There is a pitfall with electrolytic capacitors. If an electrolytic capacitor is visually damaged (bloated, leaked, exploded, …) it is dead. Reverse doesn't work! If an electrolytic capacitor looks good it may be working or may be dead too.

It's quite common that the input filter capacitors in aged switched mode power supplies looks good but have a large capacity loss that results in unstable operation or random shutdowns.

This is excellent information to know. It would be helpful to have a schematic of my power supply and a circuit board layout so that I could identify suspect capacitors more easily should I ever decide to attempt a repair. But I suppose that identification could be done seat-of-the-pants if one were determined to do it. Since in my engineering career I never designed power supplies, I did not know this tidbit of information. Combining that with the tidbit of info informing me that OST branded capacitors are suspect, we do move closer to a potential blind repair without a schematic and board layout. Whether I'll ever actually attempt this repair ... that question is still up in the air. But at least I am amassing information that would be of great help should I attempt it. Thanks!

Turns out replacing PSU caps is often easier than motherboard caps. Since ROHS standards started applying to motherboards, it's been tough to remove enough solder without burning the multiple layers of traces in the holes or injuring nearby components with excess heat. Most PSUs have been using easier solder to deal with, while the proximity of components to each other generally isn't as hard as you might think to work around. OTOH, determining in advance of removal the specs of many glued and/or siamesed caps is frequently impossible. I've been successful in resuscitation via mere cap replacement on motherboards 60%-70% of the time, while with PSUs, more like 90%+. I succeeded with 2 PSUs less than a month ago, both 250W Lite-Ons from HP mini towers not worth spending money on, but perfectly serviceable after replacing <$2.00 worth of parts each, all L-Tech, 5 2200uF 10V, 1 100uF 25V, IIRC.

Well, that just makes me less inclined to mess with the motherboard then. Which is a blessing in disguise. The thing uses DDR2 RAM, has USB2.0 ports, and is maxed out CPU-wise. I think it's about time for an upgrade - and I don't mean just the caps!

Have a look at this:
https://en.wikipedia.org/wiki/Switch...e_power_supply
The (commonly two) input filter capacitors are marked with "B" in first image at this page. Typical values for the two input filter capacitors are 220…470 µF / 200 VDC with higher capacities for more powerful PSUs. Use of two 200 VDC capacitors avoids use of one large (and more expensive!) 400 VDC capacitor (230 VAC gives around 325 V peak voltage).
"A" in first image marks line filter with common-mode choke, X capacitor (yellow, under the choke) and two Y capacitors (light blue) followed by primary side bridge rectifier (black).
Normally all other electrolytic capacitors inside a PSU for computers are low voltage capacitors at secondary side.

ROHS is one reason. The other reason is that PSU boards are two-layer PCBs but motherboards are four-layer-PCBs.

Hence why I wrote:

Looks like I have one filter cap rather than two. 390µF, 420v. And it's the potentially problematic OST brand.

That's a large sized cap, and should be easy to replace. Except it looks like they glued it in place around the base.

Thanks for all the help. You've given me a great place to start if I decide to repair this thing. My first step will have to be putting my new power supply in the computer and see if that fixes my original problem. We haven't proven that my problem is a bad power supply. That's just my working theory right now. I think it's a good working theory however. I did find that the fan in this suspect power supply is dead as well. I can barely turn it with my finger - very stiff, it certainly would not turn on it's own under power. But I have never seen that fan run since the day I bought the power supply. I always assumed it was because the supply was running cool enough that internal sensors decided it didn't need it. I have checked the temp of that power supply many times when the system was running. Both with "the finger test" - where I couldn't detect any warmth at all, and with an infrared thermometer gun - which always showed room temp. That would be a reasonable finding since it's a 600 watt supply sitting in a computer that probably only draws 200 watts at full load. And it's extremely rare that it has any load on it at all. It's just my email and web surfing computer. There's just nothing in there ... motherboard, CPU, SATA SSD and a 5400 rpm hard drive. No expansion cards. Well, it does have a lot of fans (part of why it runs so cool) - four 120mm case fans. But those shouldn't draw a ton of power. The whole system - power supply and all - are about as unstressed as things could possibly be.

Attached Thumbnails

 

Word on badcaps.net forums is the big primary filter caps are uncommonly the failures in PC PSUs, regardless of brand, thus not the first place to check in a failing PC PSU. Most commonly by far IME the failures are on the output side, often 3300uF 6.3V and/or 2200uF 10V or 16V, in 10mm diameter, which can be troublesome to match with replacements of same can size.

Update:

I replaced the power supply in the computer with the new EVGA model I bought. It booted up, but did a "lights out" crash again 1-1/2 hours later. Pushing the power button to reboot resulted in the same behavior as before the power supply replacement - fans would spin up, then everything went dead after one second. I had been monitoring temps, and all were in the low 30's (Celcius). Specifically, the CPU was 34 Celsius. Since this behavior does indeed make one think thermal shutdown, I immediately opened the case and touched every heat sink on the motherboard (grounded myself to the case first!) All were cool. The only one I could detect any warmth on at all, and it was barely above ambient temperature, almost undetectable as "warm", was the heatsink on the chipset (which is AMD780G).

So this shutdown problem does not appear to be power supply or thermal.

Am I dealing with a motherboard issue here? Is it at all possible that me replacing the memory a few weeks ago could be the cause? I went from 4 sticks (4Gb) to 4 sticks (8Gb). It was a complete replacement, I am not mixing old and new memory together. The computer ran fine on this new memory for a few weeks before this shutdown issue began. However, this memory replacement is the only thing I've done to the computer in eons, and any new change is a potential suspect. Even though this does not feel like a memory issue to me. I nave visually inspected all caps on the motherboard and nothing is bulged. But that is a visual inspection only.

Does anyone have any other thoughts on what could be the root cause? Replacing the motherboard (and CPU and memory and CPU cooler, since a new modern mobo would require that) would almost certainly fix the issue. Because at that point, I would have a new computer. That would definitely be hitting things with a big hammer, but if it is indeed the motherboard that is the cause, there's not much else to do other than replace it.

I guess my next step will be to put the old memory back in and see if the shutdowns continue. Other suggestions?

Did you check the motherboard for OSTs? They have a reputation for stealth failure. I wonder if your motherboard model is the same as any I had fail after warranty expiration but long before expected. Second youngest one was a PC Chips, socket 775, HT Pentium, i915 chipset, 4 RAM slots, which I originally filled with a pair of PC4200 256M. I was unable to resurrect it for more than a short time with new caps. I replaced it with an eCS board with P965 chipset, also 775 & HT P4, later upgraded to Core2Duo that is still in use, but seems may have gotten flaky recently. I replaced a bunch of its caps not long after warranty expiration. My point is, early socket 775 motherboards seem to have less expected longevity that boards pre-dating the bad caps fiasco, and those with sockets newer than 775. AFAICT, 775 mostly coincided with ROHS implementation teething pains better worked through only as 775 and DDR2 were becoming obsolete.

Given DDR2, yours is clearly old, and it's a shame you spent the money for 4 new sticks you won't be using in anything you replace with that isn't as ancient. I agree, reducing electrical load by going back to the two old 2 RAM sticks might help. Maybe try a pair of the new ones alone.

I did not check the caps for brand, only for signs of bulging. I would buy a new mobo before replacing caps simply because of their brand. I'm not denigrating your suggestion here, just mentioning what I'm willing to do. As you mention later, this mobo is indeed old. Only so much effort is realistic to try and resurrect it, if indeed it is the problem.

Specs:

Motherboard: Gigabyte GA-MA78G-DS3H ATX (socket AM2+ I think?)
CPU: AMD Phenom II X6 1045T 2.6GHz 6 core (normal run temp in the low 30's Celcius)
RAM: Was: 4x1Gb Kingston KVR800D2N5K2/2G, Currently: 4x2Gb Kingston KVR800D2N6/2G (as of about a month ago)
PSU: Was: OCZ OCZ600SXS 600watt, Currently: EVGA 220-P2-0650-X1 a.k.a. Supernova 650P2 650watt Platinum (as of a few hours ago)

In terms of power, resources and functionality, this computer is #5 in what I have sitting around me. It serves (or "served" as the case may be!) as my desktop computer - browsing the web, email, running Calibre, LibreOffice, GIMP, and other lightweight stuff like that. GIMP isn't exactly lightweight, but for the kind of stuff I normally do with it, my desktop computer is fast enough usually.

See mobo model number above.

That 8Gb of DDR2 RAM cost me $17.38, from a large eBay seller in China, to my doorstep. Advertised to be, and indeed appeared to be, new, not used. I'm not going to sweat that. It costs me more than that to get a BBQ sandwich, fries and a Dr. Pepper at lunch. Nor will I sweat the $75.68 to my doorstep the new EVGA power supply, that did not solve the problem, cost me. I wouldn't sweat the $150-$250 for a new mobo either, except I'd have to add $200-$300 for a new CPU, $100-$200 for new RAM, and $50-$100(?) for a new CPU cooler to that mobo cost. I could easily add a new M.2 SSD, but if push came to shove, my older SATA6 SSD could still work - it's only been very lightly used, plenty of life left in it. I've got plenty of new, large HDD's laying around, new in-box. But in truth, that mobo indeed needs to be replaced. It's DDR2, USB2, legacy BIOS, can't boot off of USB, etc, etc, etc. A real antique (but one that still serves my needs for the most part). The "backup computer" I am typing on now is even more antique: an Intel Core2 Duo E8400 with 4Gb of DDR2 RAM. I'm running a "frugal install" of antiX on it now, and it's actually quite functional. It's not a pain to use in any way. I'm impressed with what antiX did for it.

I don't think it's the load from the RAM chips. If a 650watt PSU can't power a mobo, a 95 watt CPU, 4 sticks of RAM, a few fans, and one little USB30 add-in card, something is seriously wrong. I was thinking about something in the RAM somehow triggering some PSU protection function. What that might be, I have no idea. Just speculating. It doesn't act like a short, it acts like something thermal. But if I can reach in and touch every heatsink on the mobo and not be able to detect any warmth at all - that's not the footprint of a thermal shutdown. I would expect my fingers to still be sizzling and seared to the heatsinks if thermal was the issue.

I wasn't trying to suggest replacing caps in a 15 year old motherboard, only a possible point of failure that could result in random lockups after everything is fully up to operating temperature.

If I found the right page, top notch for its day, with solid poly caps instead of electrolytics, so there's marginal likelihood caps are at fault, and less that you could determine which one(s) if you even considered to try to replace any.

My meaning of electrical load had nothing to do directly with the power supply wattage. The motherboard has at least one voltage regulator. It also has a zillion components and traces connecting everything. Some little bit somewhere could have become out of spec, thus occasionally overloading something else, manifesting as random old age failure. We like to think of Gigabyte is reliable, but when old age has set in, anything is susceptible.

Yeah, that's it. Tomorrow I will pull out the new memory, inspect the slots (and memory sticks) closely with light and magnifying glass. Clean as needed. Re-seat the memory sticks and see if things miraculously improve. If not, I'll put the old memory back in for further testing. I did mark the old sticks so I know which slot each of them came out of. I must have had a premonition of bad things to come when I marked them - that's not something I have routinely done in the past, although I should, and I will from now on.

It's way too early to declare victory. However, after doing a really good cleaning with Dust-Off of all the memory slots, and re-inserting all the memory sticks into different slots than they came out of (still using the 8Gb of new memory) - the computer has stayed up for six hours so far. When it's been shutting down in the past, it has been shutting down in the two-hour-ish timeframe after booting.

One thing I did notice when re-inserting all the memory sticks, was that they seemed harder to fully seat than they did previously. This may simply be because I was at a different, more awkward angle when working on the motherboard this time. Or maybe I did not have all the sticks fully seated the first time. Or maybe I'm just imagining it. I do know that memory sticks can sometimes require more force than you'd think to fully get seated, that is not a new concept to me. So I would have thought I'd have remembered that on the first go-round. But who knows, maybe I didn't. I have my fingers crossed hoping that the computer will stay up this time.

I am not going to begin trusting this computer again until it's stayed up for a week, and that's still a long way to go, but at least I am slightly - slightly - encouraged at this point.

I think I have found a smoking gun for my computer problems. But I will start a new thread on the issue rather than continuing it here.

My problem appears to be heat related. Or more accurately, "false heat related". When I exercise the CPU, and it rises to 48 Celsius (which isn't that high - well within the specs of my CPU), my system shuts down. I have checked BIOS, and it is set to "alert" at a CPU temp of 70 Celsius. So there should be no thermal shutdown going on at this low 48 degress temp, but that's what appears to be happening none-the-less.

Anyway, I'll start a new thread with updated info.