Help me diagnose this shit

[GoGeTa006]

Im helping my roomate building his first new rig, he did most of it himself but now it just randomly restarts, sometimes before even loading the startup screen, sometimes 5 minutes in, sometimes 1 minute into system startup

System:
- MSI motherboard
- Radeon HD 6870
- 2x 4GB DDR3 1600
- AMD Quad

anyways, I have tried the following and I am like 90% sure its the motherboard:

- Using only 1 RAM stick, trying in the different slots, still had the same issue
- Starting in safe mode and updating all the mobo drivers and graphics drivers to latest version
- Reinstalling windows
- Installing my video card (GeForce 9800 GTX+ with newest drivers)


So far, all of them have yielded the same results, I am assuming it has to do something with the motherboard or something because if I start in safe mode it doesnt crash (it doesnt load all of the drivers), I thought it could be the video card but after trying with a different video card thats known to work (I  use it) it still had that issue.

Anyone know any other "tests" I could perform to be 100% sure its the motherboard?

what I will try next is connecting the video card to the 2nd PCIe port, but that'll wait for tomorrow since I have to take out the PSU (I hate bottom-mounted PSU's)

anyways, any suggestions for diagnosing are very much welcome

arigato

[kureshii]

Run memtest86 on each memory stick, in each RAM slot. Should help you eliminate RAM or mobo as the culprit (if they are in fact responsible).

[kitamesume]

it sounds like the PSU is getting overloaded and shutting down, try swapping the PSU? you didnt mention the PSU as well.

edit: try this first -

run the whole PC as only

• 1x hdd
• 1x ram
• mobo
• cpu
• psu

and nothing else, if it runs pretty stable without the gpu and others, it may be the GPU(unlikely though), the motherboard, or the PSU cant handle the load.

[rostheferret]

Like was said, systematically check everything else first. Try running without the gfx card at all, then you at least know it has something to do with the card. If it runs fine in safe mode, it won't be loading those drivers either. It's possible the PSU is being overloaded - in fact, it sounds likely - but it's also possible there was a compatibility issue between the gfx card and the mobo. Usually they pick up on these issues pretty quick so make sure the BIOS is updated. I had an issue where the mobo had a "power saver" feature, which cut power to anything it didn't think was being used. Ergo: it kept cutting the power to my gfx card if I left it for a few minutes causing it to freeze. Slightly different result, but eh, either way it's probably worth updating the BIOS anyway. RAM errors usually causes a stack overflow and blue screen to my knowledge, but it's best to check every possibility.

[kitamesume]

^ this reminds me of the old issues of motherboards hating certain CPUs, specially when you put an AM3+ on an AM3 board, not every AM3 board supports an AM3+ CPU.

[vuzedome]

Well do everything first, then try swapping the PSU.
For all we know your friend might be powering the rig with a generic 450W PSU without the usual logos, like 80+ and stuff.

[GoGeTa006]

well according to newegg psu calculator he should need 485W and he has a 530W PSU, thats why I didnt really think it would be the PSU. . .although I do hear a high-pitch sound comming from it every now and then. . .

I'll try memtest thank you kureshii,

and it does run without the video card, I mean I cant do anything but it doesnt reset

I think I found the culprit:


I just reformatted the hard drive and reinstalled windows, difference between last time:


First time:
- Install windows
- Install video drivers from web (latest)
*crash*
*safe mode*
- Install MSI mobo drivers from web (latest)
*crash*


This time:
- Install windows
- Install MSI mobo drivers from CD (not latest)
- Install Video drivers from CD
*crash*


so I am assuming it has to do with the video drivers, now does this mean its the video card? or a software (drivers) problem?
as I mentioned earlier the nVidia card also showed errors. . .

[Ixarku]

If it were me, the next thing I'd thing I'd do would be to swap out the video card with another card, reinstall the video drivers, and see what happens.

[mgz]

also do a chkdsk on the HDD make sure no bad sectors etc. Eliminate as much as possible. Also try not installing vid card drivers

[GoGeTa006]

Well I tried booting into MSI's winky, which AFAIK is a linux based OS and it kept crashing (with bith his and my video card)

I am currently running his video card in my rig and havent had any issues. . .so things keep pointing at it being the Mobo. . .

Already tried to fix the hard drive. . .even wrote zeros to it just in case. . .

Memtest didnt report anything unusual. . .

Ffs I even tried installing windows basic. . .still crashes. . .

[kitamesume]

still, try swapping out the PSU, even if the PSU is plenty, if the PSU is faulty then thats a problem itself.

i had experiences with this, PC kept crashing even with a 550watt PSU, the unit is barely an i3 without a GPU which should indicate that the PSU is overkill. swapped the PSU and it stopped crashing, apparently the defective PSU had a faulty OCP and trips when theres a sharp rise in current consumption even far before it hits the target limit.

although I do hear a high-pitch sound comming from it every now and then. . .

this might be an indication of it.

[vuzedome]

Probably as soon as the new graphic drivers kick in the GPU starts sucking power out of the PSU and then BAM!!
No facts to back up my claim, just guessing.

[kureshii]

With all the above info, it's probably down to mobo and PSU now. Swap out the PSU like kita said, and see if it still happens.

[rostheferret]

Someone please point out if I'm talking out my ass, but to the best of my knowledge most PSU's (particularly cheap ones) run on rails. For example a 400W PSU may have two rails capable of handling 200W a piece. If you overload one rail, down goes the system. You'll have multiple leads coming out of the PSU - and since you're running close to the limit it seems - don't plug in all from one lead. A power spike could be all that's needed to cause the crash. Normally you'll want to allow a 20% buffer for optimum efficiency. Means he should be looking at a 600W really, but you're right in that it should still run in theory. Failing that, see if you can't get a PSU from another computer for testing purposes.

[kitamesume]

Someone please point out if I'm talking out my ass, but to the best of my knowledge most PSU's (particularly cheap ones) run on rails. For example a 400W PSU may have two rails capable of handling 200W a piece. If you overload one rail, down goes the system. You'll have multiple leads coming out of the PSU - and since you're running close to the limit it seems - don't plug in all from one lead. A power spike could be all that's needed to cause the crash. Normally you'll want to allow a 20% buffer for optimum efficiency. Means he should be looking at a 600W really, but you're right in that it should still run in theory. Failing that, see if you can't get a PSU from another computer for testing purposes.

theres a flaw in this concept, you shouldn't aim for pure wattage rating, hunt for ampere capacity per rails, aim for 18A(for single-mainstream GPUs) to over 22A(required for high-end GPUs).
i've aimed for 350watt PSUs (seasonic SS-350SFE) with some awesome 12V amperage, at 18A on each of it's two rails(total Amperage mustn't exceed 26A) it can run a 180watt GPU(15Amps) for all it cares without breaking a sweat.

edit: now that i think of it, i wonder which rail is the P4 power is tapped into, or does it have it's own rail... hmmm...

ah and about the power spikes, it rarely happens on a PC, since a PC's power consumption is capped.

[Tiphares2000]

It's all guesswork untill we have more information.
Still needed is : maker & model of at least the mainboard, CPU, RAM, PSU, windows version

Right now we are uncertain if the memory is in the QVL of the mainboard, compatibility of the maiboard + CPU (as Kitamesume pointed out earlier), output of the PSU (Kitamesume again) over the 12V lines.

Failures might come from:
memory : incompatibility, wrong settings, bad ramsticks <- excluded from Memtest
PSU : configuration draws 100-150W for CPU, 150W for GPU = 300W + change for other components = around 25A from the 12V lines. Quality PSU gives around 40A combined. (edit : thats at a 530W PSU)

It can still be a driver problem, as you are stable in windows safe mode. Be sure to remove the drivers completely when changeing from ATI to Nvidia (or back).

You can try running a linux distro to see what happens.

[rostheferret]

Someone please point out if I'm talking out my ass, but to the best of my knowledge most PSU's (particularly cheap ones) run on rails. For example a 400W PSU may have two rails capable of handling 200W a piece. If you overload one rail, down goes the system. You'll have multiple leads coming out of the PSU - and since you're running close to the limit it seems - don't plug in all from one lead. A power spike could be all that's needed to cause the crash. Normally you'll want to allow a 20% buffer for optimum efficiency. Means he should be looking at a 600W really, but you're right in that it should still run in theory. Failing that, see if you can't get a PSU from another computer for testing purposes.

theres a flaw in this concept, you shouldn't aim for pure wattage rating, hunt for ampere capacity per rails, aim for 18A(for single-mainstream GPUs) to over 22A(required for high-end GPUs).
i've aimed for 350watt PSUs (seasonic SS-350SFE) with some awesome 12V amperage, at 18A on each of it's two rails(total Amperage mustn't exceed 26A) it can run a 180watt GPU(15Amps) for all it cares without breaking a sweat.

edit: now that i think of it, i wonder which rail is the P4 power is tapped into, or does it have it's own rail... hmmm...

ah and about the power spikes, it rarely happens on a PC, since a PC's power consumption is capped.

Blergh. Amps vs Power: School-level physics. Well now I feel like a dolt. Still, it was a theory, even if my explanation had more holes than swiss cheese.

[Pentium100]

A cheap PSU won't provide the power it claims to.

I recently repaired a cheap PSU and it claimed to provide 30A at 5V, but the diode used was rated for 2x10A, so it could only provide 20A at 5V at most.

Good quality PSUs will provide the rated power and might even survive a slight overload. With a cheap one be happy if the PSU does not blow up when you try to pull the rated power.

Also, if the PSU has more than one "rail" be careful to load balance between them - you might overload one of them and the power supply with shut off or release the magic smoke.

[kitamesume]

A cheap PSU won't provide the power it claims to.

I recently repaired a cheap PSU and it claimed to provide 30A at 5V, but the diode used was rated for 2x10A, so it could only provide 20A at 5V at most.

Good quality PSUs will provide the rated power and might even survive a slight overload. With a cheap one be happy if the PSU does not blow up when you try to pull the rated power.

Also, if the PSU has more than one "rail" be careful to load balance between them - you might overload one of them and the power supply with shut off or release the magic smoke.

is that 2x10A continuous or pulsed? because pulse rating is approximately 3-4times higher than continuous, and it should be obvious that a switching powersupply is pulsed with a certain duty cycle.

if it was a quality PSU, even if you overload one rail it shouldn't send off a magic smoke, the OCP sensor could get fried when repeatedly done though.

[GoGeTa006]

YUP. . . Swaped the PSU and its working fine now. . .

Thank you kind sirs