4

I want a way to evaluate and get a score of how well a power supply is functioning, without removing the power supply from the computer.

Background
Where I work, we still have a bunch of 6-8 year old Windows desktops in production that we're about to (finally!) refresh. Most of them have had an upgrade at some point, but the typical machine is still an old P4 with 1GB RAM. By far, the single most common point of failure on these machines are the power supply (hard drives are #2).

Because of budget constraints, we're replacing the old machines with off-lease 3 year old machines. The new specs are Core 2 Duo 2.33Ghz with 2GB RAM for about $200. The replacement machine was a high-end desktop when it was new, and so the build quality is very good. That's a huge step up for not a lot of money. My one concern is that these are small form factor units, with non-standard form factors for the power supplies.

Goals
I have two purposes in mind for the rating the power supplies.

First, as the old machines begin to go out of service I want to harvest the old power supplies for use in others that haven't been replaced yet. The machines are slated to be replaced soon anyway. It makes no sense at this point to spend any money replacing a dead power supply with a brand new one for a machine we won't keep long. To be effective at this, I need to know that the replacement part isn't nearly as far gone as it's predecessor. I want to work on a "best available" replacement policy, but to do that I need to know which of my inventory are the best units.

Secondly, since the new machines are already a few years old and given our recent concerns with this specific part, I want to have a better idea of the status of each power supply I'm putting out for the new machines. Even if I need to replace the power supplies of all my new machines within the first year, I still feel like I'm getting a good deal. It's just that in that case if I can test up front and know that this will happen I can save a lot of effort and frustration by doing it up front. On the other hand, if they're in reasonable shape budgets are already tight and so I'd like to avoid the expense. Hence I want to be able to easily and quickly get a score for the power supply on each new machine when it arrives.

Summary
I want an easy way to pin the health of a power supply down to a simple score. I should be able to do this without removing the power supply from the machine. A windows software program would be my ideal option, but a linux tool that comes with a live cd or an inexpensive hardware device like a kill-a-watt that reads draw is a possibility also.

3
  • 3
    Do some of the SFF systems have graphics cards installed? If some end users won't need the accelerated graphics, consider removing the card on those systems to lower the internal temperature. We have a bunch of gx280 SFF's and gx620 SFF's where I work and we experience similar issues.
    – LawrenceC
    Commented Jul 22, 2011 at 19:46
  • Is it okay if you have to open the machine as long as you're not extracting power supplies?
    – Shinrai
    Commented Jul 22, 2011 at 19:46
  • @shinrai - I'd prefer not to, but if that's what I have to do so be it. Commented Jul 22, 2011 at 19:58

4 Answers 4

2

There is no way to predict ATX PSU faulures without even taking a visual inspection at the unit. But if you opt for taking an inspection, consider that the 110V (230V) rectifiers are the components most prone to failure due to severe heat up caused by poor PSU cooling due to fan dust, sleeve bearing wear, undersized heatsinks typical of cheap PSUs, etc.

These rectifiers are designed to hold up to 130°C before shorting, which they do. A visual inspection will preventively determine if there are smoldering burns beneath the rectifiers before the unit will fail during use.

Also, as Spiff mentioned above, PSUs from the early 2000's may have been assembled with certain brands of Taiwanese capacitors which had an unstable electrolyte causing eye-visible bumps.

0
1

Unless you put the power supply under some form of realistic load, you are not going to get any reasonable indication of performance. Even then, getting any form of expected future life indication will be near impossible. In reality, you could do some electrical tests and an internal visual inspection & clean of fans and components, but you should cost the value of your time doing this against the cost of a replacement PSU - I know you said they are a non-standard size, but I'd still try and track down a price and compare this with, say, 30 minutes of your time.

1
  • This is very true. The best guarantee of longevity is to pick a PSU that is manufactured by a reputable manufacturer (yes, the one with actual factories, brand name is not interesting). If you already have old PSUs at hand, apart from looking at obvious signs of wear and tear, it's very hard to predict their future.
    – Zds
    Commented Aug 9, 2011 at 14:40
1

Have you done failure analysis (FA) on the PSUs that have failed? It would be nice to know if there's a particular failure mode that you're looking for.

For example, 8 years ago, or maybe even 6 years ago, some vendors were still getting hit by the capacitor plague (see also BadCaps.net), where low-ESR aluminum electrolytic capacitors had the wrong electrolyte formula that was eating away at the aluminum, building up hydrogen gas, and eventually causing the caps to vent and release the magic smoke. In my experience, those caps usually bulge on the top before they vent.

If your FA data says that most of your PSUs failed due to capacitor plague, then if I where you, I'd try to shine a light in the vent holes of the PSUs and try to see if any of the caps are bulging. If you have good FA data then maybe you know exactly which caps you need to be looking at. Maybe rig up a tiny camera component and bright white LED (like the kind used as the camera and flash of a smartphone) on the end of a little stick that you can insert in a vent and have a look around. Kinda like a homebrew tiny version of a RIDGID SeeSnake Micro or a Milwaukee M-Spector.

0

I'm afraid there's no way to get a useful value/score for the health of the PSU without opening the case. Once you do so however, you can get your scores relatively easily with a PSU tester. Here's an example device (rated highly on amazon, reasonably priced)

http://www.amazon.com/Coolmax-Power-Supply-Tester-PS-228/dp/B002R06PGE/ref=sr_1_24?s=electronics&ie=UTF8&qid=1311375363&sr=1-24

I have no affiliation, just want that clear.

Update: I apologize - I thought that the unit above applied a load. The other commentors are correct, to really test the PSU you need to apply a realistic load to it, then read the assorted voltages. Apparently none of these low-cost PSU testers do much more than read the unloaded voltages; useless for determining if the PSU is dodgy.

2
  • Do you know if that product or any similar product tests the PSU while under load? I've seen PSUs with bad caps that put out proper voltage under no load, but fail under load.
    – Spiff
    Commented Jul 23, 2011 at 5:12
  • These types of tester generally just give a voltage reading for the power rails. For proper testing you'd need a load, as you suggest.
    – Linker3000
    Commented Jul 23, 2011 at 7:29

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .