9700K Overclock
9700K Overclock
LLC and offest voltages add voltage on top of the current settings. LLC (Load Line Calibration) helps avoid vdroop-induced instability. But the timing isn't ideal. If the CPU shuts down, higher voltage might still appear for short times or during the time the load is greater. Adjustments for components like AVX will also add extra voltage when those instructions are active, and similarly when they're not running.
In your most recent suggestion, it recommends more LLC or raising the base core voltage, provided temperatures permit.
Setting a negative offset manages temperature and power consumption. It's common to lower the CPU core clock during AVX execution as well. Given the increased load, this trade-off is justified for improved AVX performance.
I've tried various settings and wasn't satisfied with auto vcore, possibly because it wasn't optimized. Here are the outcomes I got:
- All cores set to 49x multiplier
- Load Line Calibration (LLC) at 5
- Vcore manually at 1.310
- AVX disabled
- Hottest core reached 95 using IBT stress test
Second configuration:
- All cores to 50x multiplier
- LLC at 5
- Vcore manually at 1.310
- AVX disabled
- Hottest core 93 with IBT stress test
I'm not sure if a higher clock speed is better with AVX -2. When I ran my race simulator, the average clock was 4.82, but under AVX instructions it dropped to 4.800... so my stable 4.9 with AVX 0 seems more consistent and better.
This would require extensive testing to evaluate. The ambient temperature was gradually increased during gameplay.
During gaming, the temperature reaches 80 degrees at full capacity. For this setup:
I use avx -1.
All cores are synchronized with a 50x multiplier.
Load Line Calibration set to 5.
Vcore (manual): 1.310
AVX -1 is active.
The hottest core was 95 during the IBT stress test, which was quite intense. I also ran a prime95 blend test for two hours and observed a maximum temperature of 82 degrees.
Eximo, just another point... initially, thank you for sharing this. I’m not sure if I’m correct, but in my opinion, the basic small and large FFT operations are risky and don’t align well with our processor requirements. It seems to be a tool made by people who enjoy pushing limits, regardless of practical use. I’m referring to it because I experienced rounding errors during a small FFT and the processor actually failed (100 degrees in under a minute at 5.0Ghz AVX-1). After reverting to the default settings, the results remained consistent—only without any noticeable rounding issues over three minutes. I have a well-cooled system (deep water cooler, high-quality paste, robust case), a powerful CPU without overheating protection, and sufficient power supply. The stock configuration performed reliably, whereas prime95 exceeded expectations. In my view, this software should be avoided in discussions about stability—it’s far beyond typical user needs, which are mainly for gaming or personal tasks. It poses a real risk, especially since many users have budget components that easily exceed their limits.