How does a PSU die?

I've been wondering this for a while, probably due to my paranoia of my PSU getting "old." And it being an OCZ. Obviously, they can burst into flames or spark and die, but do they ever just wear out? By that I mean do they gradually decrease their power output over time and simply fail or become no longer able to meet its demands?

I've had my PSU for around 6 years, no less, and I'm getting worried. Do me a favor and help me sleep at night, or justify the expense of a shiny new unit!

Caps wear out mainly. Thus voltage regulation and ripple increase, to the point of it finally dieing.

Electronic parts sometime faill with smoke and fire. Rare.

6 years is pretty old, but if working fine, great.

Make sure your new purchase is a upgrade in wattage and has great amps on the rails. You know you want the newest CPU/GPU in a year. I'd just wait till then unless it dies.

JohhyGuru is the place you should visit.

I'm not so sure how fine it's working. I'm not too handy with multimeters. I used to be able to do certain high OC's on my rig but now, with the same settings, it isn't happening. I've ruled out a lot of things except the PSU. I'm eyeballing an upgrade to the Bulldozer, and I think it's about time to start with a new PSU. SeaSonic seems to fill the bill, a X-750. Wouldn't mind putting my current PSU on a basic light-duty computer. But if it's about shot, I'm not sure the point of it.

Conumdrum said:

JohhyGuru is the place you should visit.

Click to expand...

+1 on that, most definitely.

Given the risk that your PSU could take out some/all of the valuable hardware it's connected to when it dies, I'd certainly making the small investment in a new PSU.

Also, make sure you are using a high quality surge suppressor instead of a power strip.

PicodeGallo said:

I'm not so sure how fine it's working. I'm not too handy with multimeters.

Click to expand...

If you cannot use a multimeter, then you have can never use something so complex as a mobile phone or iPod. Set the dial to 20VDC. Touch probes to wires. Read number. Why is that difficult? Fear. Only widely promoted fear.

Conumdrum has defined a most common type of degradation. Sometimes a supply is purchased defective. It still boots and runs a computer for months or years. Eventually degrades into failure. A failure that could have been averted if originally confirmed by the meter when the supply was under maximum load.

Is your supply degraded? Six years should never cause a supply to fail. That is for supplies in general. Nobody can say anything about your supply until you provide numbers. Every answer is only as useful as facts that you first provide. Especially those numbers.

BTW, no power supply failure should ever cause any other computer damage. As was true with power supplies long before the IBM PC existed.

westom said:

BTW, no power supply failure should ever cause any other computer damage. As was true with power supplies long before the IBM PC existed.

Click to expand...

Might want to verify that statement with many known instances and ask JonnyGuru too.

Conumdrum said:

Might want to verify that statement with many known instances and ask JonnyGuru too.

Click to expand...

I said no power supply *should* ...

Most computer assemblers have no idea how electricity works. Do not know that all supplies most contain a long list of functions. Many computer assemblers never verify manufacturer spec sheets. Would not know what to read. Often use hearsay as proof that a supply is good.

Computer assemblers create a market rich for dumping. Larger wattage supplies, missing essential functions, can be recommended by computers assemblers. They only view dollars and watts. Assume if a computer works, then the supply is good. A problem traceable to insufficient electrical knowledge.

Worse, a computer assemblers often does not know that only they - not a supply manufacturer - is responsible for meeting those design and safety standards. No supply manufacturer is required to meet any standard.

No supply must even damage the load. And no load must ever cause a supply failure. A requirement that existed in computers long before the IBM PC even existed. That existed even when some disk drives used motor oil to move the heads.

The word was 'should'.

westom said:

If you cannot use a multimeter, then you have can never use something so complex as a mobile phone or iPod. Set the dial to 20VDC. Touch probes to wires. Read number. Why is that difficult? Fear. Only widely promoted fear.

Click to expand...

Well don't I now feel insignificant for not knowing how to use a multimeter? I in fact do know how to use one, I just don't have one. I've never tested PC's with one and I'm not an electrical engineer. After reading the other recommendations, I've taken their advice and have realized how I might go about testing a power supply, as well as gaining information and insight into other aspects of PSU's. Too bad you couldn't reference information as well.

Next time, try teaching rather than condescending remarks. If you aren't up for that, just keep to your own happy place. I don't mind you being blunt, but when you can't back it up with useful information, it doesn't come off very well.

PicodeGallo said:

Next time, try teaching rather than condescending remarks. If you aren't up for that, just keep to your own happy place. I don't mind you being blunt, but when you can't back it up with useful information, it doesn't come off very well.

Click to expand...

He once said, in a Usenet newsgroup, under the user name w_tom, that quality PC PSUs contained crowbar circuitry to shut them off in case of overload, but several people who understood electricity, were proficient in use of 3.14159-digit multimeters, and who lived in castles protected by whole-house surge protectors said he was wrong. He didn't just say those PSUs 'should' have contained crowbars but that they actually did. He also mentioned foldback current limiting, but I thought foldback limiting was about cutting the output current way back from the short circuit current so the PSU wouldn't keep pumping full current into the load, while PC PSUs instead just shut down completely and won't restart until the AC is disconnected. Here's the message thread, from Dec. 19-26, 2001:

http://groups.google.com/group/aus....1b5b4a/26fd2a261537e17e?hl=en&q=w_tom+crowbar

http://groups.google.com/group/aus....1b5b4a/26fd2a261537e17e?hl=en&q=w_tom+crowbar

larrymoencurly said:

He once said, in a Usenet newsgroup, under the user name w_tom, that quality PC PSUs contained crowbar circuitry to shut them off in case of overload, but several people who understood electricity, were proficient in use of 3.14159-digit multimeters, and who lived in castles protected by whole-house surge protectors said he was wrong.

Click to expand...

Yup, the same self-professed "expert" who came trolling around in this thread. Nothing like googling a huge host of buzzwords and then quoting yourself as an indisputable "authority."

PicodeGallo, you may want to get a new unit and then see if the folks at Johnny Guru would be willing to test the old one. If it tests well, you can keep it as a backup/htpc/folding/NAS PSU.

Good reply, BTW, PicodeGallo, don't let the trolls get under your skin.

larrymoencurly said:

[...] but several people who understood electricity, were proficient in use of 3.14159-digit multimeters, and who lived in castles protected by whole-house surge protectors [...]

Click to expand...

I almost spit out my morning coffee reading this line.

I've been testing with P95 and it's not going well. It crashes my computer shortly after starting. I've checked voltages with CPUID HWmonitor and they look a little off.

+3.3 - 1.20V
+5 - 4.89V
+12 - 11.83V

These are averages. The 3.3 & 5V rails are pretty stable, but the 12V rail never hits 12V and I've seen it dip to 11.78V. I've checked pretty much everything in my system, with the PSU being the only thing left. I still don't have a multimeter to check actual voltages. I just wanted to run this by you guys to see what you think while I browse PSU's on Newegg.

PicodeGallo said:

I've checked pretty much everything in my system, with the PSU being the only thing left. I still don't have a multimeter to check actual voltages.

Click to expand...

The onboard volt meter is historically unreliable until calibrated with a multimeter. If your 3.3 volts is that low, the machine is not even boothing. If it boots, then numbers reported by CPUID HWmonitor make no sense.

If psu is causing problems, then other system components should be behaving as if defective.

PicodeGallo said:

+3.3 - 1.20V

Click to expand...

Picode, can you confirm that number, or is that a typo?
Just making sure I'm understanding the situation properly.

Thanks bud,
-JT

CPUID hwm shows 1.20V. I don't think that's correct. The Asus PC Probe is showing 3.29V. It also shows higher voltages on the +5 & +12, 4.99 & 12.17 respectively. I'm a little wary of these monitors; I know they aren't very accurate.

But this Prime 95 thing is telling me the real story. When I set P95 to "Blend", it will run until about pass 8 or so. When I set it to "Small FFT's" to test the CPU, it crashes within a minute. Memtest shows the RAM is good. When I first put this computer together, I was achieving some pretty amazing OC's P95 stable. Now, I can't come close. The only thing I've done now is raise the multi a bit--everything else is stock. I can't do much more this if I want the machine to boot. I'm also experiencing crashing in games or anytime the system sees a load for more than a couple of minutes.

I know the temps are in check because: 1. The monitors never get above 47C (@ 1.42 Vcore) under load (before it crashes), 2. My Corsair A70 blows the air out of the top of the case and it's not very warm when I put my hand over it. I've tried various Vcore voltages and the effects are all the same.

So, I'm at a loss.

I'd lean towards the numbers shown by the Asus PC Probe. Having said that, definitely invest in a DMM when you can. A cheap/starter DMM can be had for ~$20 @ RadioShack.

Still replace the PSU, definitely, but a DMM wouldn't hurt either.

I just looked up my PSU and here's what's listed:

+3.3@36A,+5V@30A,+12V1@18A,+12V2@18A,+12V3@18A,+12V4@18A,[email protected],[email protected]

http://www.ocztechnology.com/ocz-gamexstream-600w-700w-power-supply-eol.html

Now, I have the older one, the one without the 6+2 PCIe connector. I believe this is the old 4x12V EPS12V style in which the PCIe connectors are on their own rails instead of on just one. Is it possible my GTX 470 is "stealing" power from my other components? After all, in 2006 video cards didn't need the power they now do, and 18 amps per rail doesn't leave a lot of room for other stuff, and I've got a lot of other stuff.

I could be talking out of my ***, as I'm still trying to figure this stuff out. I think I'll drop by radio shack today and find a multimeter.

Don't believe the numbers Asus PC Probe gives you. With my P6T mobo, PC Probe says my 3.3 rail is at 2.93v, 5 volt rail at 5.53v and 12v rail at 13.32v. Bios shows voltages right around those figures too. Checking with a digital multimeter, my 3.3v rail is 3.35v, 5v rail is 5.08v and 12v rail is 12.11v. So the board is misreporting all 3 voltages. Don't ever trust any readings the mobo gives you as being "real". Use a meter to check voltages. Once you know what the real voltages are as compared to what the mobo is reporting them as, then you can use the board's readings as a guide as to whether you are having a power delivery problem by looking at a change from their "normal" readings.

If your OC fails when it used to work you could have a voltage droop issue with either the psu or the motherboard. The components that wear out in a psu can also wear out in a motherboard.

I have a 10 year old enermax psu that has been in use much longer than it ever should have, but it is still alive and kicking (needs to be put out of use, but not until I get a new psu for my HTPC). The main thing that will happen when the capacitors age in a psu with be decreased efficiency. Voltage can droop lower than they used to, but it would be somewhat easy to see that happening. You could use OCCT to run a stress test and monitor all the voltages to see what is going on (not as accurate as a DMM, but will at least tell you something). That is a pretty good place to start. Also, run this test without an overclock. No sense damaging your cpu if it isn't getting enough voltage.

Issues like this aren't that easy to pin down. As long as your PSU is a good quality PSU, then you shouldn't have issues unless something goes wrong. I've had VERY old AT psu's run in old 486 builds that were well over 15 years old.

I tried OCCT. While the voltages are all wrong, I decided to run it anyway. I liked the readouts it gives you. I ran a short test to see if I could detect anything. There were some Vcore dips, one while monitoring then the rest of the time, it just dipped and stayed there. I ran the test again intending to let it go all the way...then my computer crashed--it shut off again just like in P95. Unfortunately, there were no graphs for this.

I find myself coming to the conclusion that power delivery is very suspect. Is the PSU bad? Probably not--it does allow me to surf the web and play music and movies. However, I don't think it's up to the task of supplying enough clean power to demanding components anymore. My particular power supply is notorious for excessive ripple, and I think it's taking its toll on the unit resulting in tired regulators. I think this is resulting in the Vcore dips I saw in the graph, and why the computer is shutting off.

So, I think I've settled on a Corsair TX750 v2. Good performance and good price, and it will likely be relegated to a BD/Sabertooth or Crosshair V build later on. I do need another PSU, as I can use this OCZ in another computer that won't be OC'd and doesn't play games (a very simple HTPC).

Thank you all for the help. If this doesn't work, I'll be back!