Prime95 CPU test error. Do I change voltage?

[TYTSoldier]

Shortly after starting Prime95 test using blend on my i5-3570k at 3.4Ghz, Workers 1 and 2 have passed several tests, but almost immediately after test 4 and test 9, Workers 3 and 4 had this fail message: fatal error rounding was 0.5 expected less than 0.4. I bought this CPU used several months ago and hadn't tested it until now, but I've been testing a lot of my hardware due to frequent crashes in a certain game (Crysis 3). I did 1 pass on the RAM stress test and there were 0 errors, but the CPU had the errors mentioned above. I've been looking around for solutions to the problem, and it sounds like I need to increase voltage? If that's correct, how do I go about doing that? (I have a Z77 Extreme4 and my 16GB RAM 1600 is on XMP 1.3).

I have a cooler master 212 plus as my cpu fan.
I also have Hardware monitor on and it says that my fan RPM is 1854 min 1877 max and auxfanin1 is 1157 min 2419 max during this test.

Also I have an HX 750 power supply and EVGA GTX 780 Superclocked ACX

 

[cemerian]

yes increase voltage, if it fails in prime then its not stable, increase voltages by the smallest increment possible at time and test it like that, as soon as one core failed you have no more use of running prime, once a core fails restart go into setting and increase voltage a bit

 

[TYTSoldier]

yes increase voltage, if it fails in prime then its not stable, increase voltages by the smallest increment possible at time and test it like that, as soon as one core failed you have no more use of running prime, once a core fails restart go into setting and increase voltage a bit

These are my voltage numbers:

CPU Voltage: Auto
CPU Load-Line Calibration: Level 5
IGPU Voltage: Auto
IGPU Load-Line Calibration: Level 5
DRAM Voltage 1.350V
VTT Voltage 1.076V
PCH Voltage: 1.059V
CPU PLL Voltage: 1.832V
VCCSA Voltage 0.925V

Which ones do I change and how much would you recommend?

 

[cemerian]

cpu voltage but looking at the settings and your cpu speed i would suggest lading fail safe defaults and then checking if prime is stable first, if its not stable increase the cpu voltage by the smallest amount possible at a time

 

[TYTSoldier]

yes increase voltage, if it fails in prime then its not stable, increase voltages by the smallest increment possible at time and test it like that, as soon as one core failed you have no more use of running prime, once a core fails restart go into setting and increase voltage a bit

It turns out that increasing the cpu voltage didnt do anything at various values. I tried changing the ram voltage from 1.35 to 1.5 and now it's made several passes without error. I wonder if that's what caused my crashes in Crysis 3. I guess the 1.35V low profile didn't work as advertised.

 

[cemerian]

well i have heard of some issues with that voltage rams, wich tests were you running when it crashed in prime, small fft, large fft or blend?, if its large then try using small fft test wont stress your ram at all, after or before you can run memtest86 to check you ram for errors and stability, if it doesnt crash in small fft then leave cpu voltage at default and change ram to 1.5 volt mode

 

[TYTSoldier]

well i have heard of some issues with that voltage rams, wich tests were you running when it crashed in prime, small fft, large fft or blend?, if its large then try using small fft test wont stress your ram at all, after or before you can run memtest86 to check you ram for errors and stability, if it doesnt crash in small fft then leave cpu voltage at default and change ram to 1.5 volt mode

I was running blend. And before I ran prime, I did memtest86 and made 1 pass with the RAM without error.

 

[cemerian]

ok, but are you stable now in prime aswell if so then you should be good?

 

[TYTSoldier]

ok, but are you stable now in prime aswell if so then you should be good?

Yes, I am.