Uncertain about interpreting the readings during stress testing of the OC
Uncertain about interpreting the readings during stress testing of the OC
See my build details in my profile if they help.
In general, I’ve been able to maintain a 100% stable overclock on my 5820k at 4.4 GHz with 1.264V using the Noctua NH-U14S aircooler.
Before that, I’d told myself I couldn’t push past 4.5 GHz without a better watercooler with two 140mm fans.
But lately I’m questioning whether that was the right target.
I’ve been monitoring core temperatures during stress tests in Aida64 and while gaming or doing other heavy tasks, I mostly ignored the overall CPU temp, focusing instead on keeping the hottest parts under control.
Now I’m wondering if the best approach is to run a stress test and just keep an eye on the average temperatures rather than chasing the absolute max.
For instance, during a stress test, I notice that hotter cores stay around 70-75°C, cooler ones around 66-70°C, and the whole package stays near 66-67°C.
There are regular spikes—especially in core #3 (84-87°C) and core #1 (82-84°C)—while other cores spike at 74-76°C.
Aida64 shows average temps of 64-66°C for all cores except core #3, which is at 70°C, well within safe limits.
I’m trying to understand whether these spikes are dangerous or just normal, and whether I should aim higher based on averages or stick to the safe range indicated by the spikes.
Could it be that reaching higher safe temperatures is possible if I rely on average readings, or have I already reached the limit considering the fluctuations?
You can find my build details in my profile if you find them useful.
I'm currently running a 100% stable overclock on my 5820k at 4.4 GHz with 1.264V using a Noctua NH-U14S aircooler.
Before, I told myself I couldn't reach 4.5 GHz without a better watercooler with two 140mm fans.
But lately I'm questioning whether that was the right target...
I've been paying more attention to core temperatures during stress tests in Aida64 and while gaming or doing other CPU-heavy tasks.
I've mostly overlooked the overall CPU temperature, as it seems...
personally i dont stress test
see no reason to torture the cpu way beyond what real life use would do to it
though some people think stress testing is mandatory
i just load it right up with the things i actually do on my pc
try coretemp for monitoring the cpu temps those little spikes may just be an error
I dont stress test much either, if its stable enough for what I'm doing i.e gaming or something thats fine for me. I will run a stress test for maybe 10minutes or so just to make sure i'm not crashing immediately and temps aren't too high if so I start to game and if I dont crash gaming or using the computer normally for a few days I call it stable. I'm sure some people would say thats a lazy hack way to overclock but it works for me.
I would say those temps are fine especially for an aida64 stress test, I would try to push it a little higher. I would just make sure no cores are throttling, aida64 if I remember right has a little thing on the bottom that goes red for each of your cores if any reach TJ max and start to throttle.
You see my build listed in my profile if it can help. In short, I'm currently achieving a fully stable overclock on my 5820k at 4.4 GHz with 1.264V using the Noctua NH-U14S aircooler. Earlier, I told myself I wouldn't be able to push past 4.5 GHz unless I upgraded to a better watercooler with two 140mm fans. But lately, I'm starting to question whether that's really accurate... Since I've begun suspecting I might have been concentrating on the wrong temperatures. I've been monitoring the peak temperatures in the cores during stress tests in Aida64, as well as while gaming or doing other CPU-heavy tasks. I mostly overlooked the overall CPU temperature, thinking it would stay lower and that I should focus more on the hottest component. But is this the right approach? I'm wondering if it's better to run a stress test and just keep an eye on the temperatures without focusing too much on the absolute max, so I can decide whether it's getting too hot before the average shows up. For instance, if I run a stress test now and observe the temps in real time, the hottest cores might be around 70-75°C, while cooler ones sit at 66-70°C, and the entire package stays near 66-67°C. However, there are frequent spikes in the cores—especially core #3 (reaching 84-87°C) and core #1 (around 82-84°C). Other cores spike to about 74-76°C. On the other hand, Aida64 calculates an average of 64-66°C across all cores, with the exception of core #3 at 70°C, which is well within safe limits. I'm trying to understand this better—those spikes are definitely high, but they're not constant. They reach elevated levels occasionally, yet they don't stay there long enough to be dangerous. What should I consider when deciding whether to push further? Should I compare maximum temperatures or average temps? How do core temperatures compare to the package temperature? Could I safely exceed these averages, or have I reached the upper limit (given the spikes)? This seems to address your concerns. Check out this link: In short, focus on the average. Your tests reveal some extreme cases. Stress tests mainly check CPU stability and a cooler's efficiency. You've confirmed stable overclocking and accurate temperatures under load. After that, you can trust your real-world temps will stay within safe ranges. Also, note that most coolers come with one fan for push mode; you could add another to improve heat removal and possibly reduce those spikes—though it's not necessary yet. I'm unsure why my answer was posted twice. My apologies. If a moderator wants to remove it, just let me know. I can delete it using the trash icon. For more details, see this detailed Intel temperature guide: It's packed with useful information and should help clarify things without raising further questions.
Thanks for the links! Now you know how to keep an eye on it. Just checking if thermal testing in Prime95 is still needed for my case, since I'm not doing it just for benchmarking temperatures.
Regarding the recommended max temperature of under 80°C and around 75°C, it seems aimed more at testing Prime95 specifically rather than general stress testing. If a simpler test like Aida64 or gaming made the CPU reach about 75°C without extra load, it might still be too hot. So even if other apps run fine at that temperature, it’s safer to keep it under control.
If it’s safe for your setup, you could stick with around 75°C maximum. You’re already using a heavy real-world workload, so adding a second fan to the heatspreader makes sense—especially since you’re not switching to watercooling. There’s some debate about whether it really helps, but balancing a quieter fan curve and a responsive thermal curve could help. A slight manual throttle at higher temps (like -200MHz and Vcore adjustments) might also make a difference.
everything really depends on what you're comfortable with personally
I’d just use the strongest real-world application you can, but it should still stay under 75°C
it’s still your choice how loud the fans are versus how cool the CPU feels
I prefer quiet operation, so my fans are kept low
but I have an overkill custom liquid cooling setup
my temperatures rarely go above 60°C even when I run everything at full capacity all in one session