Corsair H60i experiences unusual temperature fluctuations.
Corsair H60i experiences unusual temperature fluctuations.
I've owned a 3570K for a while without much concern for overclocking, as game performance was satisfactory. However, over time this device is beginning to show its wear in certain titles. To address this, I opted for the "one-click overclock" feature from my motherboard (Asus P8Z77V-LK). This adjusted the BCLK to 102 MHz, the multiplier to 41, and the RAM speed to 1359MHz (24GB DDR3-1333), achieving a combined speed of 4181 MHz.
To confirm stability, I began using Prime95 and monitored temperatures using Speedfan initially, then HWMONITOR. I observed some unusual patterns: the hottest core would climb to 92°C during about 8–10 minutes of Prime95 use, then quickly drop to 78°C. This cycle has occurred roughly four times now. Have you encountered similar behavior with a Corsair H60i?
The CPU isn’t throttling, as confirmed by CPU-Z. The fan speed readings via PWM consistently fall between 4455 and 4500 RPM, though I’m not entirely convinced. What could be causing this?
I5 3570k equipped with h55 on Asus p8z77-v LE at 4.3GHz 1.08v. P95 26.6 after one hour of testing, minimal FFT readings stay below 70°C.
The issue lies in the OC settings. Not all 3570k units are identical; they come from various silicon batches. Stability at standard clock speeds across all CPUs is only assured by raising the high voltage. Adding a software OC not just increases BCLK but also boosts the voltage, affects the multiplier, and may push current limits. It works, but it comes with trade-offs. Removing the OC brings things back to normal. Then simply increase the multiplier. That’s all you need. You can easily achieve a 4.2GHz setting by adjusting the multiplier; the stock voltage is already quite high.
You should use an Intel CPU, and the best display is RealTemp (compatible with P95). Also stick to the original P95 v26.6 version—new releases often include high AVX instructions that unnecessarily raise temperatures.
Adjusting the voltage clearly made a difference. The auto overclock was at 1.68 volts, and I switched it to +0.005V manually, bringing it down to 1.12v. Temperatures are significantly lower now, though they still fluctuate with peaks and valleys. The readings begin around 68°C, rise to 83°C, then quickly return to 68-70°C. I also tried setting the pump and radiator fan in BIOS to maximum speed, but I'm not certain it worked.
Using the h60 as intended, the setup remains unchanged. The CPU header on these boards supports only PWM signals. Any 3-pin configuration connected to that header (as per Corsair guidelines) will maintain a constant 12V output regardless of the setting percentage. At speeds above 4000 RPM, the fan reaches its maximum capacity.
I might have identified the unusual temperature changes. Using Small FFTs keeps temperatures steady around 79C to 81C. With the "Blend" workload, temperatures vary more widely, between 67C and 83C. It looks like the workload is affecting the cooling, not just the settings or configuration.
I'm still observing higher temperatures than Karadjgne. It seems I can only adjust offsets in the BIOS, so I've set it to +0.005V. The voltage reported by HWMONITOR is about 1.211v, while CPU-Z shows 1.104v. Which reading should I trust?
The stock voltages remain elevated. Unless you're aiming for a drastic jump to 4.4, you won't need to exceed the standard levels. Switching the offset mode to negative is possible; however, since CPUs differ only slightly near stock speeds, a manual OC using a negative multiplier works well. Use Asus RealBench for consistent stability checks. P95 should be targeted at maximum temperatures. Conduct a few tests with Asus if successful—no errors—then gradually reduce the offset until issues appear, then increase it by one level. This will stabilize your base performance. If you plan to boost speed, test at the same voltage; if errors occur, increment the multiplier and retry. Continue this process up to around 4.3GHz. After reaching that point, further tuning usually demands more varied parameters such as a 25% increase in LLC settings. Ensure all eco options are turned off. You can maintain turbo and speedstep, even pushing your maximum OC, then adjust the turbo setting to align before lowering the OC slightly. Perform a P95 test using a small FFT for consistent load conditions, unlike blend which causes fluctuating loads and RAM stress.
Both tools provide accurate readings: CPU-Z displays the actual CPU voltage (VID), while HWMonitor measures vcore—the voltage supplied to the CPU's VRM. Typically, vcore values are reported as voltage, whereas VID is often misunderstood.
Here’s a summary of my configuration:
AiTweaker: xmp, 100.0, enabled, manual, enabled, 43, auto, disabled, offset mode, -, auto, auto, auto, auto, auto, disabled.
CPU Ratio: all auto, enabled.
Digi: auto, auto, disabled, optimized, 100%, 125, auto, 100% auto, optimized.
These settings align closely with the default factory values. My CPU voltage fluctuates between 1.088 and 1.096 in the BIOS.
I thought about posting an update. My BIOS differs from yours—like having fewer VRM options. The highest stable overclock I can reach is 4.1GHz at 1.15v. I managed 4.2GHz with Prime95 without issues, but videos would crash at that speed. I lowered it to 4.1 and everything stabilized. Temperatures stayed around 85°C with the same workload. There’s also a mismatch between the voltage I set in BIOS and what CPU-Z and HWINFO64 display. In BIOS I set it to 1.08v, but the other tools show higher values. Any advice?
the settings defined in bios establish the standard voltage for base clock rates. additional parameters like llc, offset, turbo can and will adjust it slightly when necessary.