Min Allowable Voltages?

[bubba]

never mind, looked at the chart again, read it wrong LOL thought the 11.5 reading was from the multimeter. It was the software.

sorry.

[westom]

Well... I don't think it's power any more. ... While surfing the internet the computer shut itself off, didn't restart, just shut off.

OK. First, all measurements should have been on the Molex connector since making measurements from all those other locations were irrelevant (as your numbers demonstrate). Second, you did not include a very important numbers from the purple wire as requested.

Third, whereas the things you did for loading were not sufficient loading, still, the error margin is so large as to say those voltages are all OK. And fourth, the AIDA which reported completely defective voltages on the 3.3 and 12 volts is (according to your multimeter) seriously out of calibration.

The machine powered off. Memory has nothing to do with that symptom. Discussed earlier was the power controller. Two important sentences should have gotten your entire attention - especially after the power off. "A purple wire powers the power controller ... A power controller must have its own separate power supply - the purple wire."

But again, you still did not provide numbers from the purple wire which is quite relevant to your problem.

Imagine the power controller as a black box. That black box has inputs. If the inputs are in error, the black box may suddenly just power off the power supply. A "power system" has many components. Why did your power controller power off the power supply? A question that demonstrates "follow the evidence". Your problem involves the power ‘system’ (a power supply is only one component of that ‘system’).

a) Inputs to a power controller include a signal from the supply about voltages, the purple wire from a separate power supply inside that power supply box, and the power switch - to list the most significant.

b) All relevant voltages (on any orange, red, and yellow wires) are stable. Therefore, (again using the meter), measure the voltage on the gray wire (also on the Molex connector). That voltage should always be well above 2.4 volts because of numbers from the orange, red, and yellow wires. Gray wire well above 2.4 so that noise does not tell the power controller to power off.

c) And the power switch. Find where the switch connects to the motherboard or find access to that power switch. When the switch is not pressed, those two wires to the switch must also measure well above 2.4 volts.

Three major inputs to the power controller. Discover if any one is telling the power controller to power off. The power controller is powering off your machine. What input is telling it to do so?

List of suspects is seriously diminished now that other voltage numbers were provided. Measurements on the purple wire, gray wire, and measure between the two wires that connect to the switch. Search for a signal or noise that is asking the power controller to power off suddenly.

[Major_A]

After checking the PSU and it came back "ok" I checked the RAM again. I've run Memtest in the past error free for 3-4 hours then it starts showing errors. The insatiability of this computer similar to other machines I've encountered that have RAM issues.
A) I have the PC with me so I'll check it out today.
B) See A.
C) Why is the power switch power relevant to the computer being unstable?

*EDIT*
I haven't checked the switch yet, I will, here are the other numbers.

PSUReadouts.jpg (41.46 KiB) Viewed 3975 times

[skier]

Actually, looking back at your specs, I have a CoolerMaster eXtremePower 550W and I don't trust it in any system (was afraid to even run a Core2 off it)

some states let you take out PIII KillAWatt meters from public libraries for energy efficiency programs (Maine does at least). plug the computer power cable into the meter and let us know what it pulls for wattage at the wall (homedepot/lowes also carry them for $29.99.. They are very handy devices, I keep my rig plugged into one 24/7)

just a good thing to check, aside from that I notice your hard drive was running at 46C which is pretty toasty. It could be the hard drive failing so it makes windows shut off/reboot (Crystaldiskinfo ) I also am very fond of having CDI on all my systems (plus you can check out how many hours it has been running and # boots for curiosity, the Health Status should be Blue and say "Good") My hard drives run in the upper 20's low 30s when they're not defragging or virus scanning etc (barracuda OS 250GB 7200.10 @27C, Caviar Black 1TB @30C)

[westom]

After checking the PSU and it came back "ok" I checked the RAM again. I've run Memtest in the past error free for 3-4 hours then it starts showing errors.

There is no relationship between memory and what turns off power. If memory is defective, then it creates a task error - the task is shutdown by the OS. Or it creates an execution error inside the OS resulting in a BSOD. Neither result in power off. Both types of memory failures are reported with numbers that identify why that failure happens (ie addressing problems, internal capacitor leakages, etc).

To power off a computer, the memory must force the CPU to execute a long, specific list of codes that write to the power supply controller. A specific command to power off. That does not happen when memory fails.

The hard fact said power just cut off for no reason. Some inputs to the power controller must cause power off to occur. Unfortunately, we are running out of the inputs you can measure.

Gray wire voltage is well above noise levels. Purple wire has a sufficient voltage margin. So some other input is forcing the power controller (which is a computer different from the CPU) to power off the power supply.

Meanwhile, let's move on to how memory diagnostics are executed. No need to run hours. Intermittent memory failures are due to insufficient margins (threshold, timing, etc). So supply what changes those margins. Normal computer operating temperatures include a room at 100 degrees F. All semiconductors work at temperatures well above that without any damage. But at 100 degrees, electronics with insufficient design margins fail hard. This is how reliable electronics are identified even decades before the PC existed.

Heat selective parts with a hairdryer. Normal temperature for every IC is uncomfortable to touch but does not leave skin. Unfortunately a hairdryer will not get that hot. But heat memory with a hairdryer on its highest heat setting. Execute Memtst86. If memory is 100% defective, then its insufficient margins will temporary change just enough that Memtst86 sees hard failures. Record the numbers to learn e what is really and has long been defective.

If heated memory still passes, then move on to the large IC that talks to memory. For example, many PC traces from memory will go short to that large IC. Heat that northbridge and execute Memtst86 again.

An error report does not specifically define that chip as defective. But can identify an area or suspects that are intermittent - completely defective. Semiconductors that work when cold are susceptible to noise induced failures when warmer.

Heat is a diagnostic tool. Most who never learn how hardware works will instead cure symptoms. More fans. A completely defective semiconductor may work for hours at 70. And can be located in seconds using heat.

That switch is usually difficult to access. May be easier to find where it connects to the motherboard. Then use paper clips or sewing needles to touch those connector contacts. Hand dexterity or imaginitive solutions may be necessary. A voltage will above 2.4 must be measured when the switch is open. And that voltage will drop to zero when the switch is pressed (so an assistant to provide a third hand may be necessary).

[skier]

*snip

I would still see how much AC wattage you are pulling, because often voltage can be fine but if you are pulling too much wattage it won't work right, period. my OCZ 600w unit can pull 750w at the wall and shuts off at 800+ but really doesn't like it, with the same hardware and load with my Corsair unit it only pulls 450w at the wall. just an efficiency thing that can vary unit to unit, brand to brand. and again I would never use a Coolermaster unit again unless I read a sparkling review from here. it was inexpensive for a reason.

[westom]

some states let you take out PIII KillAWatt meters from public libraries for energy efficiency programs (Maine does at least). plug the computer power cable into the meter and let us know what it pulls for wattage at the wall (homedepot/lowes also carry them for $29.99.. They are very handy devices, I keep my rig plugged into one 24/7)

Kill-a-Watt to view things such as power factor provide the OP with no useful facts. All computers must startup and work normally even when incandescent bulbs dim to 40% intensity. Is your voltage dropping that much? Then fix a potential human safety problem inside the walls. Meanwhile, those DC voltages on the orange, red, purple, and yellow wires do not vary even when bulbs are that dim. That is what electronic power supplies do. Make AC voltage variations irrelevant to electronics.

Most A+ Certified computer techs never learn electrical concepts or learn from numbers. Therefore this may be the first time a lurker is hearing these concepts. Even Intel's original ATX specifications said PCs must work when incandescent bulbs are that dim. A voltage that is harmful to motorized appliances such as your refrigerator. But perfectly ideal and sufficient for PCs. Light bulbs will report more useful information than the Kill-A-Watt. OPs multimeter can also report same relevant facts when he was measuring any one orange, red, purple, and yellow wires.

Disk drive does absolutely nothing to cause a computer power off. Please learn power controller functions as described in previous posts. Completely obvious. Disk drives (and so many other components) are not on a list of suspects.

I appreciate you making the suggestion. Is used here to demonstrate why so many computer assemblers' will only swap good parts until something works. Demonstrated are simple computer concepts that anyone should first learn. Above are the simple things. Demonstrated above was how to 'follow the evidence'. Everyone has heard that repeatedly in CSI. Most have no idea what the expression means. If you were 'following the evidence', then a disk drive was never mentioned. A concept only learned from making mistakes - as you have demonstrated. And why your post is useful to others following this thread.

Also learn why a multimeter or light bulb does better what the Kill-A-Watt might report. A dimming light bulb should never cause a computer failure. But might report a human safety problem. No light bulb should change intensity even when major appliances power cycle.

[westom]

I would still see how much AC wattage you are pulling, because often voltage can be fine but if you are pulling too much wattage it won't work right, period.

Wow. That post even contradicts what is taught in high school physics.

If something draws excessive power, then voltage drops significantly. And incandescent bulbs dim.

No computer draws 800 watts. But when selling to people without basic electrical knowledge, then 800 watt supplies that are also missing other critical functions can be dumped into the market. Most every PC draws about 200 watts. Its power draw may even peak at 350. A 20 amp power circuit means even 2400 watts are easily supplied. And no incandescent bulb changes intensity.

Skier demonstrates speculation, hearsay, junk science, and insufficient education that forces so many American companies to go to China for better educated techs.

Learning how to 'follow the evidence' can be rather difficult for some. By following facts (ie. incandescent bulb intensity or numbers from the orange, red, etc wires), obviously, AC wattage and voltage are not relevant. Obviously, excessive power consumption would cause a voltage decrease and dimming light bulbs. Obviously, a computer will not draw excessive watts. Unfortunately urban myths are alive and well.

[XstollieX]

No computer draws 800 watts. But when selling to people without basic electrical knowledge, then 800 watt supplies that are also missing other critical functions can be dumped into the market. Most every PC draws about 200 watts. Its power draw may even peak at 350. A 20 amp power circuit means even 2400 watts are easily supplied.

So are you saying that no computer will peak more than 350 at the wall?

[westom]

So are you saying that no computer will peak more than 350 at the wall?

Not exactly. But read spec numbers for computer designed by engineers. They don't waste money on 800 watt power supplies because a PC does not need that much power.

Meanwhile, the topic is not power supplies or hearsay. The topic is about Major_A's failure. Even if a computer draws 800 watts, the receptacle will provide that power - and more. Numbers provided by Major_A also say AC mains do not explain his failure. Just a few reasons why wild speculations about insufficient power has no place in this discussion.

Major_A has suffered sudden power off. Power cuts off when a power controller orders it. Why would a power controller suddenly power off a computer? Follow the evidence. View inputs that affect that power controller. Numbers provided by Major_A eliminated a long list of suspect - including AC voltages.

Major_A, BTW, what is the manufacturer of this machine? If you said, I have long forgotten.

[Major_A]

Manufacturer... me. Pieced together from old parts I had lying around.

[skier]

snip

why do you keep referencing incandescent light bulbs? I'm dying to know

[westom]

why do you keep referencing incandescent light bulbs?

Incandescent bulb reports far more useful facts than anything you have discussed. Anything a Kill-A-Watt might report was already made irrelevant by viewing an incandescent bulb. Simplest tools (ie a hairdryer) are powerful diagnostic tools. But only if one learns underlying facts, the science, and numbers.

All PCs must work fine even when a light bulb dims to 50% intensity. Even the original IBM PC. But that was already explained.

Light bulbs have been ubiquitous for as long as you have existed. Still do not know a light bulb is a powerful diagnostic tool? Constructive questions means answers with numbers. A computer must startup even when a light bulb dims to less than 50% intensity.

A light bulb suggests AC power for any properly constructed computer is fine.

[Major_A]

The power switch isn't going to happen. I need a set of alligator clips or narrower diodes on the multimeter. I installed Windows 7 on the machine and looped PCMark Vantage earlier for over an hour and it completed fine. Maybe all this was due to something with Windows. I'm about to loop Vantage overnight and see what happens when I check it in the morning. If everything is good then I'll start installing games back on the computer and seeing how they work.

[westom]

The power switch isn't going to happen. I need a set of alligator clips or narrower diodes on the multimeter.

As suggested, use a paper clip or a sewing needle.

Moving on. Long hours of test is mostly futile. Use the hairdryer to 'aggravate' marginal semiconductors.

The remaining suspects (that you do not have equipment to locate) are all on the motherboard. Based upon what you are seeing, my number one suspect would be bad capacitors somewhere in the vicinity of the memory and northbridge. Or a larger electrolytic has failed (a cylinder with a flat top where the flat top is bulging up slightly).

Background. Your motherboard has a large copper plate completely across the entire board. Called a ground plane. And voltages all over that sheet will be different. So that ground plane voltage variations do not cause failures, a board has bypass capacitors all over. To find a defect, means using an oscilloscope. Or maybe inspecting every capacitor to confirm it exists. But that means you know where each ceramic capacitor should be and what to look for.

Bottom line - only remaining suspects are on the motherboard. A defect that may have always existed and was slowly getting worse. Using heat may simply teach significance and power of heat as a diagnostic tool. And confirm that the only remaining suspect is indeed guilty. Assuming the switch is not creating an intermittent, only one suspect remains. Sorry for the bad news. Sometimes finding an equivalent motherboard for the existing CPU can be difficult. Sometimes one get lucky - and at a good price.

[Major_A]

Visually the board doesn't have any bad capacitors. I looked at someone's computer last week that kept crashing. Three capacitors around the socket were bulging so I came to the conclusion that's where the instability was coming from.

My sister's house, where this is usually located, runs a tad on the warm side. Plus we live in the Amazon, aka southeast Texas. If a cap hasn't popped yet then I don't think it's going to. Unless there's some other catastrophic failure.

[westom]

Visually the board doesn't have any bad capacitors. I looked at someone's computer last week that kept crashing. Three capacitors around the socket were bulging so I came to the conclusion that's where the instability was coming from.

Capacitors around the CPU socket? CPU has its own power supplies. If those capacitors were bulging, then voltages inside a CPU were unstable which explains that crashing. Those capacitors cost dimes. How easily that motherboard could have been fixed.

Meanwhile, room temperatures have little relevant affect on capacitors. Electrolytic capacitors bulge most often due to manufacturing defects associated with aging. Most every electronic defect is not visually apparent. Those capacitors were the rare exception. Most capacitor failures - especially a maybe 50 ceramic (another type) capacitors all over the board - would have no visual indication.

[Major_A]

I don't know how to check/repair/replace capacitors anyway. Besides the PC with the bad caps was running a PIII, I think it's time for an upgrade.

[westom]

I don't know how to check/repair/replace capacitors anyway.

You never tried to. Therefore you would not yet know how to do it. Doing is how so much gets learned - as demonstrated by the multimeter.

A PIII is a best computer to fix because mistakes cost so little. Go to Radio Shack, Lowes, Tru-Value hardware or anyplace that sells hammers. Buy a soldering iron. Remove a defective cap. It really is trivial if someone does direct and encourage you. Learning to write in 1st grade was probably as difficult. Some also learn dexterity applicable to other tasks. Meanwhile, this is a tangent not relevant to your problem.

Well, inspection rarely finds electronic problems. But the good news. You are not shotgunning; replacing a perfectly good power supply, memory, or disk drives on speculation. The concept is called 'follow the evidence'. Some never learn it.

Confusion has diminished to one suspect because hard facts and basic computer knowledge resulted in a useful recommendation.

[skier]

The PC was running a PIII, I think it's time for an upgrade.

i'd say

though i'm still running a 600MHz for a budget/tax machine....