The SOLVED 5800x is causing issues with your game.
The SOLVED 5800x is causing issues with your game.
Cinebench works fine for stability checks, yet falls short compared to Aida64 or Asus RealBench, which handle a range of instruction sets at different intensity levels with varied RAM and GPU inputs. Handbrake and Prime95 perform even less, focusing on a limited set of instructions without covering enough complexity. In the past, simpler CPUs with fewer cores, no scheduling, and basic features made Prime95 26.6 quite acceptable for stability testing, especially since games were simpler and used similar instruction sets. After the release of BF4, things changed dramatically; modern games now use a much broader range of instructions across different intensities, including even AVX and AVX2 in a Skyrim mod. This explains why you can run 8 hours of Prime95 without errors, open a browser, and boot into BSOD. Stability testing aims to push the CPU to failure, not just verify basic functionality.
I’ve noticed that if Vccio is set too low, my 10900k can behave differently—stable one moment in Prime95, then unstable the next (normally after a reboot). It seems Prime95 isn’t sufficient as a sole verification method. I’ve consistently remained stable after completing full small ffts and an hour of large ffts. This is likely because I’m confident the system is stable before starting Prime95. Errors usually appear after about six hours, and a simple Vccio adjustment can resolve them.
I used TestMem5 with a custom configuration named
Extreme1@ anta777.cfg. Everyone’s approach varies, and I’m not claiming TestMem5 will solve all issues. Aida64 and occt worked well for me, but Prime95 small ffts failed. Even the large fft on Prime95 passed. Games seemed to stabilize once things settled. Understanding how my system reacts to overclocking has helped me achieve the desired settings. The problems I faced were unusual—once instability hit, shutting down or using a stable setting would trigger a BSOD. Fixing Vccio resolved these quirks.
Typically, after completing a full set of small ffts and an hour of large ffts, stability was achieved. The six-hour large fft sequence covered most tests with my settings. This came after two hours of RAM checks; they didn’t catch every flaw but cleared most issues. The remaining hour of large ffts resolved the rest. Currently there are no major concerns, though every overclock still brings up some stability challenges.
Cinebench isn’t suitable for stability checks since it lacks error detection. Running it for extended periods often leads to data corruption and numerous errors. That was my first attempt. Most users rely on their preferred tools—Aida64, Intel Extreme Tuning Utility, occt, TestMem5, MemTest86, and Prime95. I found TestMem5 and Prime95 to be the most effective for me. I performed Prime95 small ffts for both SSE and FM3, ran large ffts on FM3 for six hours with custom parameters, and everything worked fine. Cinebench and Aida64 underperformed; Aida64’s FPU alone is stronger. Only Prime95 can generate over 300 watts and cause crashes during small ffts. Let’s talk about it.