New SKYLAKE PC - Prime95: Rounding was 7.756809404e+013, expected less than 0.4

Toggle sidebar Toggle sidebar
R

robertjan88

Reputable
Sep 27, 2015
19
0
4,510
Dear all

Last Monday I received my new gaming pc from the Netherlands (I live in Luxembourg) and I tried Prime95 (64bit) Blend test (with the default 8 workers), to see if everything was stable.

However after 2 hours one of the workers stopped with the following message: "FATAL ERROR: Rounding was 7.756809404e+013, expected less than 0.4"

The other 7 workers went on fine for another 10 hours when I stopped them myself.

Less than 5 hours later I runned another 14 hour+ Prime95 Blend session. This time with the front door of the Define R5 case open. No errors at all!

I also runned a stresstest for my SSD and the HCI memtest for 500% using 8 threats to max out the CPU as well. No errors either.

How is this possible? Was there just a small error with the test or with another software interfering? Or wasn't there enough air in the case? The R5 has small intake holes on the sides of the door which I assumed was enough although this might not have been the case for the 2x140mm fromt fans (in addition to the one in the bottom) during the stress test.

Also, if I would run 2 more 14+ hour Prime95 blend tests and a Memtest86 and this wouldn't return any issues, would it be safe to assume there is nothing wrong or it was just a minor heating issue?

My specs:
Intel Skylake 6700k @stock 4GHZ
Mugen-4 PCGH tower cooler (with 2x120 fan)
Corsair DDR4 LPX 2666 16GB (4x4GB)
Msi Z170A gaming M5
Msi GTX 980 ti Gaming 6G
Samsung 850 Pro 256
Case: Fractal Define R5 with 3 intake (2x front and 1 bottom) and 1 outtake fans (back)
Corsair RM750i 750W 80 plus gold
Windows 10 PRO 64bit

Nothing is overclocked.

Thanks a million!!
 
Solution
jbseven


First make sure your RAM frequency, timings and voltage are set manually in bios. Then retest with prime while using hwinfo to monitor temps. if it passes, nothing to worry about. The temps will tell you better if you need to improve your fan config by adding more fans.
Sort by date Sort by votes
jbseven

jbseven

Distinguished
Dec 2, 2011
646
0
19,160


First make sure your RAM frequency, timings and voltage are set manually in bios. Then retest with prime while using hwinfo to monitor temps. if it passes, nothing to worry about. The temps will tell you better if you need to improve your fan config by adding more fans.
 
Upvote 0 Downvote
Solution
R

robertjan88

Reputable
Sep 27, 2015
19
0
4,510


Thanks for the replies. Runned another instance of Prime95 for over 12 hours and again no issues. Will run Memtest86 this evening and then Aida64 to finish. In regards to Memtest86, is it possible to run all 13 tests in 1 run for all 4x4Gb I have? Since HCI memtest (runned this 5 passes/500%) gave no errors Its just a double check so no need to have this done stick by stick for now
 
Upvote 0 Downvote
jbseven

jbseven

Distinguished
Dec 2, 2011
646
0
19,160
Did you change the settings manually in bios?

With that much testing, it's unlikely that you have any issues.

If you insist on more testing, it'd have to be 8 hours per single stick of ram with memtest86 (since if there is a problem it is very subtle). HCI only tests available ram so while it's good enough for oc testing, it's not the best choice for faulty RAM imho.

Honestly, I'd just write it off as a freak occurence. Although I am interested in max temp readings from hwinfo.
 
Upvote 1 Downvote
R

robertjan88

Reputable
Sep 27, 2015
19
0
4,510
Thanks for the feedback.

Does HW temp once installed also give me an overview of the temps in the past?

Didnt change the memory to manual yet since the timings seem to be right so far.

To sum up, during the first Prime test woth closed case door one of the workers stopped with the strange message in the title. Since then I runned 2 more 12+ houre Prime sessions and an HCI session woth the door open and none gave an error. Btw I read that the Hashwell and later series have problems with the new Prime tests so it was probably just an error from the software or a slight overheating since there wasnt enough air with the case closed.

To run memtest86 on each stick I hope I dont need to remove the others and can just set this in the program?

In addition it will ask me to set an option in the CPU selection. Do I choose single, parallel, round robin or sequential? When single, do I fill the number of cores, threats or just 1 for core no 1?

When this is fine I will stop testing though and enjoy the pc.

Thanks a million!

 
Upvote 0 Downvote
jbseven

jbseven

Distinguished
Dec 2, 2011
646
0
19,160
You will want to double check that your mb has not assigned the wrong frequency, timings or voltage to ram (with cpu-z for example).

Hwinfo does not show values outside the time it was running.

It is better to run with 1 stick at a time since the error you are trying to detect is so subtle. I believe running with multiple sticks can also cause false positives but I'm no memtest expert.

As soon as you boot into memtest, go to settings to bring up the menu and test selection to choose what tests to run.
 
Upvote 1 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom