Overloaded PSU: what effect does it have

[farns]

Hi, Im a n00b with basic computer hardware knowledge (can build and generally troubleshoot a PC etc) and no electrical knowledge but I was wondering this:

If I was to overload (by a big gaming video card) a reasonably modern cheap OEM PSU the likes of those in HP computers a (a Bestec ATX-250-12Z H5r), will anything "bad" happen or will the overload protection just kick in and shut/crash the PC?

I have googled the PSU here as being of ok quality (albeit recapping recommended though), its not Seasonic made of course but I was just wondering what overloading such a PSU in normal use actually means.

Ive heard about Hyena and Deer and the old Bestecs etc but that was a while ago and this particular model is apparently an ok PSU according to you blokes and jonnyguru.

So am I taking a sizeable risk by not upgrading such a PSU or will it just shut down maybe with its capacitor life shortened somewhat.
Am I likely to be looking at A: shutdown only B: PSU damage/death or C: PSU death and collateral damage ?

This is just for situations for family and friends where I just want to throw in a low-mid range videocard and be done, and not have to take time and expense to replace PSU.

[Th3_uN1Qu3]

Not replacing the power supply when you install a power-hungry video card is usually a costly exercise. That Bestec will do fine with a midrange graphics card but i wouldn't count on it when it comes to other PSUs.

If the overload is gross and it just shuts down then you are lucky. However, to prevent false triggers, the overload protection is always set a bit higher than the maximum capabilities of the supply. In cheap power supplies, it's always set A LOT higher... The worst thing is a power supply operating on the edge, because efficiency drops, the thing runs hotter, the capacitors age faster, and there's more junk sent to your motherboard. I've had a friend that blew his (expensive) motherboard after ~1 year of running a gaming card on a cheapo PSU. When the PSU failed it took the board with it. And it was a Gigabyte with all solid caps... Also a few years back my father came home to find his whole computer burned due to a nasty power surge while he was away. The only thing that made it alive was the keyboard. Yes, even the mouse failed. Lucky for him, the system got replaced under warranty for a better one, and he had backups of the data on the hard drives. But not everybody is this lucky.

In both cases a good power supply would have spared the other parts. It's your choice.

[everell]

For your reference, here is a thread specifically for this model Bestec power supply.

http://www.badcaps.net/forum/showthread.php?t=9079

This model also has a crowbar protection circuit. It confuses some people. For some overload conditions, the psu simply shuts down. Remove the overload condition and it comes back to life. Under other overload conditions, the crowbar causes the psu to shut down, and removing the overload will NOT bring it back to life. To unlatch the crowbar, you must unplug the psu from the wall outlet, allow a couple of minutes for the electrons to drain off, then plug psu back into wall outlet and turn power supply back on. When crowbar happens, it is an overload design feature. It does NOT mean that the power supply has a problem.

Most Bestec ATX-250 12Z power supplies I have seen have a bad, bulging capacitor C36. Replace it or "move it". Also the glue used during the manufacturing process ages and goes conductive. Any dark brown or black colored glue needs to be removed. The thread referenced above talks about these two problems.

[c_hegge]

FWIW, I overloaded an atx-300-12z to 400w. It exploded after a couple of minutes. Sure enough, the primary switching transistor was visibly burned and had a hole in it after I dissassembled it. Now that probably wouldn't have killed the motherboard, but it's a problem nonetheless.

[westom]

Originally Posted by farns

So am I taking a sizeable risk by not upgrading such a PSU or will it just shut down maybe with its capacitor life shortened somewhat.
Am I likely to be looking at A: shutdown only B: PSU damage/death or C: PSU death and collateral damage ?

Numerous questions and some misinformation are posted. First, 350 watts provided by some better manufacturers are also called 500 watts by other supply manufacturers. Neither is lying. Both are measuring different numbers. Many computer assemblers would not know the difference.

Second, irrelevant is watts. Your concern is current for each DC voltage. Nobody can define a power supply as undersized without numbers. Obtain those numbers by buying or borrowing a multimeter. A tool so idiotly simple as to be sold even in Kmart or most stores that also sell hammers. Or $18 in Wal-mart. If you get a meter, then instructions can be provided so that your question (worry) is answered using a full one minute of labor. And no wild speculation.

Third, whereas a power supply does and must have an overvoltage crowbar, that crowbar does nothing for an undersized or overloaded power supply. If a supply is undersized, a signal to its power controller shuts down a computer. Discover if a supply is approaching overloaded by using a meter. One minute of labor. Then a next post says what you really have - without doubt or speculation.

Fourth, almost nothing that causes system failure leaves a visual indication. Many failures that can be identified even when the system is still booting, again, would be defined by numbers from the meter.

Fifth, if a power supply fails and destroyed the motherboard, then the computer assembler is the reason for failure. All power supplies (even long before the original IBM PC) contained circuits (ie overvoltage crowbar) that make motherboard damage impossible. But those who know only from observation will often repeat a popular urban myth. Or the supply was installed by a computer assembler who selected supplies base only on useless numbers - dollars and watts.

Sixth - answers to those three questions - A Shutdowns. That is determined by a power controller. Controller inputs (including an undersized supply) are why a controller might power off a computer. And why the meter would define that shutdown.

B PSU damage is mostly due to manufacturing defects. Even a grossly overloaded or shorted power supply is not damaged - as was true even 40 years ago. But many want to blame assumptions and myths rather than facts.

C PSU collateral damage and other damage. As noted previously, only a power supply installed by an electrically naive computer assembler would install a supply that could cause collateral damage. Overvoltage crowbar is one reason why motherboard is not and must not be damaged.

[kaboom]

To answer the OP:

Overloading a 'good' PS that only has overcurrent protection as opposed to overpower protection will...

Overheat the secondary rectifiers, where the current limit is imposed on the primary side. For example, something rated for 20A on +12 (240W) supplying more on that rail and less on the others will result in lower current on the primary side. So the 'current limit' never trips the supply, yet the +12 rectifier(s) overheat and short. Since the resulting short is before the output inductor, the high dI/dT fault is reflected back to the primary. Unless the supply has 'current mode' control (UC3842 as in the ATX-250-12Z), the primary switch(es) explode. Although that seems to make sense, it obviously wasn't the case with c_hegge's 'experiment.' Too much power, switches, sec rectifiers and output inductor overheat, shorts, more shorts and explosions! And since I2R heating applies to the inductor and transformer, doubling the current _quadruples_ the heating! It's worse in reality because of copper's positive tempco.

Originally Posted by westom

350 watts provided by some better manufacturers are also called 500 watts by other supply manufacturers. Neither is lying.

No, no, and no. This is the same bullshit as the 'Tandy watt' wars. If a power supply has a STICKER with "500W" on it and uses two FR5406s 'on a bracket' for a +12 output rated at 20A, two MBR10100s for a +5 output rated at 40A and one MBR10100 for a 15A 3.3V output, and FR4004s for both -5 and -12, then it's a LIE. Sure, 'the sticker says so,' but looking inside tells otherwise.

Originally Posted by westom

Second, irrelevant is watts. Your concern is current for each DC voltage. --buying or borrowing a multimeter. A tool so idiotly simple as to be sold even in Kmart or most stores that also sell hammers. Or $18 in Wal-mart.

Nice insult, there. The OP's not gonna be back after your suggestion that anyone 'dumb' enough to use a hammer can 'also use a meter.' So, just what is relevant? Out of phase VARs on the AC side perhaps? I rebuilt a power supply and wound a new transformer for it. And not that useless "just put a resistor on +5 to turn it into a bench supply" shit. I could pull well in excess of what the original transformer could take, thermally. Even though the losses were skyrocketing, the PWM would just increase the duty cycle, vainly maintaining the output voltage setpoint. Rated for 10A on +12, I could pull 25A from the original transformer, and my voltage would be dead on. Sure, the varnish would cook after half an hour, but that doesn't matter, since "my voltage hasn't dropped." What'll you tell 'em, once the transformer goes pri-to-sec? The transformer I wound has a bigger core and much heavier windings. 30A continuous, 60A intermittent, thermally limited.

You will never get work done without power, or wattage. If 'watts are irrelevant,' how come most POCOs have caps on the lines to cancel out the reactance of the lines and motor loads? Is it because they want to _reduce_ the current, so as to provide more POWER to the customers with FEWER losses on the lines? Oh, no. Watts, dissipated by I2R heating of dist/trans ckts, are irrelevant.

Originally Posted by westom

Third, whereas a power supply does and must have an overvoltage crowbar, that crowbar does nothing for an undersized or overloaded power supply. If a supply is undersized, a signal to its power controller shuts down a computer. Discover if a supply is approaching overloaded by using a meter.

I'll use an ATX-250-12Z as an example here. The 'power controller' you are referring to is the PG comparator. On low line voltage or low output voltage, the PG signal (to the motherboard) is deasserted. The computer then remains in reset 'forever.' Once again, you think that if the voltages are at setpoint, there can be _no_ overload.

Originally Posted by westom

Fourth, almost nothing that causes system failure leaves a visual indication.

Conductive glue across isolation barriers
Bad solder
BAD CAPS
Holes in semis
"Special" Infineon MOSFETs
Burned output inductors
Missing components in EMI/RFI filters
No PI filters on the outputs
Smoke/burning smells just before 'the earth shattering kaboom'

A cheap DMM is not going to see the ripple caused by bad caps in a PS. And it certainly won't tell you if you have 60/120 Hz feedthrough from the primary caps being too small/dried up.

Originally Posted by westom

Fifth, if a power supply fails and destroyed the motherboard, then the computer assembler is the reason for failure. All power supplies (even long before the original IBM PC) contained circuits (ie overvoltage crowbar) that make motherboard damage impossible. useless numbers - dollars and watts.

Did the 'boys on USENET' kick you out again because of the insults, westom? "Fails and destroys the motherboard" comes in several flavors.

5VSB runs away.
Line voltage into motherboard/computer from glue across transformer/ISO barriers or PRI/SEC short in TX.
Switchers short and, because of bad caps in the PS unable to filter, trash gets into system.
No filtering because of fooyooh caps lets ripple into system, which then intermods with switching freq of CPU buck converter (VRM). The resulting beat note wreaks havok on everything else, even being audible in the fan and HDD motors.

Your "all power supplies" argument falls apart today, as there are too many supplies 'cheaped out.'

I hope you realize that 5VSB running away is a special case, as some power supplies (including ATX-250-12Es) only had 'turns ratio derived' feedback for the AUX supply. The cap fed from the feedback winding rectifier would short and/or get high ESR, causing the circuit to 'kick harder' to bring the feedback point back up. But in this case, the feedback point is effectively bypassed by the leakage of the bad 'critical cap,' so the 'harder kicking' results in the 5VSB going up to as much as 20V, killing the motherboard.

Originally Posted by westom

meter would define that shutdown.

Don't need a meter. The system is either on or off.

I've seen conventional CTs monitoring the primary current in these things. As well as a sort of 'zero-sequence' (not truly) arrangement where the driver transformer has a diode feeding a cap that provides a voltage proportional to the primary current. UC3842 and their variants use 'current mode control.'

Let's say that one of those is designed to shut the supply down when a current over 3A on the primary side is reached for 100ms. 3A*150V = 450VA. To make it easier, let's ignore the losses in the transformer, switches, rectifiers and output inductor.

Our limit is 450VA. The outputs are +12 @ 20A, +5 @ 30A. Even though there's a little headroom, if the output inductor is sized properly, that limit will shut the supply down if there's a short on one of the outputs, or if the reflected primary current of both outputs combined exceeds the limit. But +12 @ 37A will not trigger the primary current limit, at least not until that rectifier shorts. 12*37=444W, reflected back to the primary, is 2.96A. In other words, overheated and eventually shorted +12 rectifier, but no primary overcurrent trip til then.

Originally Posted by westom

B PSU damage is mostly due to manufacturing defects. Blame assumptions and myths rather than facts.

Then why don't you short one that does not have primary OC protection and tell us what happens? I hate to spoil it for ya- the switchers explode, their base resistors burn up and the 'speed up' caps go leaky from the current going through the base resistors. 40 years ago, when switchers were a marvel, everything was protected. Those IBM supplies in the XTs even had OC trips on the -5/-12 rails. I had a tantalum cap short on the mobo and the supply just shut down.

Look inside a worst buy special and there is no myth whatsoever. Just like your ASSumption that proper grounding and BONDING are 'myths' rather than steps in the right direction for lightning protection. I've seen the flamewars on sci.electronics.repair, and am really tired of you 'dictating' that the experience of others are only "ASSumptions and myths." But I will agree with you that the common mode impedance of a typical branch ckt is too high for a typical surge 'suppressor' to work. Any SPD belongs at the service, and anything in or out must be kept at the same potential, bonded. Preferably, the SPD should be kept outside, with heavy copper strap to multiple rods, with bonding jumpers to CATV, telco arrestor and service noodle.

Originally Posted by westom

C PSU collateral damage and other damage. As noted previously, only a power supply installed by an electrically naive computer assembler would install a supply that could cause collateral damage. Overvoltage crowbar is one reason why motherboard is not and must not be damaged.

An overvoltage crowbar is an SCR triggered by a zener, as used in the old linear supplies to protect the load when/if a pass transistor shorted. There is no crowbar in the 'puter supplies, only a lockout circuit.

In the power supply I rebuilt, there's a pair of NPN/PNP transistors that, when the voltage related to the primary current is too high, conduct. They are cross connected, so once triggered, stay conducting. This pulls the 'duty cycle control' pin of the '494 down, causing the output practically go away. The inverter is still running. The supply to the '494 is taken before the output inductor. So while the pulses are too short to produce any voltage _after_ the output inductor, the '494 stays powered and in control. The only way to 'reset' it is to cycle the power. This could even be implemented with an SCR instead of the NPN/PNP pair, but neither the function nor use of a particular part makes it a crowbar.

Go back to USENET and start a flamewar with Phil Allison. I'll get the popcorn ready.

IB4TL

[everell]

On Bestec ATX-250 12Z 5vsb circuit, zener ZD3, optoisolator PC3, and scr xl1225 going to pwm chip pin 4 look like a crowbar circuit to me. So what you think Kaboom?

[kaboom]

That'll certainly lock 5VSB out if it runs away. And because of the SCR, it'll require a power cycle to 'reset.'

But when I think of a crowbar, this old brute force idea comes to mind. Look how it's done in the Astron RS35.

If a pass transistor shorts, raw unregulated 25-29V passes right through, so 'short the output' was a cheap way to save the load.

But if the fuse is too big or bypassed by some idiot, the SCR burns up and the load is no longer protected...

-Paul

[farns]

Thanks for the great answers all, Im just an amateur PC user who got on the overclocking and DIY thing to save money, shortly after buying my first PC back in 2001.
I google buying decisions very thoroughly so Ive seen you chaps before on occasion

Point well taken in the first post (and the rest) about the highly-set overload protection and dangers of running it at very high capacity, especially as the thread linked above says it doesnt have great capacitors.

I wouldnt put a top end videocard and quad-core etc with this sort of PSU, but Im running a stock speed E5200 and looking to add a mid-range videocard like a 5750 or 9800GT or preferably something lower end that doesnt require a PCIE connector.

The 5750/9800GT/8800GT use a similar amount of power and barely require a single PCIE pwr connector, a few redesigned cards even run without one.

Anyway judging by a cheap KIll-A-Watt type electrical meter at the wall, the E5200 + 5750 uses only 138w load running Crysis demo which is extremely GPU intensive but doesnt load the CPU much. I would guess prime95 + Crysis might add another 20-40w.

The "250w" PSU only has a rating of 14a on +12v rail so does that mean I was very close to capacity?

Obviously to be safe I should just spend the little money on a better PSU but I was just curious as to roughly how much power I actually required.

[everell]

I just took a look at a Gateway computer for a friend. It was dead. My post tester indicated that the cpu was not doing anything, which would ordinarily lead you to believe that the cpu is bad. The power supply was a Newton NPS-160CB rated for 160 watts. So I pulled it and replaced it with a Bestec ATX-250. Everything came up working fine. A marginal power supply can cause odd problems, or even cause the cpu to quit processing. It sometimes is surprising what that old Bestec can do.

The Newton psu had 330 uF/200 volt input capacitors, ONE switching transistor with TO-220 case, and all output rectifiers were TO-220 case. The Bestec has 470 uF input capacitors, and larger silicone. Also much larger switching transformer.

[Th3_uN1Qu3]

Originally Posted by westom

Im running a stock speed E5200 and looking to add a mid-range videocard like a 5750 or 9800GT or preferably something lower end that doesnt require a PCIE connector.

That kind of computer will do fine on a quality 250 watt supply. All bets are off when we're talking about generic units.

[westom]

Originally Posted by everell

I just took a look at a Gateway computer for a friend. It was dead. My post tester indicated that the cpu was not doing anything, which would ordinarily lead you to believe that the cpu is bad. The power supply was a Newton NPS-160CB rated for 160 watts. So I pulled it and replaced it with a Bestec ATX-250. Everything came up working .

True if assumptions are not based in how a computer works. A power controller must 'approve' all inputs before it permits a CPU to even execute the first machine code instruction (BIOS). Your 'swap' implies a power controller did not get a good signal from the original supply. Got a good signal from the new supply. Then let the CPU execute.

From information provided, nobody can say if that supply was undersized or (what happens more often) suffered from a manufacturing defect. The list of failures that means visual indications are so few (maybe 100) as to be near zero.

Nothing on the load must ever damage a supply. Otherwise the manufacturing defect is due to bad engineering or part failure. Required of all supplies: short all outputs together and the power supply must never suffer damage. Intel even says how thick that wire must be.

Considering how many parts must work without failure, rather amazing how reliable all power supplies (and a computer) really are. Meanwhile, parts selected by cost controls or counterfeited are two examples of why power supplies can fail. Assume the power controller did not get a 'good' input from the supply. Therefore would not let even a CPU execute.

[everell]

Whereas the old 160 watt power supply did allow the computer to operate for a number of years, it is obvious that it does indeed have a problem, most likely a malady of old age such as dementia or alsheimers - being as it can't remember how to make the cpu work. The 250 watt Bestec, on the other hand, was still young and verile enough to get the job done. As people....and power supplies.....age, a certain amount of power is forfeited. But why would you want to spend the time and money to fix that old Newton 160 watt power supply if you have a good ole Bestec ATX-250 handy? Furthermore, a simple swapout of power supply solved the immediate problem. In short, it worked!!!

[farns]

Thanks for the help, I will use a 4670 video card with the E5200, total power draw I can expect apparently is about 140w-150w with CPU and GPU at 100% capacity (gaming more likely to be about 120w-130w).
There are also 2 x 7200rpm hard drives but if I am correct they draw from the +5v rail, so there is plenty of capacity there.

Ive read that the +12v rail supplies the CPU and GPU and the Bestec 250w only states supports for up to 14A (168w), so hopefully I am not too close to capacity.
The computer will probably be replaced in a year so hopefully capacitor aging or other PSU damage will not be a factor before then.

[jumvah]

I started having intermittent failures on my puter, checked everything i could in another machine (ram, CPU, HDD, gfx)...all fine. put back in my machine, fails again.
Bought new PSU, mobo, ram, CPU and started over. same problem.
Grrr!
Only common component in both PCs is ps2 mouse.
Is it possible for ps2 mouse to cause crowbar in the PUSs?
Gonna go buy a USB wireless kb/mouse combo and try again later.
If that don't work I'm gonna start hitting things with a hammer.

[westom]

Originally Posted by jumvah

I started having intermittent failures on my puter, ...

Undefined is even what fails. No numbers are provided. Does a program crash? Does the computer lock up? Does it simply power off? What do system (event) logs report? What do the manufacturer's comprehensive diagnostics report? What exactly is every error message? Your help is only as useful as facts that you first provide. Currently very few hard facts were posted.

Never fix something by just replacing parts. Also called shotgunning. First identify a fault using information such as requested above. Once the defective part is identified, the problem gets fixed by replacing the only remaining suspect. No more spending money to replace perfectly good parts.

An example. Does the computer just power off? The power system is more than just a power supply. Another power system component determines if and when a computer can power on. Decides when the CPU can execute. And determines if and when a computer can power off.