Please confirm my reasonings for Failing PSU

August 22, 2010 at 10:50:57
Specs: Windows 7 64, 2.128 GHz / 2047 MB
Hi everyone, I'm looking for some expert confirmation of my own (perhaps flimsy) investigative conclusion.

My system has been hanging while running both gaming and video applications. Games hang reliably at random intervals between 2 minutes and 1 hour. Video (VLC) doesn't always hang, and when it does it's completely random. I accuse the PSU, in the computer case, with the 3.3v rail! Let's see the cards...

XFX GeForce 9300 Intel 775 mobo
Intel C2D e6400 @stock 2.13ghz
2x 2gb Patriot DDR2-800
30GB OCZ Vertex SATA II ssd (OS drive)
2 x Samsung Spinpoint hdd's
Antec Neo HE430 psu

HWMonitor reports the 3.3v rail at a steady 2.18v, the -12v rail at steady -13.18v and the -5v rail at steady -8.70v, all other values fairly nominal. I don't have the know-how to confirm this with a voltmeter, and I know the onboard voltage monitor can be wrong, but this psu has been through fire smoke (6 months ago) and I've at least partly excluded a few other factors (listed below). I'm not confident about the motherboard (bought to replace my beloved, and smoke damaged, GA-E7AUM-DS2H) either, but I hope I'm just biased. Every test I try on the system, CPU GPU and memory, eventually hangs, but always before reporting any errors.

Excluded factors -

Not the ssd - installed same OS and apps onto a small partition on the newer hdd and the problem remains exactly the same.

Not strictly software - hangs occur even without background apps (avast, utorrent, and whatever system processes I can close) but closing those apps does seem to buy time.

Not a cooling issue - CPU tops out at 68C running OCCT cpu test (eventually hangs anyway), GPU hangs around 84C under OCCT but doesn't surpass mid 70's in gaming/video.

Probably not memory - Memtest86 + v4.10 ran 4 passes with no errors. The system hung 98% of the way through the fifth pass, still no errors, after 5 hours 21 minutes. I also moved both ram sticks into the next slots and keep experiencing the same issue.

I know this doesn't represent much of an investigation, I'm just not that experienced with serious trouble-shooting. Would replacing the PSU be a reasonable move, or are there more tests I should run first for more certainty?

Thanks a lot!

See More: Please confirm my reasonings for Failing PSU

Report •

August 22, 2010 at 11:59:02
The motherboard gets mediocre reviews & although the onboard GF 9300 video is one of the best IGPs available, it definitely isn't great for modern gaming. Your "beloved" board wasn't much better. If you want to play games & play them well, you need a discrete video card.

As for the main issue, since the system is locking up when running memtest86 (which I assume is from a bootable CD?), that pretty much rules out a software problem. It would seem the problem is hardware related. Start by checking your temps & voltages in the BIOS. The -5v & -12v aren't all that important, but the "big 3" (+3.3v, +5v, +12v) should be within 5% tolerance. Obviously 2.18v for the +3.3v line would appear to be a problem, but you need to somehow confirm whether the reading is accurate or just a software glitch.

Also, check your BIOS settings, especially for the RAM. Do NOT simply except the Defaults or AUTO settings. Manually input the speed, timings & voltage as per the Patriot specs.

"Not a cooling issue - CPU tops out at 68C"

You don't see that as a problem? The CPU temp should not go above 60C. Make sure you installed the heatsink correctly. Just a small thin line of paste running down the middle of the CPU is all you need:

Report •

August 22, 2010 at 12:12:56
+12v: +11.40v-12.60v
+5v: +4.75v-5.25v
+3.3v: +3.14v-3.47v
bios should provide this info. Check it there first.

Report •

August 24, 2010 at 13:31:52
Hey, thanks for the advice guys. Ha! I made more mistakes than I thought...

Regarding the 3.3v rail - BIOS reports 3.296v, while the software monitors I check (Speedfan, HWMonitor, OCCT) all report 2.18v. There's also a discrepancy in CPU temp readings, BIOS reporting 5-10C less than the software monitors. On top of that the system won't post if I put in the correct RAM timings (4-4-4-12 @ 2.1v for Patriot Viper DDR2-800) and there's no option to change RAM voltages. (This pos board has a jumper that lets you switch the BIOS to 'overclock mode,' I'm using that already.) I remember this happened to me when I first installed the board, but it was so soon after the fire there were more important things to take care of so I set it back to Auto and forgot about it. I know I had the correct timings and voltages on the gigabyte board... Might a different BIOS solve anything? (No experience at all with bios updates, so a link to good resources would be necessary if someone were to suggest changing the BIOS please.)

Regarding the temperature issues - yeah 68C is too hot, I wasn't thinking. Turns out I had the PWM set incorrectly and the fan was ramping up less than 30 rpm between idle and full load haha. That's fixed now. The CPU doesn't go above 53C (highest software report) and the GPU also runs almost 10C cooler (I bumped up the case fan as well).

All's not well though. My idle temps were too high at 37C CoreTemp reported (I have an AC Freezer 7 Pro, which reported idling an X6800 at 26C using Coretemp). So I took the computer apart to remove the cooler (have to pinch the pins from behind the board...), cleaned it and the cpu, reapplied Arctic Silver, reseated the cooler and now here I am, idle temps same as before.

Rereading your post, jam, I saw you mentioned "Just a small thin line" would do. I used at least twice that much, just like I did the time before. Am I putting too much on?

Will try to crash the computer again tonight with the better thermals, but I doubt that's the only issue... Thanks again for your help!


Report •

Related Solutions

August 29, 2010 at 09:07:24
I've definitely ruled out temps this time, so it's a component issue, but on an old chipset with old components this computer is barely worth replacing one part, much less with the risk it's not the right one. I'll just deal with it until Sandybridge launches, then I'll be sitting pretty.

Thanks for the help.

Report •

Ask Question