Overclocked Sempron 3100+ won't overclock anymore...

[Brian K]

I have a Sempron 3100 socket 754 on a fairly generic Nfoce3 motherboard.
ECS, I think was the manufacturer.

It has a gigabyte of 266mhz DDR ram installed, and I run 64-bit Mandriva
linux almost exclusively, with brief forays into 32-bit Windows Xp for
gaming purposes and 32-bit Mandriva for testing purposes.

The only overclock change I did was to change the cpu bus speed in the
BIOS from 200 mhz to 240 mhz. Bus speed might not be the correct term,
but I can't remember offhand, and I'm at work while the machine is at home.

That gave me a CPU speed of just over 2100mhz, according to /proc/cpuinfo
and various Windows-based utilities.

It has run happily 24x7 for months and months and months with a nice big
Volcano heatsink/fan and lm_sensors hourly logs shows the CPU temperature
averaged about 40 degrees celsius, with occasional jumps up to 55 or so.

Anyway, to make a long story short, the machine locked up tight for the
first time ever yesterday, and upon reboots would hang at random points
during the boot process for Mandriva, reporting kernel panic, or no memory
found, or any number of other unrelated and completely confusing errors.
Windows XP boots with a STOP 0x0000007b error.

I initially suspected a disk failure but quickly eliminated that as a
possibility, after the system failed to even boot to a BIOS memory test
screen after a reset, and even after a hard power cycle.

I cleared the CMOS with the motherboard jumper, and it boots and runs
fine, but if I overclock beyond a few mhz the system fails to boot at all.

So, what could possibly have happened? I'm not all that concerned with
the few MIPS or FPS that the additional 200 mhz had supplied all those
months, but I'm curious from a technical standpoint why a machine that was
rock-solid for probably close to a year would suddenly stop working...

 

[Wes Newell]

I have a Sempron 3100 socket 754 on a fairly generic Nfoce3 motherboard.
It has a gigabyte of 266mhz DDR ram installed,

The only overclock change I did was to change the cpu bus speed in the
BIOS from 200 mhz to 240 mhz. Bus speed might not be the correct term,
but I can't remember offhand, and I'm at work while the machine is at home.

You should also have lowered base HT link speed and base ram speed.
Without that, you also over clock them when you raise the system clock to
24Mhz.

That gave me a CPU speed of just over 2100mhz, according to
/proc/cpuinfo and various Windows-based utilities.

Yep, 9x240=2160

It has run happily 24x7 for months and months and months with a nice big

Anyway, to make a long story short, the machine locked up tight for the
first time ever yesterday, and upon reboots would hang at random points

I cleared the CMOS with the motherboard jumper, and it boots and runs
fine, but if I overclock beyond a few mhz the system fails to boot at
all.

So, what could possibly have happened? I'm not all that concerned with
the few MIPS or FPS that the additional 200 mhz had supplied all those
months, but I'm curious from a technical standpoint why a machine that
was rock-solid for probably close to a year would suddenly stop working...

Stressed ram and/or chipset HTlink too long and it gave out. CPU may have
gotten weaker. Could try some extra voltage there, but first I'd try to
clock the ram and HT down before raising the system bus. Set ram to a base
of 133Mhz, and HT link to 3x (600MHz). Raise cpu voltage by .1v and set
system clock to 250MHz and give it a go. Should get 2250MHz. If that works
lower cpu voltage and try without the extra voltage. Do all these test
with the hard drive disconnected so you won't screw up data on it if it
isn't already. Boot a memtest boot floppy or cd til it passes, then test
again with linux burnK7.

 

[Brian K]

You should also have lowered base HT link speed and base ram speed.
Without that, you also over clock them when you raise the system clock to
24Mhz.
Yep, 9x240=2160


Stressed ram and/or chipset HTlink too long and it gave out. CPU may have
gotten weaker. Could try some extra voltage there, but first I'd try to
clock the ram and HT down before raising the system bus. Set ram to a base
of 133Mhz, and HT link to 3x (600MHz). Raise cpu voltage by .1v and set
system clock to 250MHz and give it a go. Should get 2250MHz. If that works
lower cpu voltage and try without the extra voltage. Do all these test
with the hard drive disconnected so you won't screw up data on it if it
isn't already. Boot a memtest boot floppy or cd til it passes, then test
again with linux burnK7.

It appears the ram is the culprit. It's all Kingston 400mhz DDR ram, but
it will only run at or below spec now.

I overclocked to 2250, as you suggested, with memory speed locked at 133
mhz, and it boots fine. I set the lock to 166 and it runs fine, at 409
effective mhz. AUTO causes it to crash.

Thanks for the advice.