On Monday 27 May 2002 11:54 am, Jeferson Lopes Zacco wrote:
> Tom Brinkman wrote:
>> If you still have freezes the next thing I'd try is renaming
>> XF86Config-4 to somethin like XF86Config-4-nvidia, and enable the
>> original XF86Config-4 to use the nv driver. You don't need to
>> uninstall the nvidia GLX and kernel packages. Just see if the
>> freezes stop when you use the open source driver. If they do, then
>> the nvidia binaries are the problem. Are you usin ones that were
>> built against the kernel you're usin 'em with?
> I build all my drivers from tar.gz, BTw there is a new driver set
> (2960) but I will postpone the installation till I fix this.
Just 'cause they're 'tar.gz' doesn't mean you can compile the
nvidia binaries for your system. The source in the tarballs is just
a wrapper around nvidia's closed source precompiled binaries. If
there really was any useful source in those tarballs, XFree86 could
support 3d accel for nvidia cards, and we wouldn't hav'ta depend on/
trust nvidia to get it right.
> really do not think it's an Nvidia driver related problem; it never
> occurs when I'm using a particular graphic intensive app. But you
> might have a point, though. One thing I was thinking about that
> matter is that I compiled the kernel with kgcc (only way I managed
> to recompile Mandrake's kernel) and the drivers with gcc. But then
> why oh why it WAS stable?
Like I said above, you didn't compile the nvidia binaries.
Actually IME, Mandrake nvidia src.rps usually provide better results
than the vanilla tarballs. The main reason I suggested changing
XF86Config-4's to use the open source 'nv' driver was just to
eliminate that possibility. The closed source nvidia driver is a well
documented source of system freezes. I've experienced 'em with some
kernels, sometimes, as many others have also, and have reported on
the linux-kernel mailing list, along with conflicts with other
software.
I keep several XF86Config-4's in /etc/X11/. I save the original
as XF86Config-4-nv, and when I edit a copy for nvidia, I save it as
XF86Config-4-nvidia. I only use the nvidia 3d accel occasionally
(when testing the latest FlightGear release). I often change kernels
and compile my own, as I saw you do also in your original post. I've
found the nvidia binaries to be very fussy about the kernel. Often
the most apparent symptom is problems with freezes. A simple
'mv XF86Config-4-nv XF86Config-4' (ie, going back to the open
source XFree86 'nv' driver) fixes 'em as soon as X is restarted ;)
> The first thing I thought was that the fan had died, but I opened
> the case and it was running fine. I have two fans pulling air
> inside and one out, and my CPU runs rather cool - never more than
> 40C. Besides it's quite cold now here (well, not european cold) so
> its temperature is about 36-37C. On the other hand, there is a lot
> of dust in there for sure; I'll clean it over before I do the
> memory test.
Clean the dust out for sure. You should do that regularly anyhow.
Screw a cheap little 'cpu type' fan on the nvidia cards heatsink if
it doesn't already have one (it needs one). If the ram passes
memtest86, then see if the system can run cpuburn for at least an
hour ( http://users.ev1.net/~redelm/ ). The cpu temps you're
reporting are from an external probe, are only a guess at best, and
could possibly be very much lower than actual internal core temps.
At least 10 to 20C lower, maybe as much as ~30C. Bad ram can
sometimes pass all memtest86 tests with flyin colors, but I'll bet a
weak system can't run cpuburn ;>
--
Tom Brinkman Corpus Christi, Texas
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
