Talk about 100MHz devices that are expensive.
Talk about 100MHz devices that are expensive.
For roughly two years I operated my 6700K at 4.6GHz, VCore 1.344 and a maximum temperature of 84C using Prime95, with four memory sticks running at 3600MHz,16,16,36,2. Recently I found my system wasn’t very stable, so I reduced the frequency by 100MHz. Now it runs at 4.5GHz, VCore 1.28 and a maximum temperature of 74C with Prime95, using the same memory speed. Everything is now stable. The extra 100MHz cost a significant amount. Here’s a quick breakdown:
Sacrifices:
- 4.6 → 4.5 (2.17% drop in frequency)
- Gains: 85°C → 75°C (11.76% drop, more speed)
- VCore from 1.344 to 1.280 (4.76% drop)
Benefits:
Lower fan speed (quieter operation!)
Less power consumption
Greater stability
I didn’t measure the exact wattage, but I’m confident it cut about 10 watts under load. Overall I’m very satisfied with the improvements. If you’re wasting energy just by adding 100MHz, reducing it might help.
the lower temperature comes from dropping the vcore. consider reducing it further.
😉
observe the results even when keeping the 4.5
BSOD at 1.268 occurred, I believe that's all. I'll keep it at 1.280 tonight, praying it doesn't fail me.
For a more accurate and relevant comparison, a few additional details are required.
What is the temperature in your ambient room?
My room is at 26 degrees Celsius.
Which version of Prime95 are you using?
I believe it's the newest release: 29.6 Build 6.
What is the torture test setting?
Custom MinFFT size (in K) = 4, MaxFFT size (in K) = 8192
Memory allocation (in MB) = 26000 (I own a 32GB drive but didn't want to use the SSD with page files)
Time required for each FFT size (in minutes) = 6
I experienced nearly identical results with my i5-6600K. I had to lower it to 4.5 GHz from 4.6GHz for stability about two years ago. The processor gradually loses performance, and the manufacturer's built-in margin was exceeded during the overclocking process.
Mkazem, The current Prime95 release is 29.4, which likely matches the version you were using. Updates beyond 26.6 incorporate AVX instructions that may cause your processor to overheat because they can exceed the processor's capacity by nearly 130%. Even though you didn't specify a cooling method, the custom P95 setup you selected differs from the standard "Blend" test, which provides a changing workload ideal for checking memory durability. As the custom configuration cycles through smaller FFT sizes like 4K, the CPU is mainly stressed during those phases, leading to high peak temperatures. In contrast, Small FFTs maintain a constant 100% TDP load, resulting in stable core temperatures and making them better for evaluating CPU thermal behavior. Even though your room temperature is 4°C higher than normal (22°C), your core temps remain acceptable. Still, AVX can negatively impact stability. If you perform AVX tasks such as rendering or transcoding, adjust the "AVX offset" in BIOS to at least 300MHz downclocking, which will lower core temperatures and improve stability without sacrificing gaming performance. If you don’t use AVX applications, there’s little need to test AVX stability. You’ll notice that using Prime95 version 26.6 (non-AVX) lets you return to your prior overclock settings while keeping stability with a lower Vcore and cooler cores. • Prime95 v26.6 - http://www.mersenneforum.org/showthread.php?t=15504 Remember, according to Intel’s datasheets, TDP and thermal specs are confirmed without AVX. You may want to refer to Section 12 - Thermal Test 100% Workload: Intel Temperature Guide - https://forums. CT 😎