Prime95 Failure... Advice?

[Catalan]

I recently finished my new system, and since I had plans to overclock, I decided to do some testing with Prime95 before making any changes. I ran one instance with Small FFT's and another with Blend. In less than 8 hours, Blend failed...

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 7 hours, 56 minutes - 1 errors, 0 warnings.
Execution halted.

Here are my basic specs:
Opteron 170 (with stock fan)
ASUS A8N-SLI Premium
eVGA GeForce 7800GT
1 Gig Corsair DDR400
Antec TPII-550

CPU temperature never exceeded 45 degrees, the whole time. Voltages were all normal, as well. Previous runs of memtest showed no memory problems.

Should I give up on the overclock before I even get started? Should I be thinking about returning the processor? Should I run more tests? Any help is appreciated.

 

[mr-busta]

Do a CMOS reset (clear BIOS to default settings) and double check everything is running at stock settings: CPU voltage (1.35V), mem timings (auto), mem speed (200MHz), mem voltage (2.7V), HT (200MHz), etc.

My thinking is that if it had run for 8 hrs on small ffts, most likely your CPU is ok. probably some random glitch on other components that caused the failure. Check heatsinks on the chipset, are they overheating? I've fixed some failures that happened after several hours of priming by re-doing chipset heatsink thermal paste from the factory with Arctic Silver.

 

[Catalan]

Do a CMOS reset (clear BIOS to default settings) and double check everything is running at stock settings: CPU voltage (1.35V), mem timings (auto), mem speed (200MHz), mem voltage (2.7V), HT (200MHz), etc.

Pretty darn sure everything's at stock settings, but I'll give that a shot anyway.

My thinking is that if it had run for 8 hrs on small ffts, most likely your CPU is ok. probably some random glitch on other components that caused the failure.

Hope that's the case... I'm still running Small FFT's with no problems. I'll leave it running until I have time to reset the CMOS, in the morning.

Check heatsinks on the chipset, are they overheating? I've fixed some failures that happened after several hours of priming by re-doing chipset heatsink thermal paste from the factory with Arctic Silver.

Motherboard chipset? Highest I've seen the motherboard temps go is 44 degrees, but it's usually been 42 to 43. Not really sure how high is too high, there... Since the thing's passively cooled through a built in heat-pipe, I don't believe it can be removed, but I thought there was an optional chipset fan in the packaging somewhere. I could try attaching that.

 

[Absolute0]

8 hours? Seems stable to me. Run for a few days at those speeds, if none of the usual games and programs crash, i don't see what's wrong with it. The real stress test to me is hours of gaming, if a computer can do that, i dont care what prime says.

 

[Logan321]

memory timings at auto = bad. Especially if you are running performance or "enthusiast" ram. The motherboards often mis-read the ram settings, or they just aren't programmed correctly in the modules...

For example, my OCZ 3200 on "auto" yields 2.5-3-3-5, and it's rated on the modules 2-3-3-6. On auto I get some instability because of the 5 vs 6.

 

[Aaron_ATX]

Its probably a heat issue. Maybe not specifically the cpu, but it may be. My rig will prime for 10 hours, loading at 43-44c, with the side of the case off. If I put the side back on, load temps hit 45 and prime fails in less than 3 minutes. (large ffts)

Seems like during the course of 8 hours the case heats up just enough to crash it, Ambiant temps can easily change a few degrees inside my house during a day....makes sense to me at least.

 

[homersapien]

8 hours? Seems stable to me. Run for a few days at those speeds, if none of the usual games and programs crash, i don't see what's wrong with it. The real stress test to me is hours of gaming, if a computer can do that, i dont care what prime says.

Terrrible advice. Ignore him. Errors are errors.

memory timings at auto = bad. Especially if you are running performance or "enthusiast" ram. The motherboards often mis-read the ram settings, or they just aren't programmed correctly in the modules...

For example, my OCZ 3200 on "auto" yields 2.5-3-3-5, and it's rated on the modules 2-3-3-6. On auto I get some instability because of the 5 vs 6.

Great advice. Set the RAM timings Manually and run Memtest86. You might want to do a search in the Corsair Support forum, chances are Ramguy has the timings listed there. Don't forget the voltage as well, its not uncommon to have to bump up the voltage, especially with Corsair modules

 

[Absolute0]

Terrrible advice. Ignore him. Errors are errors.



Great advice. Set the RAM timings Manually and run Memtest86. You might want to do a search in the Corsair Support forum, chances are Ramguy has the timings listed there. Don't forget the voltage as well, its not uncommon to have to bump up the voltage, especially with Corsair modules

EDIT Well i still don't think it's terrible advice or anything to cry about, but i have to edit because i just noticed that he was failing prime @ stock... usually i expect to see these 170s instantly cranked to 2.8.

Oh wait a second, i thought this was already overclocked, you're failing at stock settings. As they said the ram. Since small FFTs ran, the CPU is fine doing stock, blend tells you there is a problem with system resource management (memory), that needs tweaking / addition of volts.

It is definitely NOT a heat issue. At lower clock speeds, a processor can take more heat while maitaining stability than at higher clock speeds (due to electron interference more easily affecting stability when higher frequencies are forced). loading at 45C at the stock 2.0 Ghz speed is perfectly acceptable, many load higher without errors.

 

[Aaron_ATX]

It is definitely NOT a heat issue. At lower clock speeds, a processor can take more heat while maitaining stability than at higher clock speeds (due to electron interference more easily affecting stability when higher frequencies are forced). loading at 45C at the stock 2.0 Ghz speed is perfectly acceptable, many load higher without errors.

Good point, but again I would still consider heat as a possible cause....overheating mosfets, ram, etc etc may trip some instability with prime95 after a long period of time.


If it was my system i'd give the cpu just a bit more voltage and try again. If the same thing happens, give the ram a bit more voltage also and try priming once again........BUT first and foremost get those ram timings set correctly.