Originally Posted by ollebull
Sasha_A, the problem is defnetly not bleeding edge software. My 8600M GT have not perfomed acceptable with any version of gnome/kde/xfce I have tried in the last year.
And from my experince with drivers developing, a good driver shouldn't have to be modified in any huge way to work with new software, the support should already be there as it's "only" an API to the hardware. Well, maybe that doesn't apply to video drivers, I haven't developed any so I wouldn't know realy. Anyway the bug/problem is not new, there have been plenty of time to do something about it but we don't even have a clue if they ever will.
I have tried out the nouveu driver a little and I must say it does perform quite well atm, feels even more responsive than Nvidias driver in some cases. Of cource the driver is only in an early developing stage yet and not ready for daily use, but It's interesting to see that a project built on register readings can create a driver that actually works as good and even better then Nvidias driver in some cases in this time.
thanks for these tips/info.
I too looked into the nouveau (sp?) project a while back in the hopes of providing some feedback or whatever for my hardware, for development purposes, but at the time, they already had current test results for my cards. I never did try the driver though; as I recall (or actually DON'T recall
) I didn't see a driver available for use/testing. I will have to look again.
I too know little to none about driver developing, but I have learned during my experimenting with other aspects of video interaction with the linux kernel, that the nvidia driver is much more than your typical 'kernel driver', and it provides a lot of functionality to a Linux machine (to the kernel) that is otherwise not there at all. For example, for USB devices, there are core built-ins in the kernel (a subsystem if you will), and then each individual type of USB device has a little tiny driver of its own. Same goes for IDE drives, and so on.. For a (nvidia) driver, there is really no such built in subsystem; instead, there's Xorg, a colossus of a program, running on top of the kernel.
I know it is a kind of 6-or-a-half-dozen situation as far as video hardware is concerned with the kernel, but the kernel itself provides VERY little in the way of significant hardware access or functionality when it comes to more than a basic one-card, one-screen setup, let alone acceleration et al.. Xorg and proprietary drivers take care of the lions share of work; from system-boot right till X starts, the best one can hope for is a single VGA/VESA console and/or framebuffer.
Even something which to me (again I have little knowledge about) seems would be rather simple, that being control /usage of both of my video cards (particularly without Xorg involved), is impossible without Xorg and the nvidia driver, even for simple console mode.
My humble opinion is that if there were more work done on the part of the kernel (a subsystem of some kind, to deal directly with the hardware), then the nvidia driver might be able to accomplish more, by having to do less on its own, but at the same time, I understand that there are just so many video cards out there that this is a rather lofty idea.
To clarify this concept I am trying to convey, here's a terrible comparison: pretend that a BIOS is actually the operating system of a computer, and Windows/Linux/Whatever is just a giant module plugged into it... Well, that's kinda how I interpret the nvidia driver situation: Linux kernel is a small part of the equation, but has this giant module plugged into it which is supposed to perform nearly 100% of an astronomical task, not to mention Xorg and a Desktop Environment squeezed in the middle.
Sure seems like a full time job keeping that balancing act working
If nothing else, I hope that this post just inspires further discussion. Maybe I am way off base in the way I see the situation, but again, I can only base what I say on my own experience to date.