OT: Help needed,my system becomes unstable

[tycho102]

Skybird:

If that's not your CPU temperature, or your GPU temperature, you need to check your memory.

Download Memtest86 and run that at the ms-dos prompt. You can also use just about any current version of Linux; Fedora, Suse, Knoppix, Overclockix, Ubuntu, Debian. All the Live distros (CD or DVD) include Memtest. You just run "linux memtest" at the command prompt, or choose "memory test" at the menu (such as with Suse).

Configure your memory to run "stock", and let memtest run for 10 hours or more. Don't bother with more than 24 hours. Overclocking instability will show up within the first 10 minutes. Defective ram can take 45 minutes to 10 hours, depending on the severity.

********** IF ALL THAT PASSES ***************

It's your PSU or it's your mainboard's voltage regulators. Capacitors (notably, electrolytic capacitors) since 2002 have dropped in quality. China and Taiwan don't quality-assure their production lines anymore, leaving it up to the OEM's to fail test their systems. If you ever RMA a component and receive a replacement, you can see the results of this business philosophy for yourself. You'll see Sharpie marker slashes across the majority of caps on that component (such as a sound card or video card), indicating that the troubleshooter marked each cap as "good" during rework.

Your mainboard's power regulators, and your PSU, heavily depend on eCaps to supply "clean" power. Unclean power will cause those same "blink outs" like what you have described.

With a 350w power supply, at 50 degrees centigrade, it's only going to be putting out 250w or less. Usually 175w. That same 350w PSU running it's switches at 25C will put out it's full 350w. This is partially why people buy 500w power supplies; at 45C "typical" running conditions, it's still going to put out 350w of power.


======================
So, you will need to swap out a PSU, first. If you can borrow someone's for a few days of testing, that's the cheapest route. If not, you need to "upgrade" to something with a higher 12v current rating.

If memtest ran good, then it's not your processors, so it is most likely your mainboard's caps.

[porphy]

I go with the PSU theory as well. Had a incident almost like that a few months ago. I had overclocked my P4 a bit, and everything was fine. Computer was idle and late in the evening I was about to shut it down when I noticed it already was off. I was a bit tired and thoght I had turned it off earlier. Next morning it did't boot up. Dead. PSU had obviously died the "idle death" . Got a cooltek 500w and quit all OC. Now it's fine and more quiet than ever.

Cheers Porphy

Power supply is only going to supply whatever the load is. IF there's no load, there's no draw. If the load is 200w, it'll supply 200w even if its rated at 500w. A 500w power supply will run cooler supplying 200w than a 350w power supply (it would last longer too). The bottom line: you shouldn't see an increase in the power bill because of larger PSU (unless you're talking about different computers one with 200w PSU and another one with 350w PSU. The 350w PSU will most likely be putting out 150w on a continuous basis (except at startup with inrush currents and so forth), while the 200w unit was probably putting out 100w continuously. So in that case, yeah, you'd notice that on your electric bill. BUT that is the difference between PII and PIV system.

A P4 with current video card should not run on anything less than 400w. If you put a 500w into a P4 system, it MAY draw 300w at startup, but run continuously at about 150w. So you're not going to see a larger electric bill.

My thinking its one of three problems:

1. motherboard failure
2. PSU failure
3. power mains power

With respect to #3, if you're getting brownouts, or sags on the voltage that could cause problems. Spikes and noise on the line can fry components over time. A surge suppressor only handles current surge (it won't deal with over-voltage). The only way to ensure clean power is with a battery backup system. The deal with standby-by power is the same with PSU: you want the highest rated unit you can afford predominantly so that they'll last longer. A decent rated unit's battery should last about 5 years. Units rated too low won't last as long, because every time the PC is running exclusively off of battery, the run-time decreases (even if the battery is at 100% charge). So you don't want a unit that'll only give you 5 minutes of run time, you want one that'll supply at least 30 min (1000 VA or higher).

As far as mains quality, you'd be suprised at how often BBS will trip. If your lights often flicker, or go noticeably dim you're definitely a candidate for a BBS. Even if you don't see something like that, a voltage drop lasting less than 1/10 second (blink of an eye) won't even be visible in the lights. But how many cycles does a CPU do in 1/10 second? Furthermore, a PSU doesn't draw continuous power. In the U.S. we have 115v @50/60 Hz. So the PSU is sipping power at discrete intervals. A 1/10 second interuption is 6 cylcles that's an eternity for a PSU. All kinds of things can cause mains to go low: fridge, A/C kicking on, furnace, or even somebody turning on their table saw next door.

Also, surge protector doesn't filter the power either. So if you ever hear/see noise in the TV/stereo when the A/C, furnace, fridge kick in then there's garbage on mains that is abusing the PSU. Also, thunderstorms can have an effect on the power quality of mains throughout the grid. Lightning strikes on the grid have a way of propagating through the grid as noise (and even prolonged undervoltages sa the system deals with line condition). Over time that garbage can fry components. That's why a decent battery backup with power conditioning is always recommended. Its always a good idea to plug the battery backup into a decent rated surge protector too. While most battery backups have surge protectors built into them, if it blows the unit is prett much shot. However, an external surge protector can easily be replace. Furthermore, surge protectors go bad after time when they're dealing with surges. So its not a bad idea to replace a surge protector every 5-6 years outright. Protecting a battery backups built in surge protection with external surge protection is always a good idea.

[Skybird]

Thanks to all.

However, you cannot explain why the system so far is stable when being under full working load, both CPU and graphics card. It should be then when too instabile power supply shows the most obvious, since then the demand is the highest, CPU, graphics, RAM, and sound.

I now have the symptom - currently the only one - that when I started the system yesterdy and today morning, the LED indicating HD activity is on for 1.5 - 2 minutes, while the screen remains pitchblack and nothing happens, the monitor even switches back to standby mode. Then, all of a sudden, booting begins, monitor goes on, with the Windows loading screen, without displaying the usual BIOS screens. When I switch off immediately after that, and try to reboot, it is the same. However, when the system has run let'S say an hour or so, for longer time, and components are a bit warm, rebooting is done in the usual way, and reliable, and fast. Booting a really cold system= long time, rebooting a warm system (including power switchoff)= short loading time. I also have an old graphics board inserted now, which ran fine since 1.5 years, and before that in an older system. So far no effect from that change. I will put the new card back in, and run it and the fan with a separate AC from an old system for some days.

but somehow I have a feeling now that it neither is overheating, nor power supply. I just came back from town minutes agho, and have asked in the store where I bought the components and let them built the tower. The guys there are both freindly and good on their matter. They said that due to the circumstances of having a stable system even when maximum instability should be expected (full working load), the AC probably is not the problem (as long as it has no erratically appearing problem with broken internal hardware). And according to my description of how the system it set up (open, ventilator), overheating would be a surprise to be the cause of problems. Remains: motherboard problem, RAM problem, HD problem, soundcard problem (the latter I do not take seriously into account).

My tip is the harddrive now, the thing was very hot two days ago (it used to be cold, since it get's it'S share from the ventilator, and directly), at one, as described somewhere above. I could imagine that it received some kind of damage from that. But overall heat problems I cannot imagine to be the source of problems, the tower simply is too open and receives too much fresh air from a separate ventilator.
Will exchange the HD sometime this weekend, and let the system running all day for the time being.

I have no battery backup. the talk above was only about that small Lithium cell that motherboards are used to be equipped with. We also have absolutely stable power lines over here, no blinking, no blackouts or brownouts, nothing.

Never have built a system by hand myself, only decided on components, and let other do it. I can chnage a sound or graphics board or RAM bars, but a motherboard and all cables and AC - still quite impresses me.

Assuming I try a new AC - and I will not invest that money as long as it is not safe to say it is the AC causing the problem - could I trust in that if connecting connectors really matches by form, that the according device really gets the voltage and ampere it needs, then? Or do different power supplies in those many cables of a AC nevertheless sometimes have the same plugs?

[porphy]

Hi Sky

I'm no expert in these matters, but don't you think it sounds as if when you cold start the comp, your PSU is not up to it. When the components in it get warm, it stabilizes and runs fine and therefore cope with full load use. This also fits with the behaviour that it restarts just fine after some use.
HDD failures usally go with clicking sound and the drives spinning, but you don't seem to get that and the HDD should show problems when working, not only at boot I guess. I still think your PSU is dying, but perhaps a slow death. Not a heartattack like mine. But it's hard to know for sure without changing parts around, it sounds as if you will have solved the puzzle soon though.
You could take the PSU to the shop and let them measure the output. I did that to be sure it was the psu that had gone out. Connecting a new PSU is'nt that hard as long you have documentation on your motherboard layout. From this you can figure it out, no doubt. Just check if the connector to motherboard is ATX 2.0, most PSU have a adapter if you run a older motherboard. The ppl in the shop should sort that out for you.

[Skybird]

Originally Posted by porphy

Hi Sky

I'm no expert in these matters, but don't you think it sounds as if when you cold start the comp, your PSU is not up to it. When the components in it get warm, it stabilizes and runs fine and therefore cope with full load use. This also fits with the behaviour that it restarts just fine after some use.
HDD failures usally go with clicking sound and the drives spinning, but you don't seem to get that and the HDD should show problems when working, not only at boot I guess. I still think your PSU is dying, but perhaps a slow death. Not a heartattack like mine. But it's hard to know for sure without changing parts around, it sounds as if you will have solved the puzzle soon though.
You could take the PSU to the shop and let them measure the output. I did that to be sure it was the psu that had gone out. Connecting a new PSU is'nt that hard as long you have documentation on your motherboard layout. From this you can figure it out, no doubt. Just check if the connector to motherboard is ATX 2.0, most PSU have a adapter if you run a older motherboard. The ppl in the shop should sort that out for you.

Good thought, I think that is a realistic possebility. Now that you say it I remember that some weeks ago the fan in the PSU did made some kind of noise. I needed to repeatedly shake it a bit in order to get it running smoothly again. Since it has not come back, I forgot about it completely.

I keep that one in mind, thanks.

[Skybird]

I recalled the type of PSU wrong, anyway. It is no 350W, but a 420W device. I think I mixed it up with the PSU in my former system.

[Wim Libaers]

Originally Posted by Skybird

I have no battery backup. the talk above was only about that small Lithium cell that motherboards are used to be equipped with. We also have absolutely stable power lines over here, no blinking, no blackouts or brownouts, nothing.

1. That battery doesn't matter. You can even run computers without that battery, but you'll probably have to reset the time and other BIOS settings on every cold boot.

2. You mean you do not have any obvious instability. That you cannot easily see the instability doesn't mean there are no brief errors, or slow fluctuations. (i.e. errors that occur to fast or to slow to be obvious without measuring equipment)

3. It is possible for a power supply circuit to be more stable at high load than at low load. But I have never worked on any PC switching PSU, and there must be many different designs out there, so it's impossible to say if this would be the case for your system, and I don't know if there is a type of damage that will cause such instability when a PSU in good condition would not fail.

[SUBMAN1]

Originally Posted by Skybird

Thanks to all.

However, you cannot explain why the system so far is stable when being under full working load, both CPU and graphics card. It should be then when too instabile power supply shows the most obvious, since then the demand is the highest, CPU, graphics, RAM, and sound.

I now have the symptom - currently the only one - that when I started the system yesterdy and today morning, the LED indicating HD activity is on for 1.5 - 2 minutes, while the screen remains pitchblack and nothing happens, the monitor even switches back to standby mode. Then, all of a sudden, booting begins, monitor goes on, with the Windows loading screen, without displaying the usual BIOS screens. When I switch off immediately after that, and try to reboot, it is the same. However, when the system has run let'S say an hour or so, for longer time, and components are a bit warm, rebooting is done in the usual way, and reliable, and fast. Booting a really cold system= long time, rebooting a warm system (including power switchoff)= short loading time. I also have an old graphics board inserted now, which ran fine since 1.5 years, and before that in an older system. So far no effect from that change. I will put the new card back in, and run it and the fan with a separate AC from an old system for some days.

but somehow I have a feeling now that it neither is overheating, nor power supply. I just came back from town minutes agho, and have asked in the store where I bought the components and let them built the tower. The guys there are both freindly and good on their matter. They said that due to the circumstances of having a stable system even when maximum instability should be expected (full working load), the AC probably is not the problem (as long as it has no erratically appearing problem with broken internal hardware). And according to my description of how the system it set up (open, ventilator), overheating would be a surprise to be the cause of problems. Remains: motherboard problem, RAM problem, HD problem, soundcard problem (the latter I do not take seriously into account).

My tip is the harddrive now, the thing was very hot two days ago (it used to be cold, since it get's it'S share from the ventilator, and directly), at one, as described somewhere above. I could imagine that it received some kind of damage from that. But overall heat problems I cannot imagine to be the source of problems, the tower simply is too open and receives too much fresh air from a separate ventilator.
Will exchange the HD sometime this weekend, and let the system running all day for the time being.

I have no battery backup. the talk above was only about that small Lithium cell that motherboards are used to be equipped with. We also have absolutely stable power lines over here, no blinking, no blackouts or brownouts, nothing.

Never have built a system by hand myself, only decided on components, and let other do it. I can chnage a sound or graphics board or RAM bars, but a motherboard and all cables and AC - still quite impresses me.

Assuming I try a new AC - and I will not invest that money as long as it is not safe to say it is the AC causing the problem - could I trust in that if connecting connectors really matches by form, that the according device really gets the voltage and ampere it needs, then? Or do different power supplies in those many cables of a AC nevertheless sometimes have the same plugs?

One more thing here - my mother in laws system started experincing similar things as you type here. The HD LED would just come on and hang the system forever, but it would eventually come back in her case. This again was a PSU issue, but was a Shuttle if you know what that is. Replaced the PSU with same model - fixed all the issues and works fine. PSU's go bad over time.

-S

PS. Run Seatools to see how the HD is operating. You can find it on Seagates site. If it reports everything is OK (like it did with my mother in laws system since I also thought bad HD) then it is not your HD.

PPS. Windows XP is very resilient to cracshing in circumstances like this by the way.