overclocking temperatures for the i5 6600k using an h100i
overclocking temperatures for the i5 6600k using an h100i
Hello,
I'm using an i5 6600k with an h100i cooler. The radiator fans are on intake mode, and two exhaust fans are active. The multiplier is at 45, with a vcore of 1.3v. After a 30-minute realbench stress test, I'm reaching around 80°C. Is this within the normal temperature range for a correctly functioning h100i under these settings? Any advice would be helpful!
This method is not suitable for evaluating thermal compliance. Realbench is intended for assessing stability, not thermal compliance. The correct procedure involves specific testing steps. Exceeding 80°C during thermal compliance testing is regarded as undesirable.
This method is not suitable for evaluating thermal compliance. Realbench is intended for assessing STABILITY, not thermal compliance. The correct procedure involves specific testing steps. Exceeding 80°C during thermal compliance testing is regarded as undesirable.
I agree with Darkbreeze partially here. A useful tip is to begin with an AVX check to grasp your temperature limits and stability in one initial test, followed by deeper evaluations. Avoid spending excessive time on a short, unstructured stress test if you prefer efficiency. I don’t recall the specific CPU model used, but I’d look for reference voltage and frequency elsewhere.
OCCT small sets with AVX (1.250vcore or higher results at high temps) – 15 to 20 minutes or until temperatures rise too much. (Keep an eye on the CPU continuously.) If it passes without BSOD or freezing, proceed further.
This approach works well for a quick assessment before committing to more detailed testing.
If voltage is insufficient, you’ll usually notice the system failing within 10 minutes. If it’s acceptable, move on to actual performance evaluation.
Realbench – 1 hour (skip to test 3 if you want to focus on AVX offset).
OCCT small sets without AVX – 30 to 60 minutes (varies with temps)
OCCT large sets without AVX – 6 to 8 hours
Prime95 small sets without AVX – 30 to 60 minutes (based on current 4.4 GHz, 1.210v)
Prime95 Custom blended without AVX – 6 to 8 hours
Aida64 – 6 to 8 hours
For the first test:
- If errors appear, it indicates instability.
- If overheating occurs without errors, voltage is too high. Start low and gradually increase frequency to identify safe limits.
Edited: according to Darkbreeze’s clearer separation of temperature and stress testing, his recommendations provide more depth than the summary above. For beginners, follow his suggested steps (RAM check, then temperature test) for better results.
That's a lot of stress testing, I think it's beneficial to check stability, but you'll notice CPU/cache issues often lead to blue screens during games (certain BSOD codes can guide adjustments) or minor performance drops compared to less expensive options. I also agree that extended testing is wise when overclocking RAM, as I had to reinstall several times with slightly inconsistent timing settings.
It's really the only way to stress test if you want to make sure your overclock has absolutely zero lifespan impact on the CPU. If you keep it under the manageable temps in the
temperature limit tests
, then you know for a fact your CPU cannot get up to higher temps otherwise. Games are starting to use AVX now, especially with directx12 support, so imo if you care about something totally stable its really the only way, especially on next gen games.
While lighter stress tests are an option, this way doesn't take long since you figure it out based on step 1 usually and will 90% of the time be stable on the rest. Definitely don't recommend gaming alone as a "stress test" since gaming is only going to hit your spike and full load on rare occasion, and what are you going to do next? Tweak something and wait 100 days for the next crash to assume it's off? Basically the above tests guarantee with 24/7 uptime you will not do any damage to your CPU over the lifespan it would have had if you left it at stock settings. The golden goal of stable overclocking
I've also found out some really important things from my own, like improper thermal paste application and some bios settings that have greatly lowered my temps.
For example, I just failed pushing my 9700k on test 5 for Prime95, so now realizing I will need to up my voltage for my 4.9 ghz overclock, and increase my AVX offset to compensate for the temps. That's all the tweaking you really need to pay attention to if you run this set. My mobo sucks so I can't hit the usual 5.0 ghz that most people get.
Yeah, I did run my RAM at its manufacturer XMP 3200 mhz which is probably pushing my CPU a lot also
. But anyone you ask will never recommend to not at least use the auto XMP from the manufacturer, I don't do any RAM timing tweaks besides that.
Additionally, Darkbreeze's standard guide is acceptable but could be outdated. A few years ago, fewer people paid attention to AVX loads, so they simply turned them off and measured temperatures without testing. Therefore, I wouldn't rely on that guide entirely (though Step 5 essentially references the link's results). Performing an AVX offset becomes increasingly important even with high-performance coolers, as they now contribute to a different level of heat generation.
The "stress test guide" is actually a thermal compliance guide. These two concepts are completely distinct, and recognizing this sooner would benefit all communities involved. However, this change in perspective won't occur naturally. It's simple for people to mistake these terms, even though they are entirely different in meaning and approach.
Sorry -- last question hijacking this thread I promise.
Darkbreeze, why aren't we suggesting an AVX-based test to begin again for temperature limits? You mentioned your guide briefly touches on AVX but it was more relevant in 2020 and is only improving slightly now.
If you have any suggestions on how or whether not to run an AVX test (since I’ll probably need AVX for my programs), that would be appreciated.
Are you concerned about staying under a certain temperature when running Prime95 without AVX as a starting point? Or are you thinking you can check a specific temp threshold using AVX based on the Small FFT test in your guide, without overloading the CPU?