Failed Prime95, now what?

[paperfist]

This is on a non-overclocked P4 machine. Prime 95 ran for 1 hour 6 minutes and came up with: Fatal error, rounding was .5 expected less than .4

I ran it with large FFT.

The machine will hard lock sometimes; crash and/or crash with tons of screen corruption.

It has onboard video so I'm not sure if the error means CPU for sure or if it's a derivative since the CPU also effect the RAM and video.

Any ideas?

 

[paperwastage]

temps for the CPU/mobo/machine?

clean it up a little (remove dust)

 

[paperfist]

temps for the CPU/mobo/machine?

clean it up a little (remove dust)

It's as dust free as possible, I didn't write down the temps but they weren't out of range.

Running memtest86 atm.

 

[sm625]

Can you raise the voltage?

 

[Idontcare]

I ran it with large FFT.

Large FFT usually finds instabilities in the memory path, such as the Northbridge chipset which houses your memory controller or the ram itself.

Small FFT is more suited to identifying issues with the core of the CPU itself.

Since you say it failed large FFT, it is logical to next test the memory (memtest+ or HCI memtest) as well as to test the core of the CPU with small FFT (or, preferably, by using LinX/IBT - see the OC stability sticky for download links).

At this stage of the game - identifying the root-cause of the instability - I personally recommend you do not listen to the advice to increase your voltages and what not. That can hasten the death of a possibly already dying component on your mobo, ram, or CPU.

Instead, I recommend you go the opposite direction, at least for the utility of finding the culprit. Underclock your CPU, set the multiplier to a lower one, or underclock your ram, or underclock the FSB (all while leaving the voltages themselves alone, leave at current BIOS settings).

Check large FFT again when you make individual changes and see if the instability goes away. When/if it does, then you will have a much better idea (and in a way that was safer for your hardware) as to which component in the loop is likely to be the problematic one.

Only then do you start the next phase which is solving the problem. (be it with better cooling, higher voltages if prudent, replace the faulty component, etc).

 

[Rvenger]

Check your caps on the board. I bet those electrolytic capacitors are buldging or completely dry.

 

[paperfist]

Thanks for the advice guys, I will continue to test again with the different settings.

Changing the voltage in either direction is kind of throwing me for a loop as this machine has been in use for at least 8 years with no problems till now. I guess that may indicate a capacitor problem which I'll check also.

Plus it's an old HP business desktop so I'm thinking there's no options for a voltage change, but I'll look.

EDIT: Ran HCI memtest for 176.5% coverage and no errors.

 

[THRiLL KiLL]

Large FFT usually finds instabilities in the memory path, such as the Northbridge chipset which houses your memory controller or the ram itself.

Small FFT is more suited to identifying issues with the core of the CPU itself.

Since you say it failed large FFT, it is logical to next test the memory (memtest+ or HCI memtest) as well as to test the core of the CPU with small FFT (or, preferably, by using LinX/IBT - see the OC stability sticky for download links).

At this stage of the game - identifying the root-cause of the instability - I personally recommend you do not listen to the advice to increase your voltages and what not. That can hasten the death of a possibly already dying component on your mobo, ram, or CPU.

Instead, I recommend you go the opposite direction, at least for the utility of finding the culprit. Underclock your CPU, set the multiplier to a lower one, or underclock your ram, or underclock the FSB (all while leaving the voltages themselves alone, leave at current BIOS settings).

Check large FFT again when you make individual changes and see if the instability goes away. When/if it does, then you will have a much better idea (and in a way that was safer for your hardware) as to which component in the loop is likely to be the problematic one.

Only then do you start the next phase which is solving the problem. (be it with better cooling, higher voltages if prudent, replace the faulty component, etc).

sound advice!

 

[paperfist]

So I ran Small FFT and from Fri Dec 16 15:29:03 2011 to Sat Dec 17 14:02:09 2011 it ran without any errors.

It passed HCI memtest a well.

BIOS is locked down, can't change any voltages.

Is there anything to test onboard video with?

 

[pig_10]

I think SLK's advice is worth a try.
since your computer is manufactured 8 years ago, there's highly possible that one or more capacitors fail.
maybe you can use a multimeter to check every electrolytic capacitors's capacity on your motherboard

 

[Harvey]

I think SLK's advice is worth a try.
since your computer is manufactured 8 years ago, there's highly possible that one or more capacitors fail.
maybe you can use a multimeter to check every electrolytic capacitors's capacity on your motherboard

Unless they're completely shorted, it's somewhere between very difficult and impossible to use a multimeter to check bad caps in circuit, and even then, you'd have to know how a good cap is supposed to look on a multimeter, which can vary depending on the value of the cap and which lead you connect to the positive and negative terminal, and then, only while the board is unpowered.

I design electronic products, and my experience is that many electrolytic caps fail after a few years, especially in a warm environment, although newer, solid electrolytics are better. As SLK suggests, look for bulging (domed) tops or leakage around the base of these caps, but even ven then, old caps can fail without showing external signs.

If any of them has failed, even if you can find the right parts, replacing electrolytics on a multi-layer circuit board is dicey because, when you heat the solder joint to pull the lead through the board, you can damage connections from an internal layer to the solder pad. Once that happens, the board is trash.

You should also check the electrolytics on your vid card.

Capacitors with bulging tops

Capacitor leaking at the base

Capacitor bulging and leaking at the top​

 

[tweakboy]

Get those pictures away from me LOLOOLOL talk about dust and damage .... ouchie

Well the OP hes getting it when he uses the computer as well. Like freezes and lockups like he said and screen corruption. I am going to blame possibly your onboard video card and its corrupt drivers... give us some more info also if you can..... thx gl

 

[SickBeast]

OP try running memtest to see if it's your ram.

 

[Rvenger]

I have another random suggestion.

If the northbridge and southbridge have heatsinks on them, clean up and reapply the thermal paste on them. That old stuff is probably harder than plastic.

 

[paperfist]

OP try running memtest to see if it's your ram.

Did that and no errors. Only errors I've gotten was (posted above) Prime 95 large FFT. Small FFT hasn't produced any errors in 2 days of testing.

-----

Harvey,

Thanks for the pics. Problem is the computer is that it's in use for business 24/7 so while I need to do a tear down to check it out I can't atm as it's in use.

-----

tweakboy,

That's my original thought, onboard video failure. It doesn't always screen corrupt when it locks though, a few times it just blue screen and locks. I have to check out the log dumps though.

------

SLK,

Good idea, I'll do that when I tear it down for a capacitor check. It's really amazing though that with all this testing it hasn't crashed at all or over heated. The fans and heat sinks are unbelievably clean being that it's located in a zoo.

 

[fastamdman]

If you are failing p95 it has nothing to do with your onboard video. It's either cpu related or ram related. Check your ram timings.

I would go into bios, "load fail safe defaults" and then re-run the test again.

A MUCH better test though is LinX. It will tell you if you are truly stable or not. What p95 takes 18hrs to do LinX can do in 8mins

 

[Rvenger]

If you are failing p95 it has nothing to do with your onboard video. It's either cpu related or ram related. Check your ram timings.

I would go into bios, "load fail safe defaults" and then re-run the test again.

A MUCH better test though is LinX. It will tell you if you are truly stable or not. What p95 takes 18hrs to do LinX can do in 8mins

Agree with Fastamdman, if your onboard video was toast, you'd be getting constant freezes and artifacts all over the screen.