Prime 95 stress test problem

[Finns14]

Ok so I am not a computer expert but I built my own computer so I thought changing the CPU would be no problem (going from a 3200+ to a 4200 X2) but after installing the chip and gaming I was noticing crashing. I took the HSF back off and realized I had put WAY to much thermal past on so I removed the excess and thankfully there seemed to be no issues with the pins. Just to make sure I dl'd prime95 and ran the torture test and to my dismay I had an error, FATAL ERROR: Rounding was .05 expected less than .04 . I googled the issues and the only results I got were that people sometimes got this issues when they attempted to OC (which I am not) and needed more volts to the CPU. Thanks

 

[LOUISSSSS]

try a complete reformat, reset your bios to default and try prime95 again..

 

[Finns14]

Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

 

[Amaroque]

In my experience (I ran Prime95 as a DC project for years), round off errors are usually due to an overheating CPU. To a lesser extent, it could also be a RAM problem.

Reinstalling the OS won't make any difference.

 

[Amaroque]

Originally posted by: Finns14
Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

That's because Prime95 only uses one core. Try running Orthos (a dual core version of Prime95) for testing both CPU's at once.

 

[Finns14]

I was thinking it was a heat issue but the Sensor is telling me its hitting 102F when the stress test is running (Even though its not running for long) it idles at about 99F

 

[Finns14]

Originally posted by: Amaroque

Originally posted by: Finns14
Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

That's because Prime95 only uses one core. Try running Orthos (a dual core version of Prime95) for testing both CPU's at once.

So I am confused no matter what I have been doing on the computer I don't see the 2nd core being used.

 

[Finns14]

I got the same error when running orthos but when I looked in everest when running the test it still said my other core was running at 0

 

[LOUISSSSS]

Originally posted by: Amaroque
In my experience (I ran Prime95 as a DC project for years), round off errors are usually due to an overheating CPU. To a lesser extent, it could also be a RAM problem.

Reinstalling the OS won't make any difference.

no rounding errors include system instability.

 

[LOUISSSSS]

Originally posted by: Amaroque

Originally posted by: Finns14
Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

That's because Prime95 only uses one core. Try running Orthos (a dual core version of Prime95) for testing both CPU's at once.

Prime 95 uses up to 4 cores what are you talkin about?

 

[LOUISSSSS]

Originally posted by: LOUISSSSS
try a complete reformat, reset your bios to default and try prime95 again..

 

[Finns14]

Originally posted by: LOUISSSSS

Originally posted by: Amaroque

Originally posted by: Finns14
Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

That's because Prime95 only uses one core. Try running Orthos (a dual core version of Prime95) for testing both CPU's at once.

Prime 95 uses up to 4 cores what are you talkin about?

Either test I am still receiving the same same error and I feel like the other core isn't being used, although I am not sure if everest is the best program.

 

[Finns14]

Originally posted by: LOUISSSSS

Originally posted by: LOUISSSSS
try a complete reformat, reset your bios to default and try prime95 again..

UGH any other suggestiong before I do this dreaded process

 

[Amaroque]

Originally posted by: LOUISSSSS

Originally posted by: Amaroque
In my experience (I ran Prime95 as a DC project for years), round off errors are usually due to an overheating CPU. To a lesser extent, it could also be a RAM problem.

Reinstalling the OS won't make any difference.

no rounding errors include system instability.

Highly doubtful Windows could create CPU hardware errors such as round off errors...

Straight from the horses mouth: http://www.mersenne.org/faq.htm

You should keep an eye on the results.txt file for any error messages. They can indicate failing memory or an overheating CPU.

 

[Amaroque]

Originally posted by: LOUISSSSS

Originally posted by: Amaroque

Originally posted by: Finns14
Another point of concern when I look in everest when running Prime95 it says Core #2 at 100% but core #1 at 0%?

That's because Prime95 only uses one core. Try running Orthos (a dual core version of Prime95) for testing both CPU's at once.

Prime 95 uses up to 4 cores what are you talkin about?

Correction, the latest version of Prime95 v2414 supports ONE core.

From the horses mouth: http://www.mersenneforum.org/showthread.php?t=7836

 

[LOUISSSSS]

v25.4 http://www.mersenneforum.org/showthread.php?t=8981

 

[Amaroque]

Originally posted by: LOUISSSSS
v25.4 http://www.mersenneforum.org/showthread.php?t=8981

That is the beta version... It's not available for download on the main website.

 

[Finns14]

Ok well with the side of my case off and a room fan blow directly on it the stress test seems to be working better now, I made it to test 3 where as I was failing in about 1-2 mins for test 1. Do you think its because of my mistake with the thermal past or just the fact that I am using a crappy AMD stock HSF?

 

[Amaroque]

Originally posted by: Finns14
Ok well with the side of my case off and a room fan blow directly on it the stress test seems to be working better now, I made it to test 3 where as I was failing in about 1-2 mins for test 1. Do you think its because of my mistake with the thermal past or just the fact that I am using a crappy AMD stock HSF?

Could be your TIM application, or poor case ventilation. However, the stock HSF should be fine if you aren't overclocking.

FYI: Don't test system stability with Alpha, or Beta software as LOUIS suggests. Beta code has bugs that could be mistaken for hardware problems. That's why it is not available for download on the main page.

If you look at the latest version, people are having problems with it that could be interpreted as hardware problems.

THIS IS A PRE-BETA VERSION - NOT FOR EVERYDAY USE

 

[myocardia]

Originally posted by: Finns14
Ok well with the side of my case off and a room fan blow directly on it the stress test seems to be working better now, I made it to test 3 where as I was failing in about 1-2 mins for test 1. Do you think its because of my mistake with the thermal past or just the fact that I am using a crappy AMD stock HSF?

It's because you aren't giving your CPU enough vcore. Skt. 939 CPU's require 1.35v.

 

[LOUISSSSS]

Originally posted by: myocardia

Originally posted by: Finns14
Ok well with the side of my case off and a room fan blow directly on it the stress test seems to be working better now, I made it to test 3 where as I was failing in about 1-2 mins for test 1. Do you think its because of my mistake with the thermal past or just the fact that I am using a crappy AMD stock HSF?

It's because you aren't giving your CPU enough vcore. Skt. 939 CPU's require 1.35v.

i agree. its a vcore problem.

and youre not telling us your load temps while runningprime95 v25.4

 

[Amaroque]

Could be a Vcore problem. X2's are rated 1.3-1.35v? The 4800 in my sig runs at 1.3v

1.35v couldn't hurt though. The only thing is that if there are any cooling issues, that will only mitigate the origional problem.

But yeah, Load temps would be helpful.

Don't run v25.4. It's alpha, and already been superceeded by the alpha v25.5. Just run Orthos (as there isn't any need for four core support here anyway).

 

[ja1484]

Originally posted by: myocardia

Originally posted by: Finns14
Ok well with the side of my case off and a room fan blow directly on it the stress test seems to be working better now, I made it to test 3 where as I was failing in about 1-2 mins for test 1. Do you think its because of my mistake with the thermal past or just the fact that I am using a crappy AMD stock HSF?

It's because you aren't giving your CPU enough vcore. Skt. 939 CPU's require 1.35v.

While that *definitely* could be the case, and the OP should *definitely* look into it, the blanket statement just isn't true. The OC in my sig is actually undervolted from stock, running on 1.30v, and it's 36 hour Orthos stable.

Edited to add: Also, the previous CPU I had in this machine, a 3000+ Winchester, ran @ 2.2Ghz (from 1.8ghz stock) on 1.325v, 24hr Prime stable.

 

[myocardia]

Originally posted by: ja1484
While that *definitely* could be the case, and the OP should *definitely* look into it, the blanket statement just isn't true. The OC in my sig is actually undervolted from stock, running on 1.30v, and it's 36 hour Orthos stable.

Edited to add: Also, the previous CPU I had in this machine, a 3000+ Winchester, ran @ 2.2Ghz (from 1.8ghz stock) on 1.325v, 24hr Prime stable.

Anyone can try undervolting any CPU, but no matter what lies AMD likes to tell, nearly 100% of all Skt. 939 CPU's require more than 1.30v (as evidenced by your 3000 Winchester), yet AMD told all of the motherboard manufacturers that they only require 1.30v, which is why nearly 100% of all Skt. 939 BIOS's default to 1.30v. BTW, 10,000 hours Orthos stable does not mean your system is stable.

 

[Finns14]

Ok so I am running temps of about 90 F on load when I ran prime95 for a time, also when I ran prime95 this was happening http://img201.imageshack.us/my.php?image=problemvb3.png now occasionally the CPU usage would spike to 54% in task manager and would stay close to 100% in everest. I thought this was because of prime95 but I got the same results when trying this with orthos.