View Full Version : werid psu problem, murphy is on the job ;-)

24-01-2003, 10:47 AM
i rebuilt a pc (1.3duron all-in-one mobo) and installed a 300 watt enermax dual fan psu (very nice) as the old 250 just didn't have the power to run it (suspect psu). however the enermax woouldn't turn on, so i threw in my spare 300watt aopen psu and odered a replacement enermax. poped out yesterday to install the new enermax and low and behold it wouldn't turn on. the 2nd enermax not to work. this was werid as they are very good psu's. i had a closer look at the voltage rails. the psu-on voltage was 5v on both psu's (unpluged) but when plugged in, 3v on the aopen and 0v on the enermax. disconected evrything including all the pci cards. check again. voltage 5v (plugged in) on both. plugged big hungry ihug sat card, no change. plug in modem (on its own) and voltage drops to 0v (enermax).

at the moment the pc is running fine on the aopen psu. i'm not to sure if the modem is drawing to much juice or the enermax psu just hasn't got enough juice on the psu-on rail.

trust murphy to throw a spanner in the works ;-)

Billy T
24-01-2003, 12:13 PM
Initially it sounded like the straw that broke the camel's back tweak'e, but from your description I'm not so sure. ?:|

When I strike these problems I start by monitoring the current drawn by each rail to see which has the heaviest loading and to find what the critical load current threshold is. Just knowing the wattage rating isn't enought because the loads are distributed across several 12 volt and 5 volt rails. It is possible that the quality (Enermax) psu may have individual monitoring for each rail as well and that could be responsible for the full shutdown.

However, that the Aopen supply drops to 3 volts indicates a fairly heavy load is present anyway so I wouldn't rush to blame the supply. I have seen this effect myself and reported it on F1 some time back. In my situation an elderly psu was not starting reliably and since the box was under 3 yr warranty I had it replaced. When I got it home the psu wouldn't start at all. Took it back for another psu, brought it home, same problem. Red face. :8}

Turned out it would not start until I unplugged the connection to one parallel port (I had three) and in the end I found it would start if I remove the power plug from my Imation Superdisk drive that was on one of the ports. Since parallel ports don't supply power to attached devices I can't imagine why it wouldn't start but an uprated PSU solved the problem.

I can't think why your modem would draw that much current but to get a conclusive answer you need to identify the device that causes the terminal overload and measure what it is drawing. I'd do this using the Aopen psu.

Keep us posted.


Billy 8-{)

Terry Porritt
24-01-2003, 01:48 PM
Maybe Billy has hit the nail on the head with the remark about the loads being distributed over the different rails, though I'd have thought the Enermax would have coped.

The current requirement for a 1.3 Duron is 31.5 to 34.3 Amps, quite a lot.

There is this business of the combined wattage rating of a PSU, but PSUs arent all the same, they can have different current ratings on different rails for the same overall wattage. I think this may be why some 300 watt PSUs will work ok, others dont have enough juice to start things going. It depends on whether the rails are independent, or maybe in the case of the Enermax, being high quality has shut-down protection on each rail if the current rating is exceeded.

It shows that 300W is barely adequate. Ive an Aopen 300W psu too, running a 1.333GHz Thunderbird, but the calculations of combined power say it isnt adequate, neither do the individual current raitings get anywhere near the 40Amps drawn by the cpu.
Im waiting for it to hiccup or give up the ghost :D

A similar thing is reported here (http://www.procooling.com/articles/html/picking_psu_s_for_athlon_syste.shtml)

The current for the 1.3 Duron is on page 43 of the AMD pdf specifications:
24310.pdf (www.amd.com/us-en/assets/content_type/ white_papers_and_tech_docs/24310.pdf)

(Incidently, having a hell of a job trying to preview, F1 is going bonkers again, so dont know how this will come out)

Terry Porritt
24-01-2003, 01:57 PM
A bad link to AMD, try this one:

Terry Porritt
24-01-2003, 01:59 PM
Oh well I give up on that link, just take my word the current requirement is 31.5A to 34.3A max.

Graham L
24-01-2003, 02:04 PM
So instead of buying a wimpy computer PSU you should go to an engineering supplies place, and get a DC welder? :D

This reminds me of the Burroughs 6930 --- it had a 4000A 5V supply.

24-01-2003, 02:06 PM
> (Incidently, having a hell of a job trying to
> preview, F1 is going bonkers again, so dont know how
> this will come out)

Just had the same problem, if you preview, and then click post, it won't let you (asks you to log in repeditivley), however if you just post without previewing it works fine, odd.

Billy T
24-01-2003, 03:47 PM
Well bugger me Terry :O

Nearly 35 amps? That's nearly 175 watts of heat to dissipate. And where do they hide all the copper needed to get that current into the chip and back out again to the PSU without dropping the voltage to zilch (or about 3 volts maybe?). The psu/MB plugs and sockets don't look up to that task, let alone the PC board tracks.

No wonder they need heatsinks, fans, refrigerants, cooling towers, liquid nitrogen, anti-radiation shields, carbon fibre armour, heat-seeking missiles and Uncle Tom Cobbly and all to keep working reliably.

I have new-found respect for chip designers who clearly deal with more than just mind-bogglingly complex circuit logic. They have massive thermo-dynamic issues to cope with as well.


Billy 8-{)

Terry Porritt
24-01-2003, 04:47 PM
Not quite Billy because that current is at 1.75v core voltage, thats up to 60 watts for the Duron, and 70 Watts for the Thunderbird but still a lot to dissipate.

What I dont quite understand is why the voltage regulators on the m/b dont glow red hot ?:|
They dont have huge heat sinks or fans.

Terry Porritt
24-01-2003, 04:57 PM
From memory the AMD XP current rating is a bit lower more like 28-30Amps, but you can see why the Athlon Thunderbirds run so hot, 40 amps +. Maybe a welder power supply is the answer!

I dont understand either how that sort of current flows through the m/b tracks and the flimsy cpu pins or even the sub micron? wiring technology inside the cpu.

24-01-2003, 08:39 PM
i don't think it has anything to do with the other rails as they are simply not used (after all the pc is off). the one rail (psu-on) that is giving problems is the only one they don't give a wattage rating for (or should that be milliwatts ;-) ).

Billy T
24-01-2003, 08:58 PM

Thinking out loud about "unused rails," I wonder if it is a sensing problem other than over-current. Could it be an over-voltage sensor that is folding back the output? Again, I can't see how but an unloaded output could perhaps peak a bit too high. This needs a process of elimination tweak'e, do you have any suitable dummy loads around?

Do you have any formal specs for the Enermax PSU?


Billy 8-{) ?:|

Billy T
24-01-2003, 09:08 PM
Hi again tweak'e

Found some enermax 300W psu specs.

Three things come to mind on reading these:

1) Inrush current

2) Impedance of your power feed due to faulty power point/loose pin connections etc (could be causing voltage drop at switch on surge.)

3) The 3.3 volt overvoltage protection cutsin at just 400mV above nominal. Any bounce or surge on an unloaded output could shut it down.

Voltage rating: 90V~135V AC or 180V~265V AC (selected by slide switch).
Iin=115/230Vac. 8A/5A. Frequency: 47Hz~63Hz.
Inrush current: 40A max./115V; 80A max./230V (at cold start).
O/P: 3.3V, 50mV, 4%, 0A / 28A; +5V, 50mV, 5%, 7.5A/30A; +12V, 120mV, 5%, 3.75A/15A ; -5V, 50mV, 10%, 0A/1A; -12V, 120mV, 10%, 0A/1A; +5Vsb, 50mV, 5%, 0A/2.2A. 3.3V & +5V, 170W.
TOTAL O/P: 300W. Hold up time: 17ms min. at 230V AC full load.
Over Power Protection: 105%~160% of max load.
Over Voltage Protection: +5V: 5.6V~6.9V; +3.3V: 3.7V~4.5V.
Efficiency: 70% min. at full load.

Can you try the psu in another computer?


Billy 8-{) ?:| ?:|