unexpected rebootski

[VirtualLarry]

I have a rig here that I just recently built, and it has a strange problem. It reboots, randomly, after a day or so, just being left alone to its own devices.

It's 24hrs Prime95 stable, Memtest86+ 1.65 stable, and otherwise stable for everything that I do. It just mysteriously reboots.

Temps under load are 70-71C core temp, mobo sensor is a bit whacked, it also shows 70C case temp, but it didn't used to say that, it used to be 10C below core temps when I built the rig.

In BIOS, the +12V is showing up a little bit weak, but looping Prime95 + 3DMark01 doesn't crash the system, so I doubt that the problem is the 12V load.

Components:
GA-P35-DS3R
E2140 @ 3.28Ghz, 1.41875v (BIOS)
Patriot 2x2GB DDR2-800 @ 410Mhz, 5-5-5-15, 1.8v
WD 3200JD 320GB HD
Sapphire X1950GT PCI-E 256MB video
CoolerMaster HyperTX2 HSF

Any ideas on where to proceed?

I have both this rig, and another AMD64 3800+ rig hooked up to a BE550R UPS. The UPS doesn't beep, nor does the other rig reboot, when this rig reboots.

Edit: PSU is a brand new ThermalTake 430W.

 

[GundamF91]

I assume when you have APC PowerChute installed, it'll tell you if there's a blackout event, and if there's no log for it, then the power can be ruled out.

As for the system, 70C is pretty hot for E2140. Your RAM should be fine, but I would not be surprised if the Northbridge was running too hot and caused it to restart. 410mhz quad pumped is pretty hard on the motherboard, and if it has weak circuitry then that could cause it. Does the reboot happen even if you leave the computer on and do nothing, or does it reboot only when you run stress test overnight?

 

[Denithor]

My first guess would be a flaky power supply. Do you have a backup (known good) that you could use? Otherwise switch PSUs between your two boxes and run for a couple of days and see if the problem continues and if it does, which box goes down.

 

[Jophe]

If you don't have a spare psu, you can still eliminate OS versus power. At this point you can't be sure whether it's a scheduled task in the OS or a real hardware issue. Boot with a boot CD/disk and let the system sit for a few days. No reboots mean no worries with your psu or ups.

 

[VirtualLarry]

I don't have PowerChute installed, perhaps I should. My 3800+ system is a microatx, so I can't swap PSUs, but I have a Xion RealPower 600W PSU sitting in a box here, when I get a chance I might swap it in and see what happens. I added details about which PSU I have installed - are there known problems with ThermalTake PSUs? My friend who has about 4 of them has had 2 of them go bad on him so far.

If temp1 on SpeedFan is the northbridge temp, it's around 43C, so nothing too hot there. I'm wondering if I should boost the NB voltage one step, to help with the 410FSB. I'm hesitant to do that though, burned out the northbridge on a 650i mobo by doing so.

I'm going to bust out my other BE550R, and put this rig on it's own UPS soon too. I don't know if it's overloading the UPS by having two rigs on it. I wonder if PowerChute can tell me the wattage draw of my rigs. My BX1500 UPS can do that.

Oh yeah, the system has rebooted yet again, since I posted this thread.

 

[GuitarDaddy]

I had a similar problem with a build I did for my wife, it turned out to be a device driver problem. Your first step should be to go into bios and turn off the option (reboot on fatal error) or something similar. Then instead of rebooting it will stop on the blue screen error which should give you a starting place where to look. You should also look at the log files in the windows event viewer and find the dump file created from the crash. Then you can download microsofts dump file analyser and run it on the dump file created. These are the steps I went through to troubleshoot my wifes machine and finally determined that it was a bad driver file for her wireless card that was causing all the trouble.

 

[VirtualLarry]

I have the reboot on bluescreen disabled. It's not bluescreening.

 

[GuitarDaddy]

Well in that case, I too would suspect the power supply

 

[VirtualLarry]

I'm going to try disabling the Seventeen Or Bust service, to take the load off of the system, and run it just in desktop idle mode and see if it still reboots.

Edit: So far, no reboot, hmm.

Edit: Ok, so now I decided to test it with Prime95, and it hasn't rebooted. Unfortunately, Prime95 gave an error after 3 hours. I thought that this rig was prime stable, guess not.

Edit: Ok, bumped up the vcore to 1.425v, and ran prime95 again, and now it rebooted.

Edit: Upped GMCH voltage +0.1v, running SoB, still hasn't rebooted.

Edit: That may have done it. No reboots overnight, still testing.

 

[VirtualLarry]

bump for success! (time to tempt fate)

Edit: Fate tempted. It rebooted overnight last night. Sigh.

Edit: And again, a day later.

Edit: And again, an hour later.

 

[JustaGeek]

Don't remember where, but I've read stories about the Thermaltake 430W "burning up" the motherboards.

Is this the one..? http://www.newegg.com/Product/...Tpk=ThermalTake%2b430W

And it has only 18A on the 12V rail - I would replace it, Larry.

The "Value" series can only be so good...

 

[VirtualLarry]

yes, that's the one

Would this one be a good PSU? link

 

[JustaGeek]

Well, it's only 380W, but Earthwatts is a pretty good line of PSU's.

But I believe that when you put a powerful Video Card in your rig, you need a PSU around 500W. Others might disagree, but if the PSU is your problem here, look how much time and aggrevation it could have saved you!

Perhaps 380W is enough, but it will work much harder and closer to its capacity, being obviously more prone to premature failure.

But it is definitely better than that Thermaltake.

 

[Amaroque]

I concour with the posts here. Spontainous reboots are usually a PSU issue.

Question, does it reboot at stock speeds?

 

[Markfw]

One more thing to try. 1.8 vdimm@stock but running 410 (slight OC) I would crank the vdimm up at least +.1 if not .2

 

[VirtualLarry]

Good point about the RAM voltage. The chips that I bought, someone else in a different forum tested them out and got them to 500Mhz at 1.8v, which is pretty impressive for DDR2-800, that's why I bought them. I Memtest86+'ed them overnight at 1.8v, and there were no errors, but I guess I could try cranking up the voltage a tad.

Originally posted by: Amaroque
I concour with the posts here. Spontainous reboots are usually a PSU issue.
Question, does it reboot at stock speeds?

It doesn't reboot if I just leave it at the desktop idling, which I found to be interesting. If it were just a flaky PSU, you would expect that it would reboot just the same. But yet, I'm not sure if it's the load either, because looping 3dMark01 + Prime95 doesn't cause it to reboot, at least not more than just leaving SeventeenOrBust running in the background on the desktop.
(I figured that the extra GPU load would be significant to the PSU, I have a X1950GT in here.)

I guess I should try running it at stock speeds too. I wonder if my FSB setting (410Mhz) is just too close to the edge for this chip.

I should mention that I have two indentical rigs here (except one has an X1950Pro, and one an X1950GT), and the other rig had some reboot problems too, but those seemed to clear up once I upped the vcore a notch on the CPU, although I haven't given that rig as much of an extended burn-in as this one.

 

[VirtualLarry]

Solved!

It turned out to be my UPS. I had the computer, along with another computer, and my DSL modem plugged into it. The last time that the computer rebooted, I also lost my internet connection, and as it turned out, it fried my DSL modem good. All the while, the UPS never beeped or indicated any overload condition, nor was it on battery power. So I don't know what exactly is wrong with it, other than I'm replacing it.

It seems that my mobo is perfectly happy running at 400FSB, even without the GMCH voltage bump, so I took it back down.

I also clocked down from 3.28 to 3.2 Ghz, as Prime95 failed, although I swear I was able to pass 24hr of Prime95 when I set up the system. I had always intended to clock it down a tiny bit, to leave some headroom.