Spontaneous shutdowns in new AMD64 system.

[Brian]

I recently built a gaming machine for a friend of mine that wanted to play
World of Warcraft without selling a kidney to buy a machine. It worked
fine at first, but lately it's gone south. And I'm out of ideas at this
point as to what might be wrong with it.

Spec:
AMD Athlon 64 2800+, stock retail heatsink.
MSI K8N Neo, nForce 3 250
1GB RAM
WD 80GB SATA
LG CD-RW/DVD-ROM
GeForce FX 5200

Nothing is overclocked, tweaked, modded, or otherwise altered from stock.

Here's the problem. It'll boot just fine, and run for 5-10 minutes.
Everything looks fine, but then it just *stops*. Full system shutdown,
like somebody pulled the plug. Except that the power LED on the front of
the case blinks on and off. And it won't restart until/unless I hit the
cutoff on the PSU or unpug the machine and plug it back in.

My first thought was the power supply, since the case it's built in had a
*very* cheap PSU. So I went and bought the overkill 480W PSU. No change.

I thought it might be heat, but I checked the HW monitor in the BIOS, and
it was only at around 100F when it shutdown.

So I'm stumped, and my friend really wants his game machine back. What
should I check/replace next to get this machine back on its feet?

Brian

 

[old jon]

Admin Tools\Event Viewer, does it show any errors\warnings ?. come back.

 

[Brian]

"Bother!", said Pooh, as he read old jon's latest post to
alt.comp.hardware.

Admin Tools\Event Viewer, does it show any errors\warnings ?. come back.

Nothing in the last couple days. There was an error message from a few
days back, but I couldn't get the computer to stay up long enough to
copy/paste it.

I am 100% sure this is a hardware issue, though, not software. I was
poking around in the BIOS, trying to see if there was anything wrong in
there that I could see, and it did the exact same shutdown behavior from
the BIOS config screen. No OS loaded means that the OS has nothing to do
with the problem.

Brian

 

[Jamie]

do the simple stuff, reseat your RAM and Processor, check the Rails on the
New PS. make sure all your cables are connected correctly etc.

 

[General Schvantzkoph]

do the simple stuff, reseat your RAM and Processor, check the Rails on the
New PS. make sure all your cables are connected correctly etc.

You may also want to reseat the heatsink on the CPU. Remove it, clean it
off with a solvent, use fresh heatsink paste and replace it.

 

[Joe]

Dang I could have written your post a few weeks ago. I had a computer doing
the exact same thing although it was an old Duron1200 I use as a second
computer. I tried the exact same things you did. I swapped the PS with a
Known good one and checked temps which were fine and the problem persisted.
Well in my case the computer was overdue for an upgrade anyway so I bought a
cheap Socket A board and a Sempron 2300 for $90 installed them and my
problem is gone. I am back to using the original PS and all other components
even the heatsink and fan from the old one. The only two components that are
different now are the motherboard and CPU. I do not know which of the two
was the problem but it was obviously one of them. I was also certain I did
not have any problems with something not seated well or something to that
effect as my computer had run for over a year like it was. In your case I
would say you have a bad MB and \ or CPU.
Also I knew it was not software because I could not keep mine running
long enough when booted from the winXP CD to format the hard drive.
I do not know if this helps or not but good luck.

Joe

 

[Brian]

"Bother!", said Pooh, as he read General Schvantzkoph's latest post to
alt.comp.hardware.

New PS is 215W combined on 3v and 5v, so it should be plenty. I've
stripped the computer down to just a post config (video, ram, CPU, no
drives), and I'm playing mix-and-match with ram sticks and slots.

I took it down to one stick instead of two. I've tried both sticks
individually in slot one, I guess after this I'll try slot two.

I also tried a known-good AGP card, GeForce3 Ti from an older Duron-based
system that's been running rock solid for years. No change.

You may also want to reseat the heatsink on the CPU. Remove it, clean it
off with a solvent, use fresh heatsink paste and replace it.

I swapped the stock HSF for a Thermaltake and some Arctic Alumina. In
about 5 minutes, it'll crash again, unless by some miracle I've found the
problem.

I'm watching the PC health screen in the bios, suspecting heat. It booted
at 96F, and after a couple minutes it's crept up to 100F. Which should be
*far* short of hitting any safety systems.

And it just crashed again. Zero change in behavior.

What next? Buy a new motherboard?

Brian

 

[Fitz]

I would take a look at a memory issue. If you memory is 2 X 512MB, remove
one stick and see if the system runs. If so, you might try increasing the
memory voltage by .1 and trying both sticks.

Fitz

 

[larry moe 'n curly]

I recently built a gaming machine for a friend of mine
It worked fine at first, but lately it's gone south.

AMD Athlon 64 2800+, stock retail heatsink.
MSI K8N Neo, nForce 3 250
1GB RAM
WD 80GB SATA
LG CD-RW/DVD-ROM
GeForce FX 5200

Nothing is overclocked, tweaked, modded, or otherwise altered

It'll boot just fine, and run for 5-10 minutes. Everything
looks fine, but then it just *stops*. Full system shutdown,
like somebody pulled the plug. Except that the power LED on
the front of the case blinks on and off. And it won't
restart until/unless I hit the cutoff on the PSU or unpug
the machine and plug it back in.

My first thought was the power supply, since the case it's
built in had a *very* cheap PSU. So I went and bought the
overkill 480W PSU. No change.

I thought it might be heat, but I checked the HW monitor in the BIOS, and
it was only at around 100F when it shutdown.

If it used to work but now doesn't work for long, I'd look for a short
between the mobo and case that heats up when current flows through it.
I would also measure the voltages with a meter because the stuff built
into the mobo is often very inaccurate, and I'd run www.memtest86.com
or www.goldmemory.cz for at least eight hours straight.

How many watts can that PSU put out? Just because it says 480W doesn't
mean it can put out anything like that.

 

[General Schvantzkoph]

On Sat, 11 Jun 2005 15:31:37 -0700, Brian wrote:

What is the shutdown temperature setting in the BIOS?

 

[Joe]

Did you read his other post? He said:

I took it down to one stick instead of two. I've tried both sticks
individually in slot one, I guess after this I'll try slot two.

It's not a memory issue.

Joe

 

[Brian]

"Bother!", said Pooh, as he read larry moe 'n curly's latest post to
alt.comp.hardware.

If it used to work but now doesn't work for long, I'd look for a short
between the mobo and case that heats up when current flows through it.
I would also measure the voltages with a meter because the stuff built
into the mobo is often very inaccurate, and I'd run www.memtest86.com
or www.goldmemory.cz for at least eight hours straight.

I run Memtest on every computer I build, as a matter of course. It passed
two full rounds shortly after the system was built.

I'd *love* to run Memtest for 8 hours, but considering it can't even run
BIOS setup for 15 minutes right now, that's not likely to happen.

How many watts can that PSU put out? Just because it says 480W doesn't
mean it can put out anything like that.

215W combined on 3.3v and 5v rails. I haven't broken out the multi-meter
yet, but the BIOS PC health meters all look spot on. And the behavior is
consistent across two completely different powersupplies, so unless they're
*both* bad, then it's not the PSU.

Brian

 

[larry moe 'n curly]

"Bother!", said Pooh, as he read larry moe 'n curly's latest post to
alt.comp.hardware.

I'd *love* to run Memtest for 8 hours, but considering it
can't even run BIOS setup for 15 minutes right now, that's
not likely to happen.

215W combined on 3.3v and 5v rails.

Is that what it's rated for, or is that what it can actually put out?

Because some PSUs, like those from Achieve, Q-tec, and Deer, fall far
short of their ratings and may be built with less metal inside than a
good PSU rated for just half the power. Still, I'd be surprised if
your mobo drew anything close to 215W combined power because it runs
the CPU from the +12V rail, meaning it probably draws <60W from the
+3.3V and +5.0V.

I haven't broken out the multi-meter yet, but the BIOS PC health
meters all look spot on. And the behavior is consistent across
two completely different powersupplies, so unless they're *both*
bad, then it's not the PSU.

What brands are they? About five years ago, Anandtech tested several
PSUs and found that two rated for 400W that couldn't run equipment that
was no problem for some 250-300W PSUs. Since then, the manufacturer of
those 400W PSUs has cheapened its products even more.

I'd test the mobo outside of the case, to elimate any doubt of a short
with the case. If that doesn't help, then I'd measure the CPU core and
memory voltage regulators.

Could this be a problem caused by a BIOS setting that turn on the
computer (wake on LAN, serial port, or restored AC power) or turn it
off (doze/sleep/suspend)? I've had reboots and freezes caused by them,
so I disable thm while setting up the hardware and software.

 

[Brian]

"Bother!", said Pooh, as he read larry moe 'n curly's latest post to
alt.comp.hardware.

Is that what it's rated for, or is that what it can actually put out?

Rated. I haven't busted out the multi-meter, yet, because I'm not sure how
I would test it.

Because some PSUs, like those from Achieve, Q-tec, and Deer, fall far
short of their ratings and may be built with less metal inside than a
good PSU rated for just half the power. Still, I'd be surprised if
your mobo drew anything close to 215W combined power because it runs
the CPU from the +12V rail, meaning it probably draws <60W from the
+3.3V and +5.0V.

Yah. I had someone tell me once that a quick-and-dirty way to tell a good
PSU from a cheap one was by weight. Since good transformers and such are
heavy. The one that I started with, no-name that came with the case, felt
like it was made of balsa wood. The new one (Plug-n-Power brand) is large
and heavy. And the BIOS reports the voltages either right at where they
should be, or .1 higher.

I'd test the mobo outside of the case, to elimate any doubt of a short
with the case. If that doesn't help, then I'd measure the CPU core and
memory voltage regulators.

I'll try this next, although I'm usually pretty careful when I'm mounting
the board (and again, if I'd mounted it crooked, why'd it take weeks for
the problem to emerge? This isn't defensiveness, I'm geniunely baffled.)

Could this be a problem caused by a BIOS setting that turn on the
computer (wake on LAN, serial port, or restored AC power) or turn it
off (doze/sleep/suspend)? I've had reboots and freezes caused by them,
so I disable thm while setting up the hardware and software.

I'm pretty sure I've got them all turned off already, because I don't
really like anything like that. But I'll check it again. Along with a
shutdown temp setting.

I also have a new motherboard and cpu on order from Newegg, if all else
fails.

Brian