Report on Overclocking CPU Non-K Skylake in 2020: Core Clock Details
Report on Overclocking CPU Non-K Skylake in 2020: Core Clock Details
He just boosted his CPU to 4.25GHz and RAM to 3067mHz. When testing with 3DMark, it shows the stock core clock at 3.2 and turbo at 4.25. The discrepancy is noted—previously Cinebench listed a score of 3.7, but after overclocking the score rose significantly. For 3DMark too, the results improved post-overclock. Someone can help clarify this trend.
Software may misinterpret overclocks. Check clocks with tools like CPU-Z or hwinfo64 during testing. If everything works fine, most programs won't notice. Edit note: R15 should improve performance, while R20 might drop due to AVX limitations on non-K processors.
I am using Cinebench R20. I don't know what AVX means but you're probably right about that. I will go ahead and test it with Cinebench R15. Here is a screenshot in the meantime of CPUz and Hwinfo64. Thanks for the quick response, I was expecting not to get an answer for weeks. First time using these types of forums.
This indicates a clock speed of 4.25 GHz. AVX refers to a collection of instructions that benefit from specific hardware features. However, using certain tricks for overclocking without a K core can limit full AVX capabilities, potentially reducing performance significantly for software relying on it. Cinebench R15 didn't utilize it, whereas R20 did. Intel's performance with AVX improved after Haswell, while AMD only matched Zen 2 results until Zen+. The trade-off between non-K overclock penalties and achieving Zen or Zen+ levels seems reasonable.
Don't fret about it. What matters most is whether the scores improve with the OC.
I didn't verify my CPU at 3.7GHz on CineBench R15 prior to the overclock attempt; I'll need to reset it back to that speed to see if the results improve. My aim is to push the CPU up to 4.4GHz if feasible, considering the Hyper 212 EVO cooler. I'm focusing on maintaining low voltage and temperatures while ensuring stability. I've previously tried overclocking to 4.4GHz with voltages between 1.375 and 1.395, and the system started Windows but experienced unstable temperatures ranging from 30°C to 61-63°C at idle, with multiple blue screens when idle. Perhaps 4.25 is the upper limit achievable with this CPU setup, especially given its age (around 3-4 years) and cooling capacity. My performance is currently constrained by a single-channel memory, which I should have considered during the build but was my first project.
It has been a while since I adjusted my non-K overclock on a similar 6100 processor. From what I recall, I managed to reach around 4.7 MHz, which is standard for a Skylake CPU with air cooling. It seems I was using a cooler that performed even worse than a 212°C. https://hwbot.org/submission/3674669_mac...737.92_mhz From what I remember, I experienced some system issues after hitting that frequency, so I didn’t have much time to finish the bench. I don’t have the exact numbers, but it looks like my RAM performance also depended on bus speed—so be cautious with multipliers when trying higher CPU speeds. I did come across an older benchmark of mine at 3.7, achieving a score of 413. https://hwbot.org/submission/4176103_mac...100_413_cb Personally, since I’m focused on AVX capabilities, I’d prefer to stick with stock settings for everyday use. If the gains justify the drawbacks, I can get a solid 4.2 MHz at typical voltages (up to about 1.25V), but above that the increased voltage leads to more heat and less benefit. Dual-channel RAM is helpful, but with only two cores available, it probably won’t make much of a difference here. If you’re willing to invest, adding another 8GB of RAM in dual-channel mode could help with certain games.
The i3-6100 running at 4.7 GHz with 1.2 V is truly remarkable! It feels like you won a tech lottery. Even reaching basic startup seems unbelievable. The best overclock I managed reached 4.4 GHz, pushing core voltage just below 1.4 V. My processor reaches its maximum at 4.4 GHz regardless of how close I get to the safe range set by Intel (around 1.52 V). I couldn’t boot at 4.5 GHz or higher without exceeding the allowed voltage limits. Eventually, I settled on a 4.3 GHz overclock, which let me lower the idle voltage to 1.36 V and load to 1.34 V. All these settings—core frequency, cache speed, RAM timing, and clock speeds—were fine-tuned as well.
Still amazed that someone achieved such a high frequency at 1.2 V. Clearly, stable operation at this level is unlikely unless there’s an issue with the voltage sensor. It’s safe to say 4.7 GHz at 1.2V isn’t reliable for consistent performance or booting. Yes, boosting the clock frequency also raises the base clock and RAM speeds. I couldn’t find a way to isolate the core clock frequency in UEFI settings—only three options were available. Still, RAM frequency dropped when I increased the base clock, which helped stabilize things.
I eventually reached a 4.3 GHz final setting, allowing me to drop idle voltage to 1.36 V and load to 1.34 V. All these adjustments came with changes in DRAM timings (13-15-15-28-2T-270) and clock speeds. My original 3.7 GHz score was 387 points; a 26-point improvement would be significant, possibly indicating a better build or setup.
I scored 460 points overall and 182 single-core, but I’m not focused on AVX performance. I mainly stream, watch videos, code, and browse. I admit I don’t know which apps rely heavily on AVX—disabling it might hurt efficiency. Voltage increases naturally with higher overclocking, which raises temperatures too. I struggled to keep a 4.2 GHz run without hitting 63-66°C at idle (monitored via HWiNFO), and the package temps likely had a 5-10°C margin error.
Temperatures definitely climbed when pushing to 4.2 GHz, so I avoided it. A solution came in the form of a liquid metal thermal compound and a cheap performance blade—both under $20. The blade worked well without scratches, and the process was surprisingly smooth. Delidding itself was more enjoyable than the overclocking. Before and after photos show only minor idle temperature changes, but under stress tests (RealBench, Prime95, IntelBurnTest), I stayed under 60°C.
Keep in mind that HWiNFO’s readings can be unreliable for core temps; a digital multimeter is safer. For accuracy, allow a 10°C buffer when interpreting CPU temperatures. Overall, the experience was rewarding—especially seeing the system handle stress better after delidding.
Don't rely solely on the CPU-Z data. I was operating well above 1.2V, though I wasn't sure what the exact value would be. My typical max is around 1.4V, but I can push higher temporarily.