I had this same problem on two machines running dual Athlon MP's. Earlier
versions of Mandrake and other distros would lock up a lot. Some distros and
versions would not even install. One thing that I tried was swapping the
processors so that the voltage leveled out. I had read somewhere that there
is a problem sometimes when running dual MP's where the voltage will be
different to each processor. This helped a bunch on one system.
Secondly, installing Mandrake 9.0 and using the default graphics drivers have
caused no lockups. I have tried Nvidia's drivers and have had problems but
maybe I was just setting that up wrong. But MDK 9 RC2 default everything
(kernel, drivers, etc) has worked great for me.
-Gary
On Monday 23 September 2002 06:55 am, Tom Brinkman scribbled something about:
> On Sunday September 22 2002 01:09 pm, Charles A Edwards wrote:
> > On Sun, 22 Sep 2002 09:37:15 -0500
> >
> > Tom Brinkman <[EMAIL PROTECTED]> wrote:
> > > For more on the AMD/AGP issue see,
> > > http://www.geocrawler.com/lists/3/Linux/35/175/7626960/
> > > Note: "There *is* an Athlon/AGP issue. This issue has
> > > *not* been tied to a bug with the Athlon/Duron processors."
> > > ...and...
> > > "Our conclusion is that the operating system is creating coherency
> > > problems within the system by creating cacheable translation to AGP
> > > GART-mapped physical memory."
> >
> > This is the same AGP 'bug' for which AMD issued a Win2k patch and
> > this 'bug' exists in the
> > Linux 2.4 kernel series unless the kernel has been built for Athlon
> > optimization.
> > I may be be mistaken but do you not build your kernel, Tom?
>
> Yes, currently 2.4.19-16k7. I only edit Mdk's default .config
> file and comment out i585 and enable k7, compile as usual. I had no
> idea this was a work-around for the AGP issue tho.
> Thanks for the info ;)
>
> > At any rate there are numerous threads on the kernel mailing list
> > regarding this issue when AGP is used and the always recommended
> > solution is to use mem=nopentium.
> >
> > In support of my position see:
> >
> > http://www.kickassgear.com/athlon-linux%20bug.htm
>
> Interesting, but it directly contradicts the statement made by 'Sean
> Cleveland and Wayne Meritsky of AMD' (the kernel ML link I gave). I can
> see why AMD could release a patch for W2k, even tho the bug isn't in
> their cpu's. Also, I believe the Linux 2.4 kernel has been patched
> too, for quite some time now. IIRC, shortly after Mdk 8.0 was
> released.
>
> > If others are desired just do a google-linux search for
> > mem=nopentium.
>
> That's where I got the kernel ML link ;) BTW, your link is just a
> hardware (overclockers) webpage, and is just the authors opinion. ie,
> "Here are the details.", but then doesn't say where he got 'em.
>
> > I currently use mem=nopentium on my systems and can confirm that
> > without this option random X lock-ups will occur if using the Nvidia
> > driver, and on systems with the ATI Rage128 using either the ati or
> > r128 driver if DRI is used.
> >
> > Even should men=nopentium be only a placebo I'll still take my
> > medicine.
> >
> >
> > Charles
>
> Well, besides compiling for Athlon, I don't use nVidia's closed
> source drivers either. One or both are probly why I don't need the
> nopentium parameter, and don't experience lockups.
>
> Hardware problems, eg, mobo/PSU, ram, overheating are the most often
> causes of random lock-ups, spontaneous reboots. Those areas should get
> a real hard close look before tryin bandaids IMO. 'Course, as I infer
> from your experience, if lock-ups stopped, then the nopentium option is
> probly needed, either for your hardware or to rectify a bug in the
> video drivers you're usin. ...and as you seem to say, probly doesn't
> hurt.
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
