New built system having error on prime95

[inwood07]

I ran prim95 blend test a couple seconds later worker 4 & 7 says "fatal error: rounding was .5, expected less than .4. Hardware failure dectected, consult stress.txt.file."

i am guessing it could be when i installed my heatsink/thermal paste. i didnt have the thermal paste on correctly. here are my system specs

i7 920 c0 2.66
Xigmatek s1283v
OCZ platinum ram (7-7-7-24@1.64)
gigabyte ud4p x58
WD black 1tb
corsair 750
Coolermaster HAF 932

thanks in advance

 

[mcnuggetofdeath]

What temperature is it hitting when you run Prime95. That should give you an idea whether the issue is the installation of the heatsink, or a faulty CPU.

 

[boulard83]

use CORETEMP and watch temp.

 

[inwood07]

how would i check the temps, i am still a noob.

 

[inwood07]

ok they are around 67-71 degrees

 

[mcnuggetofdeath]

reseat your heatsink and make sure its attached properly. did you apply your own thermal paste? countless guides all over google about how to do that if so. to my knowledge thats nearing the thermal limits of that chip, and it shouldnt be anywhere near that at stock speeds.

 

[inwood07]

thats what i figured i guess i have to wait for my thermal paste remover to come in and reapply it. i used the pea size method. thanks for your insight.

 

[mcnuggetofdeath]

Just use rubbing alchohol and a paper towel to remove the stuff. You dont need a special compound. Then use a credit card to get it thin enough.

 

[inwood07]

ok i reapplied my thermal paste and temps are down around 57 degrees at max load, but prime 95 still says the same thing. what should i do now?

 

[mcnuggetofdeath]

Depending on what test i run i get different results. For instance, I can run small ffts at 4.1GHz, but even trying a blended test results in an immediate crash. You might very well have a defective CPU. try some of the other stress testing software though. OCCT, Intel Burn in test, if none of those fail then your good. Frankly if you can run Prime95 for any extended period of time your probably safe. Just not 100% stable. sorry i cant be more help. you could try overclocking or under clocking it. but the simplest thing would be to make sure all your voltages are set to the right level. some motherboards default with too little or too much voltage. check with your various manufacturers to make sure everything is where it should be.

 

[orangegator]

My guess would be bad ram. First make sure you have it set to manufacturer specs. Then use memtest86+ to test it.

 

[mcnuggetofdeath]

Rounding errors are not indicative of bad RAM. RAM might cause it to spontaneously crash, not return mathematical errors. Run small FFTs to eliminate that variable all together.

 

[coraythan]

Rounding errors are not indicative of bad RAM. RAM might cause it to spontaneously crash, not return mathematical errors. Run small FFTs to eliminate that variable all together.

I just had this exact problem. I was also getting crashes in Lord of the Rings Online, and spontaneous blue screens. I ran memtest86+, it found a few errors, I took out one of the two sticks of ram, and everything seems to be running perfectly.

I don't understand why, but everything from my experience yesterday and today suggests that Prime95 rounding errors can be due to RAM.

 

[mcnuggetofdeath]

Physically damaged RAM means it doesnt store data correctly. If it doesnt store data correctly than the issues would present themselves elsewhere. IE in LOTR online as you yourself experience. It also results in random lock ups, crashes, BSODs, etc. To suggest that one would only happen upon issues resulting from bad RAM in Prime95 is ludicrous. You would notice issues almost everywhere. As to removing the bad stick of RAM, having fewer modules occupying your slots allows for a higher overclock and thereby more stability. If there are fewer things causing memory faults, you'll hit higher frequencies with fewer problems as the system has less to compensate for.

 

[orangegator]

Rounding errors are not indicative of bad RAM. RAM might cause it to spontaneously crash, not return mathematical errors. Run small FFTs to eliminate that variable all together.

So, according to your logic, running the blend test which "tests lots of ram" will only crash from ram errors and with spit back rounding errors if it's the cpu? You realize you are talking out of your ass.

@OP Ignore this guy. Thoroughly test your ram and you will likely find you problem.

 

[mcnuggetofdeath]

Im saying bad RAM will manifest itself elsewhere. A rounding error is different then it returning the wrong value. It displays a different error message. Try reading more than a single post. Way to stay classy btw.

 

[orangegator]

Im saying bad RAM will manifest itself elsewhere. A rounding error is different then it returning the wrong value. It displays a different error message. Try reading more than a single post. Way to stay classy btw.

I'll admit I'm not 100% positive, but I'm pretty sure all prime95 errors show up as a rounding error, unless it's a system lockup or crash.

 

[Twoboxer]

Whatever lol . . . if this was a C2D, DDR2, I'd be focusing on the RAM and pull sticks, loosen timings, increase voltage because I'd be guessing a memory fix would get it stable.

 

[jsc]

I'll admit I'm not 100% positive, but I'm pretty sure all prime95 errors show up as a rounding error, unless it's a system lockup or crash.

Pretty much true. I have never seen anything else. I also have never seen a properly working, stock clocked system cause a P95 error.

First thing is is the system overclocked? Second thing is what are the temperatures? What kind of HSF? What kind of case?

I set the motherboard CPU thermal alarm to 70 C. That is my "do not exceed" threshhold. Excellent stickie on thermal and voltage management:
Core 2 Quad and Duo Temperature Guide
http://www.tomshardware.com/forum/221745-29-core-quad-temperature-guide

After memtest for about 24 hours, I start overclocking by relaxing the memory timings and set them to recommended factory voltage - usually not the default 1.8 volts for DDR2.

The current Intel CPU's are good enough that you can achieve a sizable OC (maybe 15% - 25%) without increasing voltages. After that, for awhile, you get a linear increase in frequency with small increases in voltage. After that, there will come a point when you get a small increase in frequency for a larger increase in voltage. I stop at that point regardless of temps. Then I try to optimize the memory.

I stress test for 24 hours. I have had systems fail at 14 and 18 hours.

 

[inwood07]

i ran memtest86+ and got no errors.

 

[inwood07]

i also ran prime95 again. i get the same error with lower temps (56-61 degrees). idle temps are around 35-41 degrees.

 

[redgarl]

i also ran prime95 again. i get the same error with lower temps (56-61 degrees). idle temps are around 35-41 degrees.

You still got a hell of a time to assemble your hardware here... It's simple, your heatsink is not well installed.

Take your motherboard out and check your pins connected to your board. Also check your instructions of your CPU. I never worked with i7 cpu, so I am clueless about the way the HS is locked on your mobo, but I know what a HS not placed the right way do.

Sometime placing the HS is really a pain in the ass.

Also, the chance that your OCZ RAM was DOA is almost zero, so focus on your HS... but your idle temps seem okay for me...

 

[inwood07]

i am pretty sure i have the heatsink installed correctly

 

[Twoboxer]

You're getting a "memory" error under load. Stop screwing with the processor, its fine. Look elsewhere.