On Sunday 20 May 2001 06:18 pm, Eric George wrote:
> Tom Brinkman wrote:
> > On Saturday 19 May 2001 03:27 pm, Eric George wrote:
> > > Thanks for the links.
> > > I ran burnp6, burnmmx & burnbx for several hours each with no
.               ~~~~~~~~~~~~~~~~~~~

   these are tests for a Pentium class cpu with a BX chipset 
motherboard, which is why I slanted my response towards diode readings.
Last time I looked, there are K class tests included in cpuburn.
Try running 'burnK6'  

> > > problems. CPU temp never went above about 52.5 celsius.  Room
> > > temp was about 23 C and the idle temp was about 45 C.  The case
> > > temp actually read a little less then room temp, probably just
> > > calibration issues.

> I don't thinks so.  These new athlons run hot!

   Yes, Pentiums fail at ~65C, AMD's at ~90C  BUT, both x86 processors 
experience hickups at temps well below they're fail temps.  More like 
~50C plus a spike for both. 

> I follow the forums on Tom's hardware guide, and this question comes
> up quite a bit.
> Most people are running high 40's to low 50's probe temp under full
> load on higher clocked TBirds.
> The stock HSF tend to only get you around 60c under full load.

    Your temps have to be from a probe... you must add ~20C to be on 
the safe side. Even then thermistor should be at least in firm contact 
with the center of the die, lightly greased, and shielded from the 
hs/f's airflow.  That Tbird could be hitting 70+C and is getting into 
the lockup-reboot zone, if not close to failure. Which is the most 
common RMA reason for returned Tbirds.

>
> And as I said in another post, I've run it under full load for weeks
> under RedHat 6.2 with no problems.

   I agree, I've noticed a marked increase in cpu/cache/ram loads with
progressive upgrades of Mandrake 7.2, and even more with 8.0. It's not 
Mandrake I believe, but the increasing intensity of XFree86, KDE, 
Gnome, and such to become more like Winblows to the average Joe so he 
can be won over. Dubious, IMO. 2.4.x kernels don't seem to be helpin 
either ;(
>
> I'm running VolcanoII lapped, with ArticSilver.

    Both are not what's needed. Silver is a gimmic, Volcano is 
substandard. The $2 grease from Radio Shack is all I've ever used.  
Thermaltake hs/fans suc.  GlobalWin, or an Alpha, not Orbs. Still your 
max temp should be kept under 50C, even if you use a $2 no-name 
heatsink with a table fan pointed at it.
 
  Keep in mind that temps vary (internally) very drastically. Often 
10+C in a split fraction of a second. What you're seein' from a probe, 
is a guess at best, an' won't indicate temp spikes (which cause random 
reboots).  It's like tryin to measure the temperature of the wiring in 
a wall, by attaching a thermometer to the sheetrock.  Almost useless.  
AMD is finally getting up to speed, the new Athlon4's (Palomino's) will 
have an internal diode to monitor actual core temp accurately.

> >     I just d/l'd  memtest86-2.5.tar.gz  and ran 'make' which built
> > the binaries, then 'make install' which copied them to a bootable
> > floppy on an 8.0 system (?).
>
> Yea, when I try 'make' I get the error I mentioned :-(
> Haven't tried mprime yet.

    No idea on memtest? It's a small deal, builds in a second. mprime 
is as good a test, better in some ways, but you need to let it run in 
the background for a day or so. Then look at the output for any error 
lines. 

   Eric, not to be critical, but most problems in my experience are 
best approached as user - hardware - OS, in that order. The problems 
you're havin are not very likely due to 8.0 at all. If there's no 
fundamental hardware problem (eg, heat), then configuration is the next 
likely cause of locks/reboots.  Ram timings, L2 cache latency, errant 
bios settings and such. 'Course with the Tbird, power supplies are more 
critical too. Watch the voltage outputs from lm_sensors. Particulary 
the -12v, -5v readings. If they're low/vary, it's a wobbly PS.  Also, 
IIRC, you have a kk266 ?  In any event, whatever motherboard, it's a 
good idea to bump up the IO voltage (3.3v) on any mobo that supports 
it. Most of the better ones have increased IOv by default. This will 
greatly improve ram and overall motherboard stabiltity an' performance. 
3.5 to 3.7 volts IO is the preferred. Also try your ram sticks in 
different slots. If you've got more'n one stick, try putting 'em in 
different order too.

   Next thing to look at would be what pci cards you have in which 
slots. Might need to move those around by trial'n error also. Look here
http://www.fullon3d.com/kk266faq/kk266faq.shtml#about

All'n all tho, random lockups-reboots are most often a hardware problem.
-- 
Tom Brinkman      [EMAIL PROTECTED]     Galveston Bay

Reply via email to