Originally Posted by Associated
Ok, I'm onto something... I had to upp VCCIN and SystemAgent and now I am to the point that OCCT can detect an error (after 5-7min) and I don't get BlueScreen. VCCIN is now at 1.872V and SystemAgent is +100offset (0.950V). Error that is detected is not on a Core so I assume it must be memory?
You should really be testing it out with Linux and stressapptest. An easy way to eliminate Vcore and VCCIN assuming you have sufficient cooling is just to push Vcore +0.03V higher and set VCCIN to 1.98V. If you still get errors after that it is most likely VCCSA, RAM related. Only do one thing at a time. Usually when the RAM is stable, the only thing that needs changing is the Core Voltage. I guess you could say RAM > Core > Cache then. I just prefer Core first myself as getting that stable is my primary concern. Linux is worth the hassle to ensure a stable system.
Overclock RAM, test for 1 hour in stressapptest Linux. Turn off PC. Turn on again to ensure training won't affect stability, run stressapptest for 6 hours. Once that's done, RAM is stable, no more needs to be done.
Overclock Core, OCCT for 1 hour is an initial goal. Running it for 3 hours without any errors is the ideal situation.
Overclock Cache, you can use AIDA64 (yuck) or just use the system for non-critical tasks. I find that gaming helps eek out cache instability much faster than AIDA64's cache test, but it varies for others. GTA V seems to be fantastic at causing lockups and freezes with unstable cache.
Cache and RAM work together, RAM instability usually shows up by preventing Windows/Linux from booting, Core shows up as sudden program crashes or BSoDs and Cache shows up as "freezes" or sudden lockups. GPU crashes can also show up as freezes or sudden lockups, so remember one thing at a time
.Edited by Desolutional - 2/14/16 at 6:04am