9700k stable at 1.278v, unstable at 1.341v under load
9700k stable at 1.278v, unstable at 1.341v under load
Hi everyone,
I’m looking for some guidance. I have a 5.0ghz O.C on my 9700k at 1.345v with LLC set to 5. Everything seems stable. During a small FFT test (avx disabled), voltage dropped to 1.278, which is unstable since my LLC is near the medium level. I considered increasing the vcore to 1.3v and raising LLC to 7 for testing, but it caused instability and a BSOD after under five minutes.
I’m wondering why this happened and what changes might help. Should I stick with LLC5 and 1.345v? Do you think a lower vcore with a higher LLC is better than the opposite? Any advice would be appreciated. Thanks!
The default settings in most monitoring tools function adequately without changes. Vdroop is straightforward to understand, particularly with tools that include graphs. To prevent unexpected Vcore variations, it's important to apply a consistent 100% workload, like P95 v29.8 Small FFT's (with AVX disabled). The contrast between the BIOS-specified Vcore and the actual load Vcore seen in a Windows utility (such as CPU-Z) will become clear.
Overclocking is constrained by two main aspects; voltage and temperature. No two processors are the same; each has its own tolerance for voltage, thermal response, and overclocking capability. Many processors operate most efficiently within a specific "sweet spot" where they reach their peak overclock performance at optimal Vcore and core temperatures. For 14 nanometer chips, this usually means below 1.4 volts and under 80°C.
Pushing a processor beyond its ideal parameters can lead to instability, no matter how much Vcore is increased or voltages are adjusted for components like Ring (Uncore), System Agent (SA), Integrated Memory Controller (IMC), Input/Output (I/O), Phase Lock Loop (PLL) and Load Line Calibration (LLC).
LLC is intended to counteract "Vdroop," which occurs when heavy loads cause Vcore to drop significantly. Properly configured LLC ensures Vcore stays stable—within about 25 millivolts—when switching from light to heavy loads, avoiding the "drooping" that can trigger BSOD crashes. Conversely, if LLC is set too high, heavy loads may push Vcore beyond its BIOS limit, potentially harming the processor. The goal is therefore to keep Vcore steady across both light and heavy usage, allowing a slight drop rather than a spike.
For instance, if you set Vore to 1.300, you should adjust LLC so that it doesn’t sag or exceed a 25 millivolt increase, which would equate to 1.275. It’s important to note that BIOS voltage settings often use 0.005 increments, but in practice, values change by 0.008. Thus, a 25 millivolt drop translates to a 3-step adjustment (0.008 increments). On a motherboard with 12 phase VRMs and precise regulation, the best achievable stability is around 16 millivolts.
In comparison, 4th generation Haswell and Devil's Canyon processors performed well here, thanks to Fully Integrated Voltage Regulators (FIVR), making Vdroop nearly nonexistent—meaning your settings directly determine performance.
I concur with CompuTronix, it took me a while to locate the right settings for my 9700k. Check out the image with the results...
The default settings in most monitoring tools function adequately without changes. Vdroop is straightforward to understand, particularly with tools that include graphs. To prevent unexpected Vcore variations, it's important to apply a consistent 100% workload, like P95 v29.8 Small FFT's (with AVX disabled). The contrast between the BIOS-specified Vcore and the actual load Vcore seen in a Windows utility (such as CPU-Z) will become clear.
The vdroop remains a fixed value. I initially believed it was just an initial "inaccuracy" that corrected to the desired voltage once the load stayed steady for a while.
Vdroop remains a steady "condition" based on the LLC configuration, and it won't normalize if the LLC is configured too low. It's likely referring to the initial load application phase, which can cause an overshoot spike when the LLC is set too high. An oscilloscope would be required to capture that short-lived event.