Failing PSU maybe?

[canoli]

A weird symptom on this system ever since I put it together in Nov. 2010 - just wondering if you guys have any ideas what could be the source(s):

On every cold boot the board powers up and then shuts down after 2 or 3 seconds ... Strike 1... Repeat (no user action, I'm just sitting there waiting) ... Strike 2...
On the 3rd attempt it boots normally, just fine.

Extreme PSU calculator always suggests about 736W. The HX850 should be plenty. It's one of Corsair's better PSUs - Gold certified, 7 year warranty.

I don't know if this is enough info to take a guess on what's going on...but if you have any thoughts I'd appreciate reading them.

Thanks!

[edit: btw this MSI board - like many others - has a few voltage check headers and whenever I connect a meter all the voltages are on spec]

Never mind. From what I can gather a PSU won't fail slowly it will either work or it won't. Sound right? Maybe it's my mobo....something weird is happening though. Last night I came home and the system was dead. DOA. I pulled the PSU, checked it out (visually, not with a meter) and all looked normal - no bulging capacitors or funkiness inside. Put it back together and the system fired up again. Don't know what the ^(*& is going on...

[chuckula]

Never mind. From what I can gather a PSU won't fail slowly it will either work or it won't. Sound right? Maybe it's my mobo....something weird is happening though. Last night I came home and the system was dead. DOA. I pulled the PSU, checked it out (visually, not with a meter) and all looked normal - no bulging capacitors or funkiness inside. Put it back together and the system fired up again. Don't know what the ^(*& is going on...

Hrmm.. on a semi-unrelated note I just built a server box from work, moved it into the server closet after building it and testing it in my office, and the thing started spontaneously rebooting every 2 seconds when I turned it on even though it had been running fine 5 minutes before... turns out the UPS that I plugged it into is either bad or couldn't support the load (I'm not sure why since the server isn't that beefy and the UPS is not heavily loaded). Anyway, if you are looking into bizarre power-related issues, then the UPS could be part of the problem too. Another possibility is that you inadvertently fixed some issue with your PSU when you pulled it out and inspected it.

[absurdity]

I wouldn't rule out the PSU until you've swapped it with a known working unit.

Also worth noting that DOA = Dead on Arrival. Clearly not the case if you've been running the system for two years

[canoli]

Also worth noting that DOA = Dead on Arrival. Clearly not the case if you've been running the system for two years

lol yeah I know, just going for the effect there...maybe DOmA - dead on my arrival - says it better, coming home to a dead system.

My UPS is getting old but we lost power during Sandy and it kicked in just fine, gave me time to shut down properly.

Maybe I did inadvertently "fix" it like chuck said when I pulled the PSU. All I did was blow the dust off the fan but it wasn't exactly caked on or anything.

All I know is it's not normal for a system to fire up and power back down on its own - twice - before booting. Maybe I'm not factoring in my hardware correctly but I doubt it. I only have 1 GPU. RAM doesn't take much voltage to run, even if I do have all 6 DIMMs populated. All my SATA ports are full too but still, 850W ought to be plenty.

Thanks for your replies!

[canoli]

Does anyone know whether testing a PSU with a multimeter is worthwhile? Meaning, checking what voltage each pin is receiving.

I found a site that describes the process but is it worth it? Obviously it won't tell me anything about how the unit performs under load.

Thanks!

[Chrispy_]

I would say no, it's pointless unless you are trying to identify DOA power supplies before plugging them in.

I say this because I have seen PSU's that pass according to those PSU testers at minimum loads, but the voltages quickly fall out of spec once the unit is loaded beyond what the little testers can simulate.

[canoli]

gotcha - although I'm not talking about a "PSU Tester" gadget - I mean a multimeter. There's no simulation going on it's simply reading the voltage at the pin. You basically short-circuit the unit, plug it in and see what each cable reads.

In any case I think you're right, continuity and/or no-load tests should identify a faulty cable, which can be helpful if the unit is dead. But for my situation ... yeah, no point to it.

Thanks for your reply C - I appreciate it.

[ludi]

gotcha - although I'm not talking about a "PSU Tester" gadget - I mean a multimeter. There's no simulation going on it's simply reading the voltage at the pin. You basically short-circuit the unit, plug it in and see what each cable reads.

Clarification: there is no short-circuit involved with either a voltage test or a continuity test.

As for the original question, switch-mode power supplies can have a variety of failure modes that are not so obvious as "works", "doesn't", and "smells funny". The only way to know for sure involves full-power dummy loads and oscilloscopes, but the shotgun approach works well enough if you've got available spare parts to swap around.

[canoli]

okay sounds good. No oscilliscope available so I'm stuck with a multimeter. But the process involves jumping across pins 12 n 13? (i think it's 12/13, the Power On and a Grnd). So that's not really shorting it I guess - or maybe it is...I'm a little weak on the terminology. But you sound like you know what you're talking about, can I ask your opinion - is it worth pulling the unit and doing voltage checks? I'm still getting the symptoms - cold boot stutters with a few tries...then finally completes. Thanks!

[ludi]

okay sounds good. No oscilliscope available so I'm stuck with a multimeter. But the process involves jumping across pins 12 n 13? (i think it's 12/13, the Power On and a Grnd). So that's not really shorting it I guess - or maybe it is...I'm a little weak on the terminology. But you sound like you know what you're talking about, can I ask your opinion - is it worth pulling the unit and doing voltage checks? I'm still getting the symptoms - cold boot stutters with a few tries...then finally completes. Thanks!

Oh, I see where this is going...yes, in practical terms you're shorting the gray wire on pin 14, "PWR OK", to any ground in order to force the PSU to boot. I thought you were referring to the function of the meter or the actual load test. The PSU should self-protect for any short-circuits on any of the +V wiring, but sometimes Bad Things happen if that is done.

[canoli]

Okay so we're speaking the same language ... So, is it worth checking voltages on each cable, esp the 24-pin, seeing what it reads after shorting (forcing it to power up) the unit? I guess if anything checks out way out of tolerance it could be a worthwhile test...yes?

[sluggo]

Okay so we're speaking the same language ... So, is it worth checking voltages on each cable, esp the 24-pin, seeing what it reads after shorting (forcing it to power up) the unit? I guess if anything checks out way out of tolerance it could be a worthwhile test...yes?

If you can get the machine to boot, you can test the voltages under load. Put your voltmeter's ground probe into any cable header with a black wire coming out of it, then probe the 24-pin connector with the voltmeter's V+ lead. Inserting the probes is easy and safe, and they will remain in place if fully inserted so you can watch a voltage as you load down the machine with the benchmark of your choice. CPU and memory voltages are harder to check without the proper gear, but it can be done.

This sounds like a supply that's overly sensitive to an overcurrent or undervoltage condition. If the machine boots only after the third try, the first two tries may be charging up local storage, and then on the third attempt the current out of the supply will be much lower, sneaking under the (too low) threshold of the control circuitry. Try booting the machine immediately after a shutdown - if it boots right away, the current sense may be the problem. Run the test exactly the same way a few more times to confirm.

[canoli]

If you can get the machine to boot, you can test the voltages under load. Put your voltmeter's ground probe into any cable header with a black wire coming out of it, then probe the 24-pin connector with the voltmeter's V+ lead.

The machine boots fine - after 3 tries. But once the machine is running the 24-pin cable is connected to the mobo header. How can I probe it while it's connected? I'm sure I misunderstand you but maybe you can explain it again if you don't mind.

Try booting the machine immediately after a shutdown - if it boots right away, the current sense may be the problem. Run the test exactly the same way a few more times to confirm.

It never "boots right away" it always stutters, meaning the fans will spin, the lights go on...for 2 or 3 seconds, then it shuts down. It immediately starts back up (on its own) and repeats the same process. (that's 2) On the 3rd try it will go through all the way and boot into Windows as if nothing was wrong. It's been doing that since I can remember and I never cared until the other night when I came home and tried to wake the machine from sleep, found it was totally dead. It came back to life after I tinkered around a bit and as I type it appears nothing is wrong with the system. It's only that damn boot process that shows something is amiss.

Thanks again for your replies S...I appreciate it.

[sluggo]

If you can get the machine to boot, you can test the voltages under load. Put your voltmeter's ground probe into any cable header with a black wire coming out of it, then probe the 24-pin connector with the voltmeter's V+ lead.

The machine boots fine - after 3 tries. But once the machine is running the 24-pin cable is connected to the mobo header. How can I probe it while it's connected? I'm sure I misunderstand you but maybe you can explain it again if you don't mind.

When the 24-pin connector is plugged into the motherboard, all you see of it is the "back" of the connector where the colored wires enter the body of the connector. You'll notice that each wire has its own little square housing in the body of the connector. What you will do is insert the V+ probe into a housing, right alongside the wire that carries the signal you're interested in (+5V, PON, etc ..). When you get in far enough, the tip of the probe will make contact with the metal terminal crimped onto the end of the wire and you can measure that signal. There's nothing to worry about - you can't break the connection nor damage the connector itself. This technique is standard practice in the industry.

Try booting the machine immediately after a shutdown - if it boots right away, the current sense may be the problem. Run the test exactly the same way a few more times to confirm.

It never "boots right away" it always stutters, meaning the fans will spin, the lights go on...for 2 or 3 seconds, then it shuts down. It immediately starts back up (on its own) and repeats the same process. (that's 2) On the 3rd try it will go through all the way and boot into Windows as if nothing was wrong. It's been doing that since I can remember and I never cared until the other night when I came home and tried to wake the machine from sleep, found it was totally dead. It came back to life after I tinkered around a bit and as I type it appears nothing is wrong with the system. It's only that damn boot process that shows something is amiss.

Thanks again for your replies S...I appreciate it.

I guess I'd start from ground zero. Do a thorough blowout of the dust in the PSU and the motherboard. Re-flash your motherboard's BIOS and reset it to default condition. If you have a spare motherboard around try booting that with this PSU. If it passes, suspect the original motherboard. If booting fails with the new motherboard, then use your voltmeter to test the wall voltage. If the wall voltage is in spec, then you've got a dodgy PSU.

Tell me more about the UPS. You've got a lot of power demand at start-up - what's this UPS rated for? Also, does the machine go through this restart sequence when the UPS is bypassed and the machine is plugged directly into the wall?

[canoli]

...insert the V+ probe into a housing, right alongside the wire that carries the signal you're interested in (+5V, PON, etc ..). When you get in far enough, the tip of the probe will make contact with the metal terminal crimped onto the end of the wire and you can measure that signal. There's nothing to worry about - you can't break the connection nor damage the connector itself. This technique is standard practice in the industry.

Beautiful - never knew I could do that.

I guess I'd start from ground zero. Do a thorough blowout of the dust in the PSU and the motherboard.

I do this regularly - obsessively even, so it stays fairly dust-free inside my case.

Re-flash your motherboard's BIOS and reset it to default condition.

I'm a little hesitant to do this as I've always lived by "never flash an unstable system." I updated the BIOS once since I got this board - because I really had to - and all went well but because I ...

If you have a spare motherboard around try booting that with this PSU.

...don't have a spare X58 board I think I'll skip this step for now. There's an AMI update (v1.7) for my board on the MSI website, which I'd like to install once this power issue is solved.

Tell me more about the UPS. You've got a lot of power demand at start-up - what's this UPS rated for? Also, does the machine go through this restart sequence when the UPS is bypassed and the machine is plugged directly into the wall?

The UPS is rated 1000VA/500W and as I understand it the 500W signifies what's available on battery backup (for a few minutes). The volt-ampere - read the wiki, still not sure what that's all about.

Anyway it's the Tripplite OMNIVS1000, which I bought about 5 years ago - http://www.tripplite.com/en/products/model.cfm?txtModelID=2656

I never tried plugging the PSU directly into the wall but at some point in the next few days I will. (so simple it's brilliant )

Thanks!

[Convert]

I've personally ran in to three distinct causes for this kind of startup phenomenon.

1. It was either a P5B or P5N-E but the CMOS jumper was loose.. Yes I know, this sounds crazy but bending the pins slightly made the issue go away. I did not figure this out on my own, I came across a thread of people having the same problem and it resolved it for them.

2. Overclock/improper SPD being picked up.

3. Bad motherboard.

I'm sure there are plenty of other potential causes, but I haven't seen a PSU do this.

[canoli]

Thanks for your reply C - sounds like you're referring to an Asus board maybe? Not sure what those pins correspond to on my MSI board.

I can rule out a wonky OC because I dialed everything back to default. But "improper SPD being picked up" sounds interesting because one of the other "bugs" this board has is it won't always recognize all 6 DIMMs when it boots. I've found a combination of clock/latencies that give me about a 90% chance of a successful boot - but 1 in 10 tries it will see 16G instead of all 24. Whether that's connected to what you wrote I don't know. I ran memtest(86) several times and all my sticks check out fine, no errors.

Thanks again for your reply.

[Convert]

Yes, it was an ASUS motherboard. It looks like your board doesn't have a CMOS jumper and instead uses the CMOS reset button on the back of the IO plate. Check to see if this button is free from obstruction, the Intel boards that have this feature commonly have anIO plate that is punched out incorrectly so the button is always depressed. I wouldn't think this is your issue though, seems like it would be pretty obvious as it would be telling you its been reset.

Try running it with three sticks of ram since you are on LGA 1366, see if the problem goes away then. Oddly enough my P6T, which is also LGA 1366, had the same problem. I don't remember my original memory config but I think I had 6x2GB sticks and it would not see some of them during a cold boot, if I restarted though they typically came back. I don't think I had the cold boot issue on this board though, I might be confusing it with the P5B/P5N.

A BIOS update fixed the RAM issue for me. I would also be hesitant to do a BIOS update with a flaky system but I'm not sure if I would be worried about it with your symptoms. The system runs perfectly fine with the exception of the cold boot issue correct?

[canoli]

yes exactly, once it's up n running there's really no problems except for one nagging thing:

every so often one of my F3s drops offline. I thought I solved it with a new data cable but it's started again. Same drive, doesn't matter what port, what power cable, just every now n then it drops offline...then comes back on its own. It's aggravating but I've never lost data because of it. Various utility apps all report a healthy drive and the read/write speeds are right where I'd expect them to be, similar to my other F3.

The problem hits so randomly and unfortunately Windows Event Viewer tells me nothing about it, as if it never happened.

Besides the hiccups on boot and the Spinpoint disappearing once in awhile the system runs great. I do long renders out of Cinema 4D, the 980 chugging away for hours and hours at 100% (20, 30-hour renders are not uncommon) and it stays under 62C. OpenGL-intensive routines (certain filters in Premiere Pro) kick the 470 up to ~74C. So I don't know....maybe just have to live with a few nags.

Thanks for your replies C!