Randy Carpenter wrote:
>
> > > Thanks ... (I have not yet experienced anything weird)
> >
> > The best description is that the BP6 is simply not stable with dual procs
> > under Linux, regardless of kernel or hardware. Some circumstances are
> > more unstable than others, but taken as a whole, it's just a no-go.
> >
> > Single-procs seem to be solid, however.
>
> I have never had a single problem that wasn't related to a bad CPU. I've
> had uptimes of 60+ days running dual 366's at 550, running X, and plenty
> of cpu intensive apps, like multiple VMWare's or seti@home clients.
> There has to be some explanation for people's troubles. Most of the
> evidence I have seen points to problems with video cards.
>
> -Randy
>
Normally, I don't get uptimes of more than 12-14h when I have holidays
or 6-8h the other 46 weeks - as I'm too much concerned in ecology
(and my electricity bill, too ;-) ).
So I did only a short (~1.5 h) stress test yesterday with the
newly-installed
SuSE 6.3 (kernel 2.2.13 with the "6.20 eridanus uniform IDE" patch that
applied
cleanly - BTW: what's the difference between "eridanus","draco" and
"hydra" ?);
I've got 2x Cel. 466 not ocl., 3 IDE HDs attached as PM to ide0, 1 and 2
(HPT 366),
SYM8750SP U-Narrow SCSI contr. with another HD, CD-ROM, CD-writer and
DDS1 tape,
RTL 8139 PCI Fast Ehernet, Fritz PnP passive ISDN, Hauppauge WinTV PCI,
SB16 PnP,
video card is Matrox 220 "Business" PCI. I disabled APM and ACPI in BIOS
except for
DPMS and also disabled all shadow rams as well as not assigning an IRQ
to USB
(I don't need that stuff for now, but running short of IRQs you see). So
that's what I
did to stress at least the PCI and ISA buses a little :
- Compilation of XF86 3.3.4 (on the system's root part. /dev/hdc10)
- "Tarring" ~800 MB from hde1 to tape
- Copying ~ 1.8 G from hda7 to another box via NFS
- and simply watching TV ;-). Obviously, I was doing that in X.
So I had 5 of my 6 add-in-cards plus 3 onboard controllers working
simultaneously;
xosview most of the time showed 4-5 IRQs, load rose (only ?) up to 3.4.
I admit that few times tenth a second of sound from TV appl. was lost,
and I admit
the compile & backup times weren't too famous - but apart from that, I
noticed
no problems, especially no locking or something, and the log files
stayed clean.
What I realized, though, was that the freshly-compiled "sensors" (thanks
Robert!)
running in a loop showed temperatures up to 73/56/57 deg. Centigrade
(CPU0/1/system) !
So I better get an additional case fan soon (luckily, I have a spare one
in another
box) and/or (pardon my English) cooling cream !
My opinion after that admittedly more I/O- than CPU-intensive test is
that the BP6
is quite _stable_ (at least stable enough for me...) and I see no reason
I should
regret having bought it 5 weeks ago.
BUT, yes, I had those locks before I rearranged the PCI cards : before,
the TV card
was in slot 3 (the one sharing the IRQ with the HPT366) - and I could
reproducibly
"shoot" the system by a simple attempt to mount a part. on /dev/hde
while
watching TV... - now, my Mystique in slot 3 _not_ assigned an IRQ in
BIOS seemingly
doesn't make problems.
So, maybe those lockups have something to do with IRQ sharing ? (Just
guessing).
So long
Juergen
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
