Sign in with
Sign up | Sign in
Your question

Rounding Off Error during Prime95

Last response: in Overclocking
Share
January 15, 2011 5:13:57 PM

I get an occasional Rounding Off error on Prime95, specifically "Expected 0.5, received 0.4". What is this and what is it a result of?

Other than that everything runs fine. Never crashes or BSODs.

More about : rounding error prime95

a b à CPUs
a c 125 K Overclocking
January 15, 2011 5:18:03 PM

Your overclock isn't completely stable, but most likely very near.

How long does prime run before this error occurs? and what are the temps during prime?
January 15, 2011 5:24:38 PM

Temps are up to 43C and it occurs after about 15 minutes.
Related resources
a b à CPUs
a c 125 K Overclocking
January 15, 2011 5:28:25 PM

Try increasing the Vcore or backing off the overclock slighty to regain stability.
January 15, 2011 5:30:26 PM

I have only increased the VTT cpu voltage. Have never touched Vcore. I will try that and let you know. Thanks.
January 15, 2011 5:31:21 PM

P.S. Backing off an overclock is like not finishing a bowl of ice cream. I just can't do it!
a b à CPUs
a c 125 K Overclocking
January 15, 2011 5:41:28 PM

Let us know how you get on with manually settings the Vcore, And I'm the same espically if its Ben and Jerry's ice cream :) 
a c 172 à CPUs
a c 197 K Overclocking
January 17, 2011 2:35:39 PM

ahthurungnone said:
I get an occasional Rounding Off error on Prime95, specifically "Expected 0.5, received 0.4". What is this and what is it a result of?

It's a calculation error. If you are doing the small fft's test, something is a little wrong with your CPU settings. If it's large fft's or the blend test, your problem is likely the memory settings. In that case, try increasing the RAM voltage to a maximum of 1.65 volts or relaxing the memory timings.
July 18, 2012 6:19:26 PM

Hey guys,
I have the same problem but my setup is most likely different. For one, my relevant specs:
OS: WIN7x64SP2
Motherboard: Gigabyte GA-970A-D3 (Socket M2)
CPU: AMD FX(tm)-8150 Eight-Core Processor factory@3624.3 MHz
RAM: Three (3) Kingston KHX1600C9D3/4GX PC3-10700 (667 MHz) 4GB each and one (1) Hyundai Electronics HMT351U6BFR8C-H9 PC3-10700 (667 MHz) 4GB for a total of 16GB. I have removed the mismatched RAM stick as part of a troubleshooting process due to gross system instability; three or four bsod failures in every ten hour period for a few weeks, for reason of a clock interrupt not received on secondary processor.

Please note that I am not overclocking! I tried to overclock when I first obtained custody of machine and could not achieve reliable operation. I'm not a total moron - I have successfully overclocked an oldish Abit IP35PRO XE board with an Intel Core2 Quad Q9450 cpu by 550 MHz to run windows server 2008 r2 with Hyper-V, which then runs one CentOS 6.2 and two winserver08r2 virtual machines. However, with this AMD system I had a terrible run and haven't been able to dedicate more time for study in over a month. It's been running at default settings since.
Yesterday I finally decided to take out the mismatched ram stick to see if it would improve anything. The system now shows 10GB RAM with only 8GB usable - I probably need to adjust which slots are populated or maybe the board may only have either 2 sticks or 4 sticks. Not too worried about this for now.
Okay so on to the reason I'm posting here!
Last night I ran a Prime95 primality test using all 8 cores, and approximately 6 hours later in the logs it reads:
ERROR: ROUND OFF (0.5) > 0.40
and then:
Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
Confidence in final result is fair.

Then every ten minutes or so I see the ERROR: ROUND OFF (0.5) > 0.40
followed by:
Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
Confidence in final result is fair.

This continues for another five hours or so until I get back to my office and stop the test.

Does anyone have any idea what is causing this? I'd love to see factual answers and I'll even take some educated guesses!
Thanks in advance
~Ethan
!