System Powers Off - Revisited

Blue|Fusion · Posted: Thu, 14 Dec 2006 18:06:33 Post Subject: System Powers Off - Revisited

A few of you has helped me diagnose some issues with my sister's old Dell in the other trhead titled "System Powers Off." Well this time, my system seems to be doing it on occasion.

It first did it last week in my college dorm room. The system was updating, so there was alot of C compilation going on in the background. In the foreground, I was just surfing in Firefox. Out of nowhere, the system powered off. No warnings, beeps, or anything. Just went dead. I sat there stunned for a second. I then pressed the power button, however, nothing happened. I checked the plugs and all was good. I then decided to just glance in the side window. I saw that the green LED on the motherboard was still on, so there was power getting to it, so it wasn't the PSU, atleast I don't think. I tried the power button again to no success a few minutes later. Afte that, I cycled the power switch on the back of the PSU, waited a few seconds, and then tried the power button again. Booted up normally.

When it did go down, temps weren't any higher than usual for load on the P-D 840, plus, I figure it would throttle before it would actually kill the system.

Well, everything was going fine thereafter until today. I am recompiling all software on my system for some new CFLAGS to optimize the programs for the CPU. I was in about 100 of 430 packages after about two and a half hours of running at load temperatures. It was running while I was downstairs for most of the time. Just before it decided to die again, I was back in Firefox browsing while it was compiling in the background. I'm still not sure if Firefox has anything to do with it at all, but the only similarity between both accounts were that I was compiling and surfing in FF.

So here's what I did this time after the fact...
I had the system overclocked from 200MHz FSB to 214MHz FSB. All of the voltages were as low as they go in the BIOS with the exception of the PCI-E and the RAM which were at normal voltages. I ran memtest86+ several times through to ensure the overclocked RAM was working properly, and then I ran Prime95 stress test for several hours to make sure everything else was working well with the light overclock and undervoltage. No issues whatsoever.

I reduced the overclock to 210MHz on the FSB for now, but left the RAM timings as low as I can get them.

Anyway, does anyone have a good way to diagnose this and see if it's a problem with the PSU or overheating or what?

Thanks.

LaTech · Posted: Thu, 14 Dec 2006 19:41:45 Post Subject:

which model dell do you have? if it's a Optiplex 240 or 270, check the capacitors on the motherboard.

Blue|Fusion · Posted: Thu, 14 Dec 2006 23:26:34 Post Subject:

The old Dell isn't the issue - that was the other topic I started last week. This time it has to do with my custom rig. Here's the specs for reference: http://richgannon.net/?style=v1&page=content&id=5

Blue|Fusion · Posted: Fri, 15 Dec 2006 03:30:38 Post Subject:

Ugh. It happened again. This time while I was away so I don't know if the temps were high or not. But based on what I've seen, I'm thinking it is a problem witht the PSU going out when there's heavy CPU load for extended periods of time.

Any practacle way to test the PSU/mobo to see if I can confirm where the problem lies?

By the way, I completely removed the overclock this time, so we'll see if that makes any difference.

Kilamon · Rated XXX Joined: 22 Mar 2005 Posts: 811

Do you have a voltage monitor that you can use? Are you running these systems on a UPS?

Blue|Fusion · Posted: Fri, 15 Dec 2006 20:27:23 Post Subject:

The system and all components are plugged into a Belkin UPS. As a matter of fact, the power went out briefly while I was away today and I came home to a still running computer, so the UPS is running fine.

lm_sensors just got an updated driver for the Winbond chip used on the ASUS motherboard, but I believe there is a little fishyness with it as it displays the 12V rails with 6 volts. I really don't believe that's right, so I think there's an error in the math on that driver. But for the sake of things, here's the rest of the output from the voltage readings:

I'm going to try to get some readings from Windows in a bit. Also, FYI, the BIOS shows the 12V rails as at around 12V so I am pretty certain that's just bad math in the driver.

edvallie · Posted: Sat, 16 Dec 2006 03:44:49 Post Subject:

My first instinct with a problem like this would be to RUN STOCK SPEEDS. Overclocking+Low Voltages are just begging for some type of power failure.

**Modulok** · Posted: Sat, 16 Dec 2006 11:30:21 Post Subject:

blue fusion, what distro are you running? I would like to know how to get the hardware stats like you have it...looks nice.

**Doctor Feelgood** · Posted: Sat, 16 Dec 2006 11:40:15 Post Subject:

I think he is using the same thing you liked back here:

http://www.bigbruin.com/forum/viewtopic.php?t=10422

Grin

Blue|Fusion · Posted: Sat, 16 Dec 2006 12:31:36 Post Subject:

It is GKrellM with some 3rd party theme using lm_sensors, hddtemp, and the nvidia-settings temperature monitors.

As far as the power failures go, it does look like the overclocking was the issue with the RAM. With it backed down to stock FSB, there have been no problems overnight rebuilding everything for a good stress test. Right now on package 464 of 466 so looking good. I just have the memory timings a bit lower and I think I'll stick with just that and leave the speeds as they are.

Thanks for the help, guys.