Running Stress Prime 2004 and Blend Test CPU + RAM FAILS

[Ronstang]

The CPU stress test and the RAM stress test both pass with not problems. I can leave them run for the whole evening and nothing. When I run the Blend test which tests the CPU and RAM at the same time I get an Illegal Sumout failure in under a minute and it reports a possible hardware failure. What I want to know is how to interpret these results. Am I safe to assume something is wrong with my motherboard?

 

[formulav8]

You may want to mention some system specs. Especially whether you are oced or not. The cpu speed, ram speed and timings, ect....



Jason

 

[Ronstang]

It is a Lanparty NFII Ultra and I am running an XP 2500+ and 2 512 sticks of Corsair value ram DDR3200 all at stock timings. This problem just developed the other day.

 

[Ronstang]

Anyone, I have an extra Lanparty NFII Ultra in the closet, is it worth my time to make the switch? Is the mobo the likely cuplrit?

 

[Ronstang]

OK, I switched out the mobo and the problem persists so I do not know what it could be now. Any suggestions?

Is ther any possibility it could be a power supply? I have been having weird random lockups and last time it did that it turned out to be the power supply...but this is an Anted True Power 550 so I find it hard to believe although lately I have heard some switching noise that sounds like it is coming form the PS occasionally.

 

[dunkster]

Have you checked voltages under stress testing?

My guess is that there are no hardware problems - instead your problems may be due to 'draw-down' or 'voltage-droop' of power supply voltages under load, causing CPU Vcore to drop just enough for instability.

I'd try:
- Bumping Vcore a bit, so CPU stays stable even with a little drop, and;
- Test for RAM stability with Memtest86.

Hope this helps!

 

[Ronstang]

OK, everytime I think I have this fixed it comes back so I am really starting to think it is a Vcore voltage problem like you say dunkster. My question is what is the Vcore supposed to be? Motheroard Monitor lists two voltages, Core 1 which is 1.63 and Core 2 which is 1.5. I know I can change it on the mobo but I need to know how much I should go. This is a Barton 2500+ that I am simply running as a 3200+ (simply setting FSB to 200)because that is what a Barton 2500+ really is anyway. The machine has been running rock stable this way for a year.

 

[Ronstang]

So no one can help me?

 

[Ronstang]

I upped the Vcore to 1.70V and Prime now fails even sooner so I really do not know what to do here.

Actually it seems to be erratic becuase now Prime runs without an error in the first 2 minutes where just a few minutes ago it only ran about 10 seconds before failure. This is getting tedious.

 

[trueimage]

turn that vcore back to default
unless you'd like to fry things.
that may be an acceptable voltage for your cpu but i don't know.

I'd set everything back to default and list your full system specs here so we can help.

maybe post a screenshot of your mbm readings and cpu-z

 

[Ronstang]

The Vcore is back to default. The system is an AMD XP 2500+ OC'ed to a 3200+ simply by running the FSB at 200, and it has been stable this way for a year. The mobo is a DFI Lanparty NFII Ultra and I have two 512 sticks of Corsair DDR 3200 Value Ram, a Geforce 6800OC at stock settings, and three hard drives, on optical drive, and a floppy. I have tried running the processor back at default speed but it made no difference.

As far as MBM it is showing core 0 at 1.63V and core 1 at 1.50V case temp is 35C CPU is 42C and CPU Diode is 48C. all at idle but under full load it will only go up about 5C.

EDIT...while running Prime2004 the Vcore drops to 1.62 then back to 1.63 then back to 1.62 but the test is not failing this time. This is getting to be a pain.

 

[Ronstang]

I am starting to think I am chasing a ghost.

 

[Technonut]

Have you run the stress tests not OC'ed? Give it a try.....

 

[Ronstang]

Originally posted by: Technonut
Have you run the stress tests not OC'ed? Give it a try.....

Already did that, no difference.

 

[Ronstang]

2 hrs on the CPU stress test with no errors, I ran it for at least that long on the RAM only test the other night with no problems. It only fails on the blend test with an "illegal sumout" and it is not always consistent. One time it will fail almost immediately and then the test will run fine for an hour or more. The biggest problem is I have been having the BSOD (the XP Pro version of them anyway) and the computer will just restart. It is very unpredictable as to when this will happen as it does not always follow a pattern. Sometimes I have write failures and the chkdsk utility has to run and there are all kinds of problems on the HDs that seem to be fixable...as long as they do not happen on the C: drive which means I usually have to reinstall windows. This is really getting annoying.

 

[Regulator07]

i know exactely what you are talking about with the prime blend test, i run memtest86 and large and small fft tests and get no problems stock and with an overclock, then i run blend and sometimes i get errors in 1 min, 2 min, 5 min, sometimes it runs for 2 hours no problem and i just turn it off, i dont really get it, but i am getting more and more convinced that the blend test is the problem and not the hardware, i have run it at stock and overclocked and errors are always sporatic and make no sense, but the comp is stable always and have no problems. I hope someone can explain why this is happening.