2010-11-06, 07:15 AM
The thing you have to do is single out which component is causing the failure. The trifecta of failures in a high load system are the memory DIMMs, CPU, and GPU. Looks like you've singled out the memory.
I'd encourage another run of memtest to see if you get anything there. Leave no stone unturned. Allow memtest to run for an hour to simulate a high load environment.
Before you do the next tests, get HW monitor from here and run it while you are running the benchmarks. Verify that your voltages stay stable. A sudden dip in voltage will cause your CPU or Graphics Card to segfault and cause the errors you're reporting. Keep your eye on the voltages if anything happens to fail. This way if there's a dip in voltage and then a power failure, you can say it's the power supply and not the component. While you've got HWMonitor running, make sure that your voltages are within tolerance of what they should be. If it's reporting, say, 11v on the 12v rail, your problem is 95% the power supply. It could be possible that the motherboard is reporting the wrong voltages, which is why it's hard to be 100% sure. But when you've got a problem like you've got, it's worth it to know that a power supply is good and that you have a component failure elsewhere.
The next thing you can single out is your CPU. Get the SiSandra Lite Burn-in utility and make it stress the CPU. Again, run it for one to two hours to simulate high load.
To test the graphics card, get Furmark and run that for an hour. If HWMonitor doesn't report GPU temperatures, get a tool that will read it for you. I understand you've checked the temperatures before, but you can't just assume that nothing has changed in the past four months. Your problems are worsening - something has changed.
While you are running these tests, look inside your computer. Are all the fans running? Does it smell of burn at all? Put your hand behind the power supply. Is the fan working inside the power supply? Does the power supply feel hot to the touch (it shouldn't)? On some power supplies, there is a red light on the back of it next to the power switch. If that red light comes on, that means there's a wiring fault somewhere inside the case and it also means that it's definitely a problem with the power supply (100% this time!).
If each part works individually just fine, then you know it must be the mobo or the power supply. Since replacing the power supply is generally less invasive, is more likely to be the cause, and is cheaper than a motherboard, I'd recommend replacing that first.
As I'm sure you know, you have got to take care of this now. This sounds like imminent component failure, and within a couple of months you won't be able to turn on your computer at all. I would strongly advise that even if you don't have the money now, save like $10 a week for a power supply. You'll thank me later.
As far as replacing components, perhaps it's not as bad as you think. For some older systems, there are a wide variety of cheap components that is not on the mainstream market anymore. Call up a computer business and ask them if they have any old parts they just want to sell at discount to you. If nothing else, just buy a cheap component that you know works and throw that in there. You can get an older CPU model for as little as $40.
I'd encourage another run of memtest to see if you get anything there. Leave no stone unturned. Allow memtest to run for an hour to simulate a high load environment.
Before you do the next tests, get HW monitor from here and run it while you are running the benchmarks. Verify that your voltages stay stable. A sudden dip in voltage will cause your CPU or Graphics Card to segfault and cause the errors you're reporting. Keep your eye on the voltages if anything happens to fail. This way if there's a dip in voltage and then a power failure, you can say it's the power supply and not the component. While you've got HWMonitor running, make sure that your voltages are within tolerance of what they should be. If it's reporting, say, 11v on the 12v rail, your problem is 95% the power supply. It could be possible that the motherboard is reporting the wrong voltages, which is why it's hard to be 100% sure. But when you've got a problem like you've got, it's worth it to know that a power supply is good and that you have a component failure elsewhere.
The next thing you can single out is your CPU. Get the SiSandra Lite Burn-in utility and make it stress the CPU. Again, run it for one to two hours to simulate high load.
To test the graphics card, get Furmark and run that for an hour. If HWMonitor doesn't report GPU temperatures, get a tool that will read it for you. I understand you've checked the temperatures before, but you can't just assume that nothing has changed in the past four months. Your problems are worsening - something has changed.
While you are running these tests, look inside your computer. Are all the fans running? Does it smell of burn at all? Put your hand behind the power supply. Is the fan working inside the power supply? Does the power supply feel hot to the touch (it shouldn't)? On some power supplies, there is a red light on the back of it next to the power switch. If that red light comes on, that means there's a wiring fault somewhere inside the case and it also means that it's definitely a problem with the power supply (100% this time!).
If each part works individually just fine, then you know it must be the mobo or the power supply. Since replacing the power supply is generally less invasive, is more likely to be the cause, and is cheaper than a motherboard, I'd recommend replacing that first.
As I'm sure you know, you have got to take care of this now. This sounds like imminent component failure, and within a couple of months you won't be able to turn on your computer at all. I would strongly advise that even if you don't have the money now, save like $10 a week for a power supply. You'll thank me later.
As far as replacing components, perhaps it's not as bad as you think. For some older systems, there are a wide variety of cheap components that is not on the mainstream market anymore. Call up a computer business and ask them if they have any old parts they just want to sell at discount to you. If nothing else, just buy a cheap component that you know works and throw that in there. You can get an older CPU model for as little as $40.