Why does my CPU voltage differ from what BIOS shows, especially with an i7 8700K?
Why does my CPU voltage differ from what BIOS shows, especially with an i7 8700K?
Through several adjustments on my Windows system, I eventually reset it back to the default BIOS settings, but I lost my original configuration. I was attempting to recover to where I left off, yet the voltages remained excessively high regardless of the settings I applied.
I consulted CPU-Z and Corsair Link, which both indicated the same voltage levels as those set in the BIOS. Understanding this voltage is crucial to prevent damaging my computer.
Additionally, after some research on small and minimal FTT tests using prime95, they consistently pushed me to 90-100°C (very quickly), even when I was running standard settings. (I suspect it was closer to 90°C rather than 100, but still a concern.)
I turned everything off very abruptly.
My system uses an H100IV2 liquid cooler. PC Specifications: Intel i7-8700K at 4.5 GHz @1.32 MHz in BIOS (1.36 according to the monitor), previously 5.0 GHz at 1.375. It runs on 1080 Ti, 32GB RAM, Windows 10 (x64). ROG STRIX Z370-H GAMING Mobo.
The BIOS is set without an XMP profile; I use manual settings with adjustments to just the CPU core ratio and core voltage. The XMP profile was nearing 1.4 at the default clock despite being listed as 1.25, but disabling XMP and using manual settings still caused significant discrepancies. Setting the manual voltage below 1.2 would prevent the system from starting, while auto with default settings allows the core clock to run around 1.0V in both BIOS and during testing.
The machine operates stably at that clock speed, but the voltage remains high, and I have limited control over it. This was confirmed using HWinfo64, which showed the voltage exceeding what I had configured in BIOS.
I’m not very confident about ASUS motherboards since I use a Gigabyte board, but you should be able to achieve a much lower vcore for a 4.5GHz overclock.
I have XMP enabled across four dimensions: 8GB x 4 at DDR4 3200, 4.9GHz at 1.278v on an offset vcore...to reach 5GHz feels like too much vcore for me.
It seems you should experiment with your offset to establish a suitable baseline. I found it helpful to adjust the offset carefully, gradually reducing it until performance drops, then increasing slightly. Small FFT tests can really stress the CPU, especially when AVX instructions are involved, causing temperatures to rise sharply. Consider using Prime95 version 26.6 for testing without AVX and setting an AVX offset in the BIOS.
I’ve seen some forum discussions online that might offer useful insights for your motherboard and the 8700K.
View: https://www.reddit.com/r/overclocking/co...h=f363f72f
What is your LLC like? Also, XMP overclocks your RAM but shouldn't influence your CPU speed or voltages. Edit - Prime95 smallFFT (with AVX2) provides one of the most intense tests for your CPU. Many users opt for Realbench 15min stress test for better real-world stability results.
My LLC was running for a few years with this PC and board, now it's at LLC 5 as of last night when I checked the situation myself. (I searched several similar issues and found this one solution common, but it didn't seem to help me.)
I'm taking a moment to review the thread you linked, but it's important to remember that I set it to manual mode instead of offset, and I probably didn't mention this in the original post because I was a bit embarrassed. However, I tried switching it to offset and added a small offset adjustment to manual mode. My temperatures were normal, but the CPU reported high voltages (around 1.5-1.6V) on HWmonitor, so I restarted my PC and switched back to manual mode.
Later update:
Also, in the thread you shared with optimized defaults, could you clarify what the power limit is for the long and short duration packages? I understand what they do, but I'm curious—just to be sure, I don't misunderstand anything, especially since I'm not sure what the 4095 value actually means.
It's really clear how resetting to manual mode works with those vcore values! I had to spend some time getting used to the offset, particularly on the Gigabyte board, which felt a bit more challenging.
Sure, just confirming the details. I'm going to adjust my settings and check how it behaves, making sure it stays stable. (I'm a bit worried because my computer is crucial for my hobby and school, so please give me a chance if I need to ask a few questions.)
Also, regarding the voltage, I'm concerned about what numbers I should expect. I understand 1.4 volts can be high, but if it goes up to 5 and draws more current, I might panic and shut everything down before it fails.
I understand you mentioned the settings you received earlier. You’d like me to test those and let you know what I find. I’m aware that the chip’s reported value often differs from your BIOS setting, and I’m unsure if it’s safe because temperatures aren’t spiking as expected. Also, regarding the TJMax at 100°C—does reaching 90°C mean it won’t fail, or is 100°C the point where it stops functioning?
You need to become familiar with OC concepts. It doesn't mean your approach is incorrect, just incomplete. For most overclocking scenarios, you won't require more than 50-66% of the core voltage, though this varies depending on the BIOS. Excessive LLC is equally problematic as insufficient voltage. Core voltages are always changing based on the load; higher loads cause a drop in Vcore. When the CPU suddenly increases its workload, Vcore drops significantly—a phenomenon known as vdroop. LLC was created to prevent vdroop by applying a small voltage before it reaches unstable levels, and also to reinforce spikes when the load decreases again. Adding too much LLC can lead to excessive voltage spikes combined with higher Vcore, resulting in extreme voltage peaks that can damage the VRM components. On the other hand, insufficient LLC means inadequate voltage support during the drop, causing instability that may only be temporary, allowing the CPU a chance to recover, though eventually leading to shutdown.
Other factors such as phase controls, PLD, ring voltages, VDD, offset, and various ECO settings (like c-states and EIST) should also be considered during overclocking. Simply setting core voltages and boosting Vcore isn't enough; it's not just about adjusting multipliers or voltage levels.
The 100°C threshold isn’t meant to be a hard cutoff for immediate failure. The risk of sudden CPU shutdown due to overheating is far lower than the odds of finding a mouse in Argentina while carrying an elephant in China. This temperature marks a point where the CPU will enter 'oh chit' mode, attempting to reduce clock speeds and voltages to manage heat. While this can occur after reaching 80°C, it’s more likely at 100°C. If cooling efforts fail, the CPU will shut down immediately to prevent permanent damage. It's safer to allow shutdown rather than repeatedly trying to lower temperatures, as some harm may happen during those attempts. The CPU will determine when it has reached its limit.
I recommend thoroughly researching forums like Asus ROG discussions, focusing closely on your specific CPU. The principles of overclocking remain consistent across brands and BIOS versions—what matters is the CPU itself. Avoid copying others' results; each CPU reacts differently.
Prime95 with AVX disabled works well for reaching maximum temperatures. It provides a steady 100% workload similar to gaming, making it ideal for testing limits. For AIO tasks, a half-hour run is sufficient. Stability isn't the priority here.
Asus RealBench is designed for stress-testing all components—CPU, GPU, RAM, storage, PCIe buses, etc.—in extreme conditions. It simulates worst-case scenarios and is best ignored in terms of temperature concerns.
Ultimately, a balanced approach using both Prime95 and RealBench will help you achieve stable performance at optimal temperatures.