Z390 Full Performance Boost Around The Clock
Z390 Full Performance Boost Around The Clock
Hello, I previously relied on a fixed voltage and ratio, ensuring my CPU consistently ran at 5Ghz and 1.4v even when idle. My MSI Z390 Gaming Plus supports an adaptive/offset method similar to the dynamic multiplier. For overclocking (i5 9600KF), I set the maximum safe voltage and increased the multiplier by 5, which works well for most Intel CPUs. If the CPU runs smoothly, I keep raising the multiplier until it crashes. I usually play for an hour, running benchmarks like Cinebench or XTU. When crashes occur, I raise the voltage by 0.025v (25mV) and then return to 3.4, 3, 3, 4, 3, 4. Eventually, I may need to add more voltage than necessary, so I prefer a lower clock speed instead of extra voltage. Currently, I achieve stable overclock at 5Ghz 1.4v and 5.1Ghz 1.465v. If I want to maintain 5Ghz and 1.4v with a lower multiplier and voltage while idle, I need to adjust the settings accordingly. I understand that using offset voltage and LLC 4 (flat) keeps the voltage around 1.436v or similar, not exactly 1.4v.
When C states are active, the system should naturally slow down and lower voltage as usual. I’m unsure if adaptive is exclusive to Asus, but adaptive only adjusts voltage when frequency exceeds its limits at default—making it a better choice. Keep in mind that both offset and adaptive rely on auto-voltage, so changes happen based on CPU usage. More power leads to bigger voltage drops and vice versa, which explains the constant fluctuations. Also, the LLC mode that provides stable voltage is actually too high; it hurts regulation significantly, so you should allow a 0.1V drop from idle to full load at least in manual mode.
Activate speedstep, speedshift, C-States and enable adaptive voltage. This is the preferred configuration for my overclocks. Check with HWInfo to confirm adaptive voltage is functioning correctly and isn't delivering excessive voltage.
The concept I find confusing is that higher power equals voltage fluctuations. It seems contradictory—like pushing harder on a car’s gas pedal should reduce fuel use, which doesn’t make sense. So how does increased power relate to voltage drops? Or are we discussing current instead of power? For me, power should reflect CPU usage, like the voltage droop caused by the load. In overclocking, the Flat LLC is recommended.
Vdroop and lower LLC works well because it provides a smoother response, reduces extreme voltage changes, and improves overall stability. It functions similarly to an orifice at the end of a pipe, with the CPU acting as that opening. When the CPU is idle, the opening is narrow, creating higher pressure (voltage) and low flow. Under load, the opening widens, lowering pressure (voltage) and increasing flow.
P equals VI, meaning you can boost power even when voltage decreases by increasing current flow. View the VRM like a water pump—adjusting the nozzle lets you get a focused high-pressure stream or a broad slow flow. Fully opening it still pushes more water overall, though pressure drops. I checked this: your point holds true only for CPUs that can’t accurately read voltage on boards with analog PWM controllers for the VRM, not for high-end Asus systems. In short, your reasoning is correct, but the voltage measurement might be off.
The author doubts whether German documentation covers this topic, but in English it clearly warns that running an LLC to deliver consistent voltage to the CPU at idle and load is problematic due to unseen peaks and dips. He also notes that the voltage displayed differs from the actual value received by the CPU.