Supermicro PWS-0060 *failure*, please advise

[tiresias]

Hi guys,

Yes, as the thread's title suggests, all is not well with this PSU, or at least, this particular specimen I've been using for a few months now.

This is what happened:

My workstation (NB. Hardly power hungry, a Tyan S2865 with opteron 146), suddenly shut down this afternoon - which I found bizarre since stabilitywise, it's always been flawless. Shutdown was instant, with +5vsb still being supplied, and all other voltages not.

After restarting it, it shut down in the same way after 5 minutes. On a third restart, it went out after a few seconds. And after that, instantly.

After much cursing and replugging and pulling the MB out of the case, etc., I tested another PSU to find the MB/CPU combination worked fine.

I subsequently tested the 0060 PSU and found:

- With no load, the 3.3vdc line is at an extremely abnormal 1.9v (verified). All other rails are OK.

- Any load on 3.3vdc causes it to shutdown.

Any advice (in terms of how to repair), what secondary damage could have been caused, etc. would be much appreciated.

[jonnyGURU]

Re: Supermicro PWS-0060 *failure*, please advise

Maybe the same issue I saw with the voltages suddenly dipping and then popping back up.

Undersized caps???

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

Maybe the same issue I saw with the voltages suddenly dipping and then popping back up.

Undersized caps???

You mean that there could be (undersized) caps that have failed? As in physically blown? Or induced a failure elsewhere?

Note that this PSU was used with a machine of power consumption far, far lower than its rated output, on all rails. I've also seen such 0060 units run power-hungry dual Xeon workstations for years without problems.

[jonnyGURU]

Re: Supermicro PWS-0060 *failure*, please advise

Induced a failure elsewhere.

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

I'll test this out on my rig and report back tonight. What is the exact procedure you followed tiersias and what kind of a setup are you running? I'm gonna try and duplicate your setup as closely as possible.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

I'll test this out on my rig and report back tonight. What is the exact procedure you followed tiersias and what kind of a setup are you running? I'm gonna try and duplicate your setup as closely as possible.

My setup is actually very low-power, the kind of thing that even a 300W PSU should handle without any trouble:

- AMD Opteron 146
- Tyan S2865G2NR board
- 2 GB Crucial ECC RAM (2x1024M)
- 1x WD Raptor ADFD 74M, and 1x WD Caviar RE 160M, both on the nVidia MCP's SATA ports
- All contained in a Yeong Yang W201B server box (cooled by a 120x38mm Sanyo Denki fan (0.52A))

The strange (and most frustrating) thing is, that it failed like this after a few months. So it's not a case of it being DOA, or due to "aging" components.

Note that I'm even using the onboard ATI Rage XL, which uses a few watts at the most. Not even running burnk7 for hours in a warm room would be enough to cause the PSU's fan to speed up past its minimum.

As far as procedure I followed, you mean to test it once it had malfunctioned, or?

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

Mate,

Here is what I have set up.

Opteron 146 (at stock)
1 Gb RAM (BH5 at stock)
1 SATA 36Gb Raptor
1 x1800xl (stock)
1 PCI VGA card (generic crap)
1 Optical
1 DFI NF4 SLI DR MB

No add on cards no nothing. I boots up fine even with the HDD or Optical. Just Video RAM and CPU. Swapped the x1800xl for the generic video card, still no shut-downs.

As of now, I saw no problems. I'm going to fire up the scope and check the rails in a little bit. Still its odd that your unit completely failed after a few months?

I am going to turn up the volts on the CPU, MB and BH5 and run Memtest #5. I'll report back with anything interesting I see.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

Still its odd that your unit completely failed after a few months?

It *is* very strange, particularly since I can't find anything obviously wrong with it after some preliminary testing.

Note that also it worked absolutely flawlessly until this early afternoon. The machine was just idle when it started to fail.

I had, shortly after getting it in March, put it through something of a load test, up to about 400W of current draw using a standard machine and additional wirewound resistors on the 12vdc out. No problems at all at the time.

I'm hoping that one of our resident electronics experts has seen something like this before and may be able to point me to the right direction as far as a defective component is concerned....

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

Just noticed Jonnys voltage jump on the 12V line. When I'm messing with the BIOS, DMM reads 12.14V, when I reboot and begin to run Memtest, rail jumps to 12.24 V.

No such thing with 5V though. Very odd.

Overclocked settings:
VCore 1.56V
VBH5 3.3V
VChipset 1.6V
CPU at 300MHz x 9
BH5 1t 246MHz 2-2-2-5.

I'm going to try to get this on my scope and use excel to plot it (very crappy way of doing it, but it should work).

PS*
Mate, what exactly did you do in your WW Resistor test?

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

Exact repeatable sequence:-

Pre-POST -12.12 V
During POST- 12.25, 12.19, 12.20
During Memtest - 12.21V steady.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

PS*
Mate, what exactly did you do in your WW Resistor test?

It was rather basic really - just to see if it wasn't DOA before I opened it up to take photos.

In essence, I connected it up to an old K7 machine with a couple of drives, to avoid crossloading, and then added, via two adapted molex connectors, two aluminium housed /wirewound resistors, 1R each.

Each of these drew, according to my Fluke, about 12.1A as you would expect. So both combined increased the draw on the 12vdc line by 24A, or 300W.

Total power was still well within the PSU's specs though.

...

PS. 12vdc line voltage decreased from 12.10 to 11.85 (about) when these were connected.

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

12.12 V in Windows and running 3DMark 06. This is odd. Why did the 12V jump up to 12.21 steady during memtest, while it is close to pre-POST voltages in Windows and 3DMark 06?

Hmm...that is also odd. Why should the 12V drop to 11.85 (not very good news if you ask me) on light loads? Now I'm throughly confused with the stuff you are saying and what I am seeing. Basically, these two units are almost the same, right?

When I turn the DMM to measure AC instead of DC, I see a max value of 12 mV. Ofcourse, this is in no way accurate as a scope reading, but still the AC noise seems pretty O.K to me.

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

Dude, is yours a Rev 3.0? Mine is.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

Hmm...that is also odd. Why should the 12V drop to 11.85 (not very good news if you ask me) on light loads?

Well, it wasn't much of a light load. In the end, the 3.3v wasn't that much loaded, but the 5V would have been pretty loaded up (I think it was 21A in total) by the overclocked K7, running BURNK7.

Add about 28A of +12v draw, and it would have been about 475W total.

...

PS. Rev 3? The PSU? Let me check for you.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

About the revision - which marking are you referring to exactly?

The large toroid and transformers are marked "Rev. A", while the PCB itself has, as far as I can see, a datecode only, 0447. In addition to the "LiteON", PSU model, etc.

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

Sorry for not being specific, I had to rush. I'm attending to an overclocked Presscot

The Rev is on the big label on the outer box. After the Achtung! warnings (bottom of the sticker).

Now, the Presscot + Video card booted up fine. Since Jonny's tests showed a few quirks, I wanted to see if this was enough to cause the system to shut down. I did discover a quirk alright. It was that the P5W-DH doesn't like a PS/2 Keyboard..lol. I had to switch to USB to get it to post.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

Now, the Presscot + Video card booted up fine. Since Jonny's tests showed a few quirks, I wanted to see if this was enough to cause the system to shut down. I did discover a quirk alright. It was that the P5W-DH doesn't like a PS/2 Keyboard..lol. I had to switch to USB to get it to post.

Right. I'll check that and get back to you.

I don't think even any overclocked dual CPU machine will draw enough power for that 46A beast to shut down due to overload. There is clearly something very defective with my one here, and I still haven't figured out what.

[Super Nade]

Re: Supermicro PWS-0060 *failure*, please advise

What I'm trying to do, is keep the 3.3 and 5V rails verly lightly loaded and load up the 12V. In Jonny's tests, he reported a 0.1V step. I'm trying to see if I can replicate that.

So, did you try it in another rig? That is really strange. Is there a fuse inside it which blew? I thought I saw something resembling a fuse (not sure if PSU's even have them).

My biggest complaint with this unit is that it does not have a freaking power switch. Why? Why? Why? Why?

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

So, did you try it in another rig? That is really strange. Is there a fuse inside it which blew? I thought I saw something resembling a fuse (not sure if PSU's even have them).

It has a fuse - it's under the little circuit board that "facing down" as it were. The fuse in series with the AC input though, so if that blew, the PSU would act completely dead - not the case here.

[tiresias]

Re: Supermicro PWS-0060 *failure*, please advise

As for the strangeness of its mode of failure, I agree, it has me baffled.

I've retested it with the fluke clamp meter, and the defect does appear to involve *solely* the 3.3vdc rail.

The 5v, 12v, etc. seem all still capable of putting out power, but the 3.3v is a fluctuating, unstable 1.9-2.1v at no load, and crashes to zero with load applied.

3.3Vsense is properly connected - I've checked that. Besides, a poor Vsense connection tends to send 3.3V high, rather than low.