9700k overheats beyond 1.210 voltage
9700k overheats beyond 1.210 voltage
Hello,
I'm preparing a 9700K overclock aimed at both gaming and game development in Unreal Engine. To simulate real-world usage, I've incorporated AVX in my stress tests to reflect its likely role in UE4. During an OCCT preliminary test, I couldn't exceed 1.210 volts without overheating two cores between 85-95°C. So, I'm currently sticking to a 1.210 volt 4.4GHz all-core overclock. Here are my specifications and test details:
Specifications:
- Intel i7 9700K @ 4.7 GHz (testing started at 5 GHz)
- Asus z390-P motherboard
- G.Skill Ripjaws V-Series 32GB (2x16GB) at 3200 MHz
- Noctua NH-D15 (single fan, research suggests it's sufficient for my setup)
- Seasonic Focus 750W PSU, 80 Plus Platinum
Current settings and test parameters:
- RAM XMP II Profile (3200 MHz at 16-18-18-38), no adjustments beyond auto
- LoadLine Calibration: 6 (previous attempts at 5 didn't reach the required threshold)
- Asus MultiCore enhancement: Disabled (appears to restrict CPU behavior unintentionally)
- SVID Behaviour: Normal
- Sync All Cores
- AVX Offset: 0
- CPU Cache: Auto
- Core Voltage: 1.320
- VCCIO: 1.1
- CPU System Agent Voltage: 1.1
- Standby Voltage: Auto
- CPU Current Limit: 140% (used to confirm max at 170 I believe)
- Settings for 4.4 GHz: stable and cool as of step 5
Recent HWMonitor readings from around 5 seconds into the OCCT test (details on testing method):
Please disregard the System Agent voltage reading, as HWMonitor is not consistently reliable in this context.
Ideally, I'm following these stress tests to anticipate AVX workloads while OCCT operates at its highest with AVX enabled. This should help determine if the configuration holds up.
If you need further guidance, feel free to ask. Thank you for your time.
Hello there, fresh face! I'm curious about which OCCT version you're using. Maybe check if changing the cooler setup impacts your experience, as some reviewers mention mount pressure and temperature effects. Also, could you share your typical ambient room temperature? Also, your post looks very detailed for someone just starting out here.
Hello there, fresh face! I'm curious about which OCCT version you're using. Maybe check if changing the cooler setup impacts your experience, as some reviewers mention mount pressure and temperature effects. Also, could you share your typical ambient room temperature? Also, thank you for taking the time to post—your effort is impressive for someone just starting out!
Your cooling is inadequate for a 5.00Ghz Overclock. A H110i AIO or equive would fix that if it will fit your case?
I would apply a negative offset of -2 voltage on the two cores that are overheating. Core boost can raise temps on those especially when using AVX apps
LLC should be set to 7 for 5.00GHz to prevent droop.
And there's more.
Here is the rewritten version at the same length and structure:
I started by re-seating the cooler. I made a mistake with the thermal paste initially, but it turned out okay after some stress testing beforehand. Following Noctua's instructions should have helped.
As for the heatsink, it felt tight, so I tightened the screws just to be safe. Did you mean the four screws that secure the mount? I might have missed reapplying the thermal paste when I removed it. I’m pretty sure I did it right this time.
Maintaining room temperatures around 70-72°C and no humidity (west coast) worked well. Thanks for the advice! From what I’ve learned, the Noctua DH-15 air cooler is top-notch, outperforming water coolers and built for full 9700K overclocking. Did you mean this model specifically? Also, LLC cooling has been tricky for me—I isolated it and ran tests at level 6, which showed much higher temperatures. My budget ASUS motherboard (around $150) doesn’t accurately sense voltages or CPU temps. I suspect the issue is due to its limited accuracy, possibly because of voltage calibration issues. The lower setting (5) seems better since it compensates for sensor errors. I’ve tried LLC up to 6 but still overheated. Buildzoid’s video helped explain this. My board probably can’t handle 7 safely.
Regarding the voltage offset, you’re right—it might be that my cheap motherboard can’t fine-tune voltages so precisely. Maybe adjusting the core frequency by -2 is necessary. It seems my ASUS board struggles with accurate readings, possibly due to incorrect sensor calibration. I’ve tested up to LLC 6 without success.
I should also note that I recently achieved 1.305 volts for a 4.7 GHz overclock, even under stress testing for 15 minutes. For AVX tests at 90°C, the max temps were still high at 1.305 V. I found that leaving Standby Voltage and VCCIO auto increases cooling, but it doesn’t help with AVX. I tried a 4.9 GHz test with -2 AVX offset, but it was unstable.
In short, to get cooler performance, I’d need to manually adjust the voltages for AI Tweaker settings—most guides suggest skipping that. I’m not very familiar with standard i7 voltages here, but it seems my Asus Z390-P board isn’t capable of fine-tuning its own voltage, which is why other manual overclock attempts can still fail.
Thanks for the insights!
Recent adjustments have been made based on valuable feedback. Several indicators suggest a mismatch between RAM and my motherboard, using an inexpensive Asus z390-P while attempting to operate DDR4 dual channel RAM at the manufacturer's XMP profile of 3200 Mhz (XMP II, not XMP I).
The positive outcome is that I can now sustain core voltages up to 1.365V without overheating during full-load tests with AVX enabled on OCCT (limited runs). On the hottest cores (85-87°C), I maintain temperatures within the 80°C range, thanks to specific tweaks since my last update. These changes were implemented individually, not in combination, to clearly identify their impact.
Key modifications:
- LLC setting set to 4 (significantly improved temps; LLC 5 is unsuitable beyond 1.300V)
- Intel TVB disabled (minor effect on voltage stability, per online reports)
- Re-mounted and re-pasted nh-1 onto the heatsink (mount backplate possibly misaligned before; no temperature change observed)
- Adjusted CPU fan direction on the Noctua fan to direct airflow toward the VRM (VRM has a metal cover; likely already optimal)
- AVX offset reduced to 3 (minimal temperature impact, possibly 1-2°C max)
- Increased clock speed to 4.9 GHz after stability tests with no BSOD crashes
- Updated cache auto at 4300 MHz – is this a concern?
- Possible voltage offset from Loadline Calibration; BIOS shows 1.350 Vcore vs expected 1.365 Vcore
- DRAM voltage at 1.350V under XMP II indicates a warning (yellow in BIOS)
- Core frequency fluctuates between 4900–4600 MHz on non-AVX tests with Sync All Cores enabled
- MEMORY reading shows instability during FTT tests, but kernel32 logs suggest voltage may be involved
Regarding CPU cache operation at 4300 MHz: it appears normal. For game engine work, higher speeds (e.g., 3200 MHz) are preferred despite potential stability trade-offs.
Further concerns:
- Voltage offset from Loadline Calibration could affect stability; BIOS warning persists
- Core frequency instability may impact long-term performance in demanding workloads
- Memory error tests remain stable under stress, but deeper diagnostics recommended
Following adjustments appears to have addressed the main concerns regarding the System Agent and VCCIO voltages. All other parameters were verified under the original problem conditions, which explains the improved temperatures as well as the inclusion of standard voltages previously set on Auto, as mentioned by Intel (PCH Core, Standby voltage).
I’m leaning toward repositioning the heatsink as the optimal solution. There are indications that the previous screw tension on the heatsink mount was minimal, which could have contributed to uneven pressure across the CPU. This might help slightly without drastically affecting temperatures.
It’s worth considering whether anyone using this fan should secure the motherboard properly—especially if the case allows mounting—and ensure the mount is vertical when screwed in, avoiding any gravitational effects that could cause uneven tightening.
The main issue seems to be a lack of sufficient voltage for AVX stress tests, particularly with XMP RAM at 3200 MHz or higher, leading to instability and crashes. I’m not fully satisfied with the current overclock yet, but temperatures are now manageable.
For an eight-hour Realbench test reaching up to 75°C core temperature, this appears to resolve the primary concern. Perhaps investing in a better motherboard next time would be wise instead of sticking with the budget Asus option.