Checking an old i7-6700k: seeking tips to reach 4.5GHz
Checking an old i7-6700k: seeking tips to reach 4.5GHz
I'm working with an older system built mainly around hardware from about five years ago, and I'm finally starting to optimize it. My goal is to squeeze a bit more performance out of the existing components without needing a full upgrade. However, I'm encountering some stability problems and overheating that don't match what I've seen for this processor model. I'm hoping you can assist in boosting the speed while keeping things cool and reliable.
Here are my specifications:
i7-6700k @ 4.00GHz (base)
Asus Z170-A motherboard
2x8GB Crucial Ballistix Sport 2666 MHz CL16 DRAM
Noctua NH-U12A cooler
Phanteks P400A-Digital case
GIGABYTE RTX 2070 Gaming OC 8G
The case and cooler are brand new, bought last week because my old setup was poorly configured and made quite noisy. The graphics card was added after I experienced a failure with my 980Ti in late 2018.
My initial attempts at overclocking turned out surprisingly effective. By just adjusting the multiplier and leaving other settings unchanged, I could easily push the clock to 4.4Ghz. Running Prime95 stress tests confirmed this, with core temperatures staying below 75°C. I then increased to 4.5GHz, which also performed well, though temperatures rose into the high 70s. Trying 4.7Ghz was feasible, but it reached a point where my Core 0 temperature hit around 95°C—clearly too high. So I reverted to 4.7GHz.
To lower temperatures further and try again at 4.8Ghz, I read about proper OC techniques for i7-6700K and discovered a mistake: I had left the motherboard's automatic VCore enabled. This likely caused the overheating because the board was trying too hard to keep stability. I corrected it by following a guide from TweakTown, especially the flowchart provided.
Despite careful adjustments, I still faced serious stability challenges even at the recommended 4.5Ghz with 1.3v Vcore. Prime95 would trigger a "HARDWARE ERROR" and IntelBurnTest would freeze. After several weekend experiments and new stress tests, I ended up settling on a 4.5Ghz configuration at 1.34v Vcore. This setup could handle 20 minutes of Prime95 (AVX disabled) and 15 minutes of RealBench stress testing without exceeding 85°C on the hottest core—Core 0 is typically 3–4°C warmer than the others. Higher multipliers would cause instability, and higher voltages would push temperatures beyond safe limits.
Could this be a case of the hardware being too old for these demands? Or are there other ways I can improve performance without risking damage? Maybe improving thermal paste would help, though my cooler came with NT-H1 paste, so I used that instead.
Any advice would be greatly appreciated!
Apologies for the extended wait. I've been quite occupied.
I successfully restored the BIOS to its original configuration (version 3802). I also noticed that turning off ASUS MultiCore Enhancement was necessary to achieve genuine stock temperature readings, as this setting is enabled by default and causes excessive overclocking and voltage adjustments on its own.
With the CPU correctly set to run at its maximum of 4.0GHz and a stock vCore of 1.296v, I observed idle temperatures in the mid-20s and load temperatures (Prime95 Small FFTs for 10 minutes) around the mid-60s. HWINFO logged one 69C reading, though the maximum core temperature seldom exceeded 66C during observation.
Currently, I'm executing Prime95 Small FFTs using these BIOS settings and a multiplier of 44, with vCore manually adjusted to 1.300. After about 12 minutes, temperatures have stabilized in the low 70s, except for Core 1, which averages around 65C.
EDIT 1: The Small FFT stress test performed well, so I plan to begin a one-hour RealBench stress test and then unwind.
EDIT 2: I stayed up late to complete the RealBench test at 4.4Ghz (44x100.0) with vCore at 1.300. It passed, maintaining temperatures in the high 60s and low 70s, depending on the core. However, it briefly reached 79C a few times and hit 80C at least once.
The occurrence of 80°C at 1.300v is making me reassess my thermal paste application. I’ll reapply it tomorrow after work.
For now, I’ll rest.
I repositioned my CPU cooler and noticed several issues. It seems I applied too little thermal paste, leaving almost no material in the middle of the IHS after removal. The surface was nearly transparent, making most of the text easy to read. This suggests I may have over-tightened the cooler, pushing all the paste to the edges. For this next attempt, I added a bit more paste (the Noctua formula doesn’t have a sharp nozzle, just a flat hole) and tightened it less firmly. The outcome has been encouraging. I observed a 3-4°C reduction in average temperatures during Prime95 Small FFTs—highs in the 60s instead of low 70s, with a peak of 77 versus 79—and RealBench also reached 77, averaging high in the 60s. I’m currently testing a 4.5GHz model and running RealBench. After fifteen minutes, temperatures are consistently in the high 60s, with only a brief spike at 80 on Core 0. This is still the highest temperature under load, though Core 1 remains significantly hotter when idle. Surprisingly, this 4.5GHz unit is stable at stock vCore of 1.300v. In earlier trials at 4.5GHz, I struggled to boot with a 45 multiplier at that voltage; reaching 1.340v was necessary for stability during stress tests. Now I’m stable at 1.300v, which is great.
I've abandoned 4.6GHz because it doesn't stay stable at any voltage below 1.36, and it's also too warm. Now I'm focusing on the 8-hour tests for 4.5GHz at 1.310v, which looks promising. It seems I might have missed out on the silicon opportunity or my CPU isn't up to the task for these high-end overclocking attempts. Still, I'm proud of myself for trying everything possible and gaining valuable experience. I managed to get a 12.5% OC with temperatures low enough for my fans to stay quiet—definitely a win. Plus, I picked up some useful knowledge for future projects!
It's worth noting I wasn't actually running at stock voltage (1.300v). I was using "Auto" mode, which caused my motherboard to increase the voltage automatically to keep stability during higher clock speeds.
Since you reopened this discussion, I wanted to share more about my overclocking experiences.
I removed the CPU using the razor method and achieved a good temperature reduction of 10-12°C. This motivated me to try again at 4.6 GHz. I found it stable at 1.400v, surviving overnight tests with RealBench and Prime95 Blend while keeping core temperatures under 85°C and averages in the mid-70s. However, during a gaming session, things failed after three BSODs in a single day of Prey play. I had to revert back to 4.5 GHz at 1.34v. After another long session, it's now working well, so I'm planning to keep going until I upgrade.
The need for more than a .06v increase just to maintain partial stability at an extra 100 MHz, especially with a CPU that others often push up by 200 MHz beyond what I could achieve even after delidding, suggests I definitely missed out on the silicon lottery here. 🤷
I’m going with exactly what I was trying to do, but it wasn’t very consistent. The i7 6700k isn’t great for overclocking. I recall using an i7 4790K at 5ghz without any problems and stable temperatures, lmao. Back then I was really inexperienced and didn’t realize how things could go wrong. Still, I wouldn’t suggest going too far with the i7 6700k, as it becomes unstable above 4.8ghz. Trying to hit 4.6 or 4.7 is probably too stressful just to keep it running at only 90° under full load for 0.3ghz.