consumer coverage over time.
Maybe you live in a parallel universe. That's definitely not true as I always find errors, just times I am too annoyed to repeat me everytime and nothing gets fixed. 8-11 release showed definitely that _nobody_ cared about a fixed Ubuntu build script - as I use it in my script I found it out the hard way - and fixed it. Also there are lots of other error in there which have been introduced some month ago or even more than a year ago. That's absolutely nothing I would call fixing problems.
Originally Posted by bridgman