Question Overclocking stability question

[mjbn1977]

So, I am in the long process of stability testing my delidded and air cooled (dark rock 3 pro) 8700k to find the Vcore for 5.0Ghz, 4.9Ghz, and 4.8Ghz. My goal is to have some different overclocking profiles for different situations, while planning the 4.9 or 4.8 (depending on temps) to make may 24/7 overclock. I mostly play games and develop games (using Unreal Engine, Visual Studio and Blender). So I do once in a while rendering and building lighting and stuff that is tough on the CPU. I want an overclock that doesn't cause crashes and especially doesn't have micro errors that will eventually corrupt my OS or my Unreal Projects

Here is what I do for stability testing and in that order (thermals were tested with 15min run of prime95 v26.6 small ffts):

1. run 1hr real bench to find out how low I can go with vcore before a get error
2. then run 8 hr real bench if no.1 was stable
3. if no.2 stable rund 2hr Aida64 FPU stress test and 2 hr Aida64 FPU, CPU, CACHE, MEMORY
4. then a few runs of Cinebench R15 and R20
5. one run of Blender benchmark
6. 8hr run of prime95 blend v26.6

Now I noticed something. For example...stress tests 1 - 5 did pass without problems at 5.0Ghz with vcore 1.31 and LLC4. But I had blue screens in test 6 (prime95 blend). I had to go all the way up to 1.35v in order to pass prime 95 blend for 8 hours without blue screen.

Currently I'm doing the stress test for 4.9Ghz and started with 1.27v vcore. I once again performed test 1-5, no problems, no errors, everything passed. But my 8hr run of prime95 blend v26.6 bluescreened again.

First, should I change my testing routine and from now on start with prime95 blend, since it seems to be the limiting factor here?

Also, is this test so much more stressful on the CPU then all the other test?

How worried would you be in terms of stability and micro errors?

By the way...in all described tests my temps are in the 70s. I only start touch 80C at 5.0 @ 1.35v with very intense benchmark runs.

Thanks!

 

[mike_ep45]

Hello, have you seen this stability guide? I use prime95 until I get no errors. I've ran it for 24 hours with no errors, then run it right after a second time got an error.

So I set the sizes to both 8K, FFTs in place, 15 minutes. That usually catches an error in under an hour if it's a CPU voltage problem.
When overclocking RAM I do 800K both, like 75% of RAM, 15 minutes (don't run FFTs in place we want to test RAM).

That might help track down some problems sooner. Also you might want to increase Loadline Calibration Control in your bios, maybe adjust DRAM voltage or PCH voltage (small increments), or lower DRAM speed to slow if the 800k test fails.

 

[mjbn1977]

That might help track down some problems sooner. Also you might want to increase Loadline Calibration Control in your bios, maybe adjust DRAM voltage or PCH voltage (small increments), or lower DRAM speed to slow if the 800k test fails.

I have load line calibration on LLC 4 right now on my MSI board. I seems like it is usually 0.005v to 0.01v higher under full load than the set Vcore. It seems like its kinda stable with no spikes. If I have lower or higher LLC the vcore is much higher or much lower than the set vcore