New nvidia beta driver: 180.11

To install, do: sgfxi -co 180.11

We'll see if this addresses the lingering desktop memory leak issues which has so far made the 180 series utterly unusable for me.

release notes
Back to top

Just to add that I have also found the 180 series to be unsatisfactory. It has noticeably slowed down 'normal' desktop use.
Back to top

Howdy,

I have not tried the 06, but have used both the 08 and 11 on a fresh sidux install running fluxbox, latest slh kernel. Both are slow, or at least no faster in bringing up all graphics (4 conky's audacious, and wallpaper) from a blank root window with a toggle. As a separate question:
Would a different kernel be better at this for any reason? and, is it ok to use a different kernel running sidux. I think it is time for me to make a few changes.
Back to top

I don't think the sidux kernel is the problem, but the latest 'testing' nvidia.
I'm using a MEPIS and sidux 2.6.27-7 kernel on antiX, fluxbox and icewm, and the nvidia 8 series is awful. It makes fluxbox run like kde with low RAM ;)
I went back to the 'stable' 177.82 nvidia driver and all is ok.



Edit: Now trying the 180.11 driver and it seems to be ok, at the moment. We''ll see how it fares later on today.
Back to top

sgfxi 4.1, updated all kernel/xorg detections to be more dynamic, and removed some driver cruft for nvidia so that sgfxi can move ahead another year.

All should work now without much manual updating of the detection parameters for kernels/xorg versions.

I'm starting to implement more my realization that any time bugs creep in, you should ideally not only fix a bug, but fix the cause of why the code made it easy to miss, or create, the bug in the first place.
Back to top

sleekmason, any performance issue with nvidia drivers is unrelated to the scope of sgfxi itself, but need to be dealt with on the nvidia support forums. Make sure to read the how to file bug reports thread there if you do post issues.

A driver being the same speed as the last one is neither a bug or a problem, in this case, if you see the bug, it reveals itself as a gradual memory leak, easy to see with firefox/iceweasel with many tabs open, it simply slows down your whole desktop, window painting, etc, until it's totally unusable. This is not subtle. This is the matter in question re 180.xx drivers.

Re kernels, they don't matter, if your hardware is working fine with an older kernel, expect it to do one of 3 things with a newer kernel, in order of likelihood: 1 - work the same; 2 - work slightly better; 3 - stop working, or start working buggily.

There's a reason major distros all stick to one or two kernel versions per release, and that's part of the reason.

I saw some good improvements in file transfer speeds, usb handling, in I think 2.6.26, that's the last time I noticed any significant change.
Back to top

excellent. Just what I wanted to know. I will pay more attention to the topic next time, and will also attempt to find a way to do a simple benchmark for the future.
Back to top

Version 4.2.0

New feature and option: -B

This will check for latest nvidia beta driver, and use it if one is available. If none is available, it will use the latest stable driver for your card type.

This feature is now also available via smxi, but smxi will not know if a beta driver was actually available, sgfxi will however always tell you if it's using the beta or the stable when you launch it.
Back to top

Also, a small refactoring of sgfxi option variable names. If you're using an sgfxi options config file, you'll need to update all boolean true false option names in the config file to start with B_<option variable name>.

like so:

This change only affects Boolean variables, ie, ones that are set to 'true'. All other variable names are unchanged.

I needed to do this to make the overall script easier to read, and maintain.
Back to top

Evening, You may wish to double check using the -cB option. The script showed me to be loading the latest stable driver with tripple buffering enabled. This being the lowercase -b option I figured you may have had a typo. Could have been my deal but I checked it twice. Using -o still works.
Back to top