Cant get i7-4930k above 4.1ghz ? (msi x79A-GD45 plus )

[jamesjgaul2]

Hi,
I was hopping i could get some help or input as im having a ton of difficultiy getting my i7 4930k to be stable at anything above 4.1ghz .

Most recently i tried setting my BIOS settings exactly as "omagundman" (who was able to get 4.5ghz and posted BIOS screen shots), the only difference is im trying at 42x ratio (4.2ghz) and im *NOT* oc'ing my RAM in anyway (left settings / speed for ram totally stock at 1600mhz) also he is at 4.5ghz (45x) , same CPU, Same MB (different BIOS versions, he is running 17.8 im running 17.9). his settings are in this thread:

https://forum-en.msi.com/index.php?topic=182247.0

Is it maybe that my definition of "stable" is harder than most ppl test for? I run p95 torture test at small FFTs (max heat) and i expect it to run 24+ hours with no issues, then i start adding other apps while p95 is running (once it passes 24hours).

however at 42x i cant even have p95 run for more than 5 min without a reboot or blue screen. Users everywhere are running this chip at 4.5ghz + and im stuck at 4.0 / 4.1

any ideas?
thanks

 

[tweakboy]

Weird.. What mobo you using disable speedStep

 

[hunkeelin]

that's cuz ur running the avx -2 instructions.

 

[hunkeelin]

I get bsod 30 second in with my rig if I run p95 with avx-2 XD.

 

[tweakboy]

What you got your CPU clock OCin ?

 

[BonzaiDuck]

There is some common advice currently: Turn off the AVX/AVX2 features of stress-tests. Those instruction-set extensions were not meant to be used in the way that the stress programs hammer them.

I'd like to think that stressing your system as you do -- 24 hours Prime sFFT, followed by "adding running applications" -- is a reasonable idea.

But there are other views on this. First of all, some of these stress-programs are called "power viruses," meaning that they aren't programmed to get the same results in the most efficient and coolest manner. Think about the Dan Brown novel "Digital Fortress."

That's why some of us are leaning more toward such programs as OCCT:CPU (OCCT also has a LinPack test, with option to turn off AVX). Another one worth trying is Intel's Extreme Tuning Utility.

We all used to follow the advice to test with Prime95 sFFT for 12, 18 or 24 hours. But the best stress testing program will find instability much sooner than that, so IMHO -- running Prime that long is unnecessary.

Put it another way (someone may correct me here yet again). I believe that errors from insufficient OC settings are Poisson-distributed. That is, the probability of failure showing insufficient settings is most likely to occur early. Thus, finding a sweet-spot for CPU multiplier and voltage requires finding failure within the first 10 or 20 minutes. Adjusting voltage, you may not see it for a few hours. But if you get beyond 4 hours for one of the more reliable test programs, the probability of failure occurring in the 5th hour is lower. If you make it to 6 hours, the probability of failure in the 7th is lower and so on.

I use Prime95 in a medley of tests as a "prelim stressor" for about 5 hours. The LFFT test also has a capability to determine RAM instability. But ultimately, if you need quick and no less accurate confirmation of stability or instability, 3.5 to 4.5 hours with the OCCT:CPU program will do it.

I've often had occasion to run affinitized LinX (see a thread by IDontCare to explain this), while running Media Center to feed my Home Theater rig at the same time. But if you're going to load up a pile of stuff with that . . that . . that power-virus, I think you're asking a bit much. Most OC guides actually advise turning off the internet connection and Windows Update while stressing.

Also, if you are running sensor-monitoring software or anything that accesses the motherboard VRMs to get real-time voltage and power data, DO NOT use two such programs at the same time. This increases the chances of "false-positive" BSODs, freezes and resets, when the system may be really rock stable.