Let's play "what part is failing?"
 

This should be simple...

Athlon 1300
256MB RAM
60GB WD HD
not sure on make and model of the case or power supply (local computer company built it for me a couple of years ago)

Lately, the system's seemed to make a bit more noise on startup, as if something was running sluggishly. It'd go away once the system was running for a minute or so.

Tonight, I plug everything in after a weekend absence, hit Power, and the system is on... on... on... shut off after about three seconds, with a blinking light under the light-bulb icon and no other noise or movement. I shut it all the way down, waited a minute, tried it again... same behavior.

On about the eighth attempt, it did finally come on and stay on, which is how I'm typing this now.

It did this once before, a few months ago, but it only "misfired" once.

I would really like to fix this before I _can't_ get it to come back on. First impulse is the power supply -- am I likely wrong?

(Ironically, I have a second fixer-upper right now; I hijacked my niece's malfunctioning Dell 450 with the intentions of wiping the HD and installing a fresh coat of 98SE. I would really love to have at least one functioning computer in the house, though.)

power supply would be first
 

Heya!

I'd look at the PS first, and then your motherboard second here.

M

Do you have a floppy drive? Try booting from that. It's possible you have a frozen hard drive -- though I haven't seen one of those in years.

Here's a thought...my computer doesn't want to boot when it's cold. If I warm up the box first, it boots fine. I have no clue what causes it though...maybe a cold solder joint somewhere.

Was your house cold when you tried to turn it back on?

Hmmm. My instinct is, it's not the power supply, or it wouldn't be powering up at all.

It sounds like your motherboard started up, found something it didn't like, and then shut down--something like a CPU fan not working? (It can happen for high temperature as well, but 3 seconds is hardly enough to overheat.)

The apartment was at a normal temperature, around seventy.

If it was an overheating problem, I would think that it wouldn't stay on for any length of time at all. Once this gets over the hump, so to speak, it'll stay on for days. I'd just leave it on and plugged in, but it's thunderstorm season and I'd rather not tempt fate even a little.

If it was a car, I'd get the starter replaced, but computers don't have those (that I know of).

My next step in the meantime is to get some compressed air and clean the hell out of the innards, particularly the fans.

I can't think of anything Mobo wise that'd make it shut down, usually if the mobo finds something it doesn't like it'll scream at you (series of beeps) that corrospond to which component it's have a bitch over. Full-on shutting down suggests to me a safety tripping in to PS, the fact there was no noises (such as the CPU fan whiring up) increases the chances of that. Unless the heatsink has completely fallen off or something it can't be overheating, if it was it wouldn't stay on anyway.

Most powersupplies are utter pieces of overrated shit made for around $3-4USD (trust me, I worked for an importer), my money is very much on that.

The one component that can cause anything inside computer to create the crash is the PSU. PSU is the foundation upon which everything else depends. However only a fool would wildly replace anything only on a whim.

You need two tools - a screwdriver and and a digital multimeter. The second is sold everywhere from Sears, to Home Depot, to Radio Shack. Even though computer fails only intermittently, the bad power supply symptoms exist constantly. You must measure those DC voltages. Those voltages must measure within the upper three-quarters of limits in this chart Once power supply is confirmed by meter, then move on to other suspects.

Could be an interrupt conflict.

I had one of those once. The video card and something else were fighting over an interrupt and the computer would instantly shut off.

I assume you have tried alternate bootups. Someone above suggested with a floppy - do try that.

Also, try booting up while holding the F4 key. That will dump you into the BIOS config area before the OS cranks up where you can putz with the settings.

Seems that its happening too fast to be an interrupt thing but you never know. Also, consider not turning it off. I never turn mine off.

I also wonder if you have a bad power switch.

You may have something there.

When it quits, the boot cycle doesn't last even remotely long enough for any of that to kick in. It's on -- it's off, maybe two seconds apart.

I live in an apartment complex, and do not trust that the building is properly grounded and insulated. I have a surge protector on my system, but do not take it on faith that that would be a 100% effective defense against power surges and lightning. I've lost a couple of modems in my lifetime to such things already, and have no interest in frying anything else; hence, if thunderstorms are likely and I'm not going to be right there to monitor things, it's getting unplugged.

Possible, but I'm wondering how I test that. I will note that the power switch behaves perfectly when I do the hold-for-five-seconds-to-kill-the-power thing.

Power switches on modern motherboards are soft, so just unplug the jumpers and touch across the two with a screwdriver (with a wooden or plastic handle).

Really does sound like a PSU issue to me.

In terms of surge protection, I usually use APC gear, kicks ass.

Very worst case: the noise is your CPU fan, and it's making noise because the heatsink is uncoupling from the CPU. When the sink is in place the system boots and acts normally. When it's slightly off, the CPU heats up so rapidly that it reaches 80 degrees C in 2 seconds, at which point the mainboard CPU temperature sensor tells the BIOS to shut the whole system down to preserve it. When the sink eventually fails with the system running, it causes a fire which spreads to the papers you've stacked behind the system, and your whole place burns down.

It's probably not the problem, but it *could* be!

Thank you, Tony. Now I will NEVER SLEEP AGAIN anywhere where there's a computer in the house. Yeeeeeegh!

I hade the same problem once , it was the power supply .

I had a friend that called once asking if i had a spare power supply , his had gone up in flames !!!!!!!!

So go buy a new one , if you have a 250watt get a 300-350watt , the price diff is SMALL !!!!!

There is a reason for that. It's the same thing. The only decent PSU out there are from brands like Antec and TopPower.

$23 Sparkle 300W from newegg. They have a good reputation if you want to go "cheap".

A terrible reputation because even specs in that URL say so. Two most critical and essential functions in a power supply - even found 30 years ago - overvoltage and overpower protection. They provide almost no numerical specs and still the power supply is obviously crap.

How do you think they sell a power supply for so far less than $60 retail? They simply forget to include essential functions. Goes right back to a previous thread where I discussed the technical ignorance by so many computer people. The Sparkle supply is dumped into N America at higher profits because so many American technical experts don't even look at specs let alone understand them. It says right there - They could not have been more blunt. The supply is missing essential functions necesary even to keep power supply from destroying motherboard, disk drive, memory, etc.

Look at the long list of speculated reasons given. I can probably number the number. And yet most every suggestion would be made redundant in minutes with the 3.5 digit multimeter. You have no numerical facts, some symptoms, and a load of speculation. Get the meter now and don't even waste time. It is that important, even to say if the power switch is sticky or intermittent. You can guess all we want, but without something to see electricity (with numbers), then we all will spend long hours playing witch doctors.

Get the meter. There are three subsystems just in the power supply - the PSU, motherboard control circuit, and power switch. Get the meter so that power supply can be eliminated as defective OR isolate the problem to one of three possible problems. All done in minutes.

BTW, you will also learn how the computer really works.

Not mentioned is that if the CPU was an Intel, then overheating only slows processor down; no damage. This function has been in all Intel CPUs as far back as the 486DX2-66 ; and maybe even longer.

However if an Athlon runs with heatsink disconnected, then Athlon will be toast within less than 5 seconds and usually also damages motherboard:
Hot Spot - How Modern Processors Cope With Heat Emergencies

Onboard temperature sensor only protects Athlons if heatsink is attached enough that CPU does not overheat quickly.

But this is irrelevant to your problem. In fact, running a system in a 100 degree F room is how one debugs a defective computer. 100 degree F is normal temperature for any properly functioning PC.

FWIW I have tested a Sparkle with a multimeter and the voltages were sound, but if you want to pay more you may. Of course a replacement motherboard is $75 these days so an additional $50 for a top class PS is up to you.

I don't keep a 100 degree room personally but there are those who might.

my own personal PSU

The blue LEDs are helpful cos they tell you the power supply is on when it's dark and your system is under the desk. But it's $84.

Only MBA trained management believes one can test for quality. Every defectively designed power supply will provide correct voltages during their lifetime. Those voltage measurements say nothing about supply quality. But then readers of W E Deming's book "Out of Crisis" (last time I looked, it was in its 26th printing) would recognize that voltage measurements are classic symptoms of an organization that had no quality. The point here being that every reader should have immediately recognized that voltage measurements report nothing useful. If it was not that obvious, then concepts of quality have not yet been learned. Concepts of quality are not always so obvious until after they have been learned.

The Sparkle power supply is dumped in America because so many so called computer experts recommend technology they don't even understand.

BTW, that voltage test is the exact same reasoning that sent 7 Challenger astronauts to death. "It worked just fine yesterday - therefore it will launch just fine today." IOW "engineers are only dumb idiots. Thank goodnes we managers are so much smarter."

The Sparkle tech numbers say it is crap. The O' rings on a similar and (one year) previous shuttle launch said quite loudly that Challenger should not be launched. Like the Sparkle recommendation, the shuttle manager 'expert' need not first read the numbers or understand technology principles. He just knows like so many clone computer assemblers who buy crap power supplies on one spec - price.

VSP - I am suggesting where your problem may have been created. However we first need those numbers from the multimeter.

Notice the price and specs provided by UT's "my own personal PSU". It claims to include essential functions - therefore costs maybe double.

http://cellar.org/2004/asusprobe.gif

My Asus mainboard came with a utility that records the fan speeds and temperatures and voltages as measured by the mainboard. I have never seen these voltages waver one bit even under the maximum load I could manage to put on the system.

Only after calibrating that voltage monitor with a 3.5 digit mulitmeter will that motherboard monitor report anything useful. They are selling motherboards on the cheap. 5% resistors are far less expensive than 0.5% and 1 %. Did you notice that little descrepancy in the specs? Or did they just 'forget' to mention that critical information?

Now let's return to fundamentals of quality. There are good components, bad components, and unknown components. Three categories. Unknown components and good components both measure good voltages. Then a sudden failure with massive damage happens. Power supply failure also takes out disk drive, Ram, motherboard, CD Rom, etc. Where was a voltage measurement going to predict the crappy power supply from the unknown category?

Stable voltages only suggest a good supply. Bad voltages say, beyond a doubt, it was a crappy supply. No voltage measurement is going to detect a missing functions in that defective Sparkle supply.

3.5 digit multimeter can identify a failure. It cannot detect all design defects. Many conditions are necesary to define a minimally acceptable power supply. That Sparkle supply is the classic example of defective - even though it always measures good voltage.

Bad voltage measurements will explain why VSP's computer is failing. But a quality supply - the first of three possibilities - demands that the supply meet a long list of critieria. A stable voltage tells us nothing about a good supply. It only says the supply is not bad - today. That Sparkle quickly violates criteria for acceptable - even though its voltage measure OK today.

At this point I am appaulled I should even have to explain simple, fundamental technical concepts. PSU measures good voltage? Therefore it will never damage any computer? Where does this religion come from?

Once VSP has a meter, only then will we start a 'step by step' process to fix his machine. In the meantime, it would be nice to have specifications for his power supply - to avoid other future problems. Today, we need a few minutes of voltage measurements to only begin an analysis. Start by eliminating PSU as reason for failure - before even speculating on anything else.

All I know is that my motherboard costs more than my multimeter.

My 3 1/2 digit Fluke was $650.;)

Question from a computer user, not a computer builder:

Can you pull a power supply from a cheap older name brand computer and have it be more reliable? Seems to me a power supply is a pretty basic component, and if they made them better in the olden days, and you can get old, obsolete computers for less than $50, you can just pull a good power supply out of an old PC.

answer to a computer user
 

Not a good idea because the ratings (wattage) will likely not be sufficient to cover the newer computers' power requirements.

I used to have old computers that would work if I reassembled them, and the PS would have something like a rating of 200w.

My computer now needs more like 300w and IT is an older model as well...800Mhz Celeron, one 10g HD a CDROM and a CDRW.

My next one will have all this and more and thus will have a higher power requirement.

Brian

Besides what Brian said, my experience in fixing "store-bought" computers is that they often skimp on the power supply. I've seen relatively new systems with 70 watt ones!!

An older power suply is also likely to have accumulated a lot of dust, which usually cruddies up the fan bearings eventually, and make it noisy and/or fail.

Again lets return to the 'numbers game' lies from clone power supply vendors. Many brand name computers now use 200 and 250 watt power supplies. If those supplies were rated using the deceptive technique of clone vendors, then those same 200+ watt supplies would be advertised as 300 and 375 watt supplies. What I have found: brand name computer supplies actually output the power as claimed. Clone supplies routinely do not output that power AND are also missing essential functions.

These problems were also demonstrated in tests in Tom's Hardware at
http://www6.tomshardware.com/howto/0...021/index.html Even in published tests, clone power supplies failed to provide full power. They also did something that no power supply must ever do. Some self destructed before full power could be obtained because it was defective by design. Clone products intentionally manufactured to sell at a low price to 'computer experts' who never even read numerical specifications. Wattage number provided only to deceive.

Another spec missing from that crappy supply. It has no overpower protection. At full power, the supply may instead self destruct - destroying with it the disk drive, its data, the motherboard, etc. Catastrophic failure directly traceable to the human who failed to learn basic and simple computer concepts. Failure that even Tom's Hardware demonstrates way too common in clone power supplies.

Power supply in the original post may be a classic example. But we cannot know until the 3.5 digit multimeter has provided necessary numbers.

Sure of that? I've worked with the kind of supplies you're talking about, utter lies on wattage, cheap pieces of shit and seen literally hundreds of the fuckers blow but I've never seen one take a computer with it.

Jaguar, I have
 

Jag,

I've seen it happen with servers. It's especially prevalent with the cheaper power supplies. My friend who owns a web hosting company (burst.net) has literally thousands of servers in place, so we have a representative sample :).

We have seen PS's from multiple vendors completely blow out motherboards, or fry certain components. It's more common than we'd like to think.

TW is right. If you have machines where the PS runs in a bad way, it will cause major issues.

Mitch

hmm ok, I'll take your word on it, maybe these ones had some kind of protection after all. I couldn't get burst.net to load, maybe his PSU has blown.

that's www.burst.net :)
 

And I just loaded the site fine. No, his PSU didn't blow out there. He uses good ones since he hates paying for replacements.