Computer Hardware Help → Re: Troubleshooting a PSU

Kris,

I used one of these to test my power supply. It showed the 5 volt line was out:

»www.outletpc.com/ep1827. ··· odO1NqMQ

It worked for me, and it's cheap, too.

Freddy

The biggest problem that I have with those PSU testers (like the one Freddy linked -- yes I have one, different brand though) is that they're misleading for troubleshooting complex problems. Let me explain:

These testers will tell you two things, and those alone:

1) If there's no voltage at all (some sort of short, broken pins, something blew up, etc.) to each individual voltage bit (3.3V, 5V, 12V),
2) What your voltages are when plugged into the PSU tester.

#1 is totally legitimate -- no beef there. It's #2 I have a problem with.

Remember: these testers are not in-line devices. Meaning, they do not sit between your PC and your PSU; you plug the PSU into the tester and look at an LCD and go "uh, yeah, okay". That's the analysis? Half-ass.

These devices in no way shape or form stress a PSU the same way a PC does. Consider voltage-wise what all goes on with a GPU hooked up, 4 hard disks, a motherboard, and possibly a 2x4 (or 1x8) ATX12V connector for a single or dual physical CPU system. Do you really think the voltages are going to be the same in that situation as when hooked to a little PSU tester?

If these testers were in-line they would be *super* useful. Imagine if one of your 12V lines dropped suddenly to 10.8V then ramped back up to 12V in the middle of the system being used. The device could maybe emit a tone/buzzer, or (what I'd love) keep some kind of log (maybe even offloaded to a serial or USB port so you could hook it up to a laptop/other system and log what the voltages are in real-time -- wow, imagine that, something we've done for years in the hardware monitoring world... ).

Take for example this exact situation of mine. One of those PSU testers wouldn't have been able to diagnose this; the 12V lines drove 12V no problem, except for very specific situations where they'd drop out, thus the problem I experienced. I was lucky to find what I did visually, otherwise I would have been replacing hardware piece by piece until the issue went away (and sometimes it took days before it recurred).

So when using one of those PSU testers, keep all of what I've said here in mind.

I hear what you're saying. I am hoping I can mimic my friend's problems by letting it run a stress-test for a couple hours and see if it flakes out on me.

I got the PC last night--I didn't have a chance to start it up, but I did poke around inside, and it has:

MSI micro ATX board
AMD 1055t CPU w/a lay-flat CPU heatsink/fan
Four sticks of two types of ram (guessing 8GB)
GeForce 500 series GPU (550 or 560)
2 HDDs
1 Optical drive
Kingwin MK-750 PSU
CM HAF 912 case

He dropped off a Vertex 3 120GB and a Hyper 212+ he said he wanted installed (no idea on the CPU cooler--I'm guessing an attempt at lowering temps) along with some money for misc. extra parts.

My plan of attack, which will probably have to wait until the weekend, is to sweep it out and do all the stability tests before I add anything, then add the SSD and new CPU cooler. His case also has no fans beyond the 240mm one in the front, so I might pick up a couple case fans for him to put on the top and side and create some positive pressure inside the case.

My biggest concern is the PSU. I hadn't even really heard of Kingwin before, and it appears that some of their higher end PSUs get good reviews (the Lazer Golds--ppl seem to like their 1000w and 1200w ones), but the one in question here--an MK-750--isn't even bronze certified and only costs $80 at MicroCenter. In fact, none of the MK series ("Superflower" in OEMs) have any efficiency certifications.

It's hard to find reviews on this model or the MK series (already checked JonnyGuru), so I'm hoping someone here could provide intelligent feedback about whether it's "good enough." If there's a specific test I'd need to run to make that decision, please let me know.

For my friend's build, I'm inclined to think that a 550w PSU is all he needs, but have not yet decided if I need to replace the current one.

You're putting too much thought into the PSU. The problem lies with the CPU overheating most likely.

If you really want to troubleshoot the problem and be certain of the fix, do NOT sweep it out first. The first thing you should do is not change anything and turn it one, run stress test or whatever and see if you can duplicate the problem. If you can, then commence with your cleaning, etc., and rerun your tests. If the problem can no longer be duplicated, then you have resolved it with your cleaning actions. If the problem cannot be duplicated, then you need to look elsewhere including something unique at his house.

So I started pretty much the same way. Booted up, verified temps were in the OK range (CPU was 25 C for a 1055t, 560ti was 29 C), and started a one-hour LINPACK CPU test.

Three minutes after it started, CPU temps were at 59 C when the computer just shut down and wouldn't start back up. 59 C should not cause that type of behavior.

Here is the feedback I am getting from the PSU--this would indicate it is operating correctly, no?



I'm going to give it 5-10 minutes to cool off and see if it will boot back up. If not, I'll start swapping out components, starting with RAM, and see what fried.

Is there a blue screen or just a quick power off?

It just died.

I'm too lazy to scroll up, but have you taken a look at the CPU? Not just the reported temps, but an actual look at the CPU? Repaste and reseat the thing.

That's on the to-do list tonight. I wanted to test it for stability before I cleaned anything, but it's possible the CPU fan and heatsink are full of dust. I'll repaste and install the 212+ and see if it boots.

If it doesn't, I'll swap PSUs. If not, then I may have fried the mobo or CPU.

Has it been able to boot since?

It's dead, Jim.

Won't turn back on. A couple of tries had all the fans (case fans, CPU fan, GPU fans) start to turn a couple millimeters before stopping. Nothing will respond now.

On the to-do list tonight is re-seat the CPU, and pull out the PSU and test in another computer to ensure it still works.

I'm not a fan of the PSU to begin with, but the fact that I could get a reading off it (see pic above) while the computer refuses to boot up lends me to believe it's a CPU/mobo issue. I am really hoping re-seating the CPU takes care of it, but if not, I may need to look into replacing either the CPU, or motherboard, or both (which sucks because I have no other AM3 CPU/mobo to try them in).

Edit: I checked the mobo for burst or bulging capacitors and found none. That said, it's a low-budget MSI AM3 board and doesn't even have a debug screen on it.

Easy things to try before re-seating CPU: Re-seat the PS connector to the MB, make sure the additional +12V plug is seated properly too.

Also try removing all unnecessary items (e.g. disks, I/O cards, extra memory, DVD drives), except minimum memory, and graphics card (if it does not have integrated graphics). Your goal is to reduce the PS draw to only that which is critical and also remove other potential unknowns. Don't forget to pull the data cables too from drives.

The fact that the power supply tester showed it ok doesn't mean that much. Having fans spin but not much else can very well mean the power supply is shot. I would check with another psu before I would bother dismantling the cpu cooler.

"The fact that the power supply tester showed it ok doesn't mean that much."

reason being there isn't a significant load on it.

When trouble shooting anything beyond swap and fit, stress testing is the best way to analyze anything. Computers, cars, you name it. I'm no engineer but know from a vast source of "jack of all trades" knowledge. Yes the tool will tell you voltages, but will it tells amps, watts? I'm no sparkie either so I'm not sure if resistance and all other types of tests can be done?

Still, for the price, I'm thinking of investing in one. Love to be a valid user of a multi-meter but haven't had anyone show me, and electrical currents being what they are, unsure is not a safe practice.

The Killawatt can show voltage, amperage, watts, watt-hours, and at least one more thing. Unfortunately, it's on the outside end of the PSU (between the PSU and the wall), so I have no idea what type of power the PSU is delivering to the computer.

I ended up getting home from work really late last night and didn't get a chance to play around with it last night. I have a new PSU coming on Thursday but I may pull out one of my own to test it on earlier than that.

Thanks for that one, it's a little different from a multi-meter. Also unless your in tech or computer repairs, you wouldn't even need to think of how to set up a test bed for loading parts.

Although an old computer laying around and try the power supply out to see if it is repeated might help; but then not everyone has a half dozen old boxes with no use for them that can afford to risk "live" testing. I always thought they would go to a good home, but with win 7 on the scene, they seem like wasted space sitting in the corner.

So I ended up rebuilding my friend's computer. The mobo was definitely dead. The old PSU would power-on but I wasn't going to hook anything up to it lest it blow up more components. He's supremely happy with it and told me that it was the computer he should have gotten two years ago. At least he's learned his lesson about taking shifty salesmen at their word and double-checking that he gets what he pays for.

As a recap, here's what he was promised:
- Corsair PSU (he got a crappy Kingwin)
- SLI/Crossfire capability (chipset won't support either)
- Crucial ram (mixture of ADATA and something that looked like this)

Also of note is that the original vendor actually took out one of the two 12cm case fans at the front of the case.

Old Specs:
- MSI 785GM-P45 micro ATX mobo
- Kingwin 750w unrated PSU
- Mixture of ADATA and bargain Crucial memory (no heat spreaders)
- 2x 500 GB HDDs in Raid 1

Here are the computer's specs now:

- Cooler Master HAF912 Case (reused)
- Optical drive (reused)
- Galaxy GTX 560ti (reused)
- AMD Phenom II X6 1055t (reused)
- HDDs: 1x 500 GB (second one from RAID 1 array was toast), 1x 1 TB (external, whose case had died but the HDD was still good)
- GIGABYTE GA-990XA-UD3 AM3+ AMD 990X SATA 6Gb/s USB 3.0 ATX AMD Motherboard
- CORSAIR XMS3 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory
- OCZ Vertex 3 VTX3-25SAT3-120G 2.5" 120GB SATA III MLC Internal Solid State Drive (SSD)
- Western Digital Caviar Blue WD10EALX 1TB 7200 RPM 32MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive
- Rosewill HIVE Series HIVE-550 550W Continuous @40°C, 80 PLUS BRONZE Certified, Modular Design, Single +12V Rail, ATX12V v2.31/EPS12V v2.92, SLI Ready, CrossFire Ready, Active-PFC Power Supply

I also beefed up the HAF912's case fans from two 12cm to two 20cm, one 12cm, and one 14cm.

Happily, his OEM copy of Win7 Home reinstalled onto the SSD despite nearly everything in the computer (except the CPU and GPU, really) changing, without it requiring a phone call to MS to unlock.