Can I use 85°C fine for CPU overclocking?
Can I use 85°C fine for CPU overclocking?
When running Prime95 26.6 for 15-20 minutes at temperatures under 80°C, it should be safe, and likely even below 70°C with any standard application or game, unless you rely heavily on AVX instruction sets—then testing thermal limits with a newer Prime95 version is advisable. Karadjgne's predictions align closely with typical expectations.
I usually prefer keeping my overclock stable under 70°C during Prime runs, and mine does so even at a full 1.1Ghz ([email protected]) on air. As long as your overclock fits your hardware realistically and you're not chasing unrealistic peak speeds, it should be achievable for most users—unless you're dealing with an especially weak chip, in which case proper CPU and case cooling become essential.
Stability remains equally crucial as thermal performance. It makes little difference if you meet the temperature requirements but are highly unstable, just like a pitbull struggling to pass a peach pit.
Darkbreeze :
If you are under 80°C running Prime95 26.6 for 15-20 minutes, you will be fine, and most likely under 70°C running ANY normal application or game, unless you are doing something that primarily use AVX instruction sets, and if you DO use those, then you should test your thermal limits with a different version of Prime95 that's newer than version 26.6. Karadjgne is almost spot on with his estimates for normal usage.
I generally like my OC to be capable of remaining below 70°C during Prime runs, and mine does even with a full 1.1Ghz OC ([email protected]) on air, so as long as your overclock is realistic for your hardware, and you're not trying for some unrealistic full time ridiculous 5Ghz, it should be easily within reach of doing for most people unless you get a particularly poor piece of silicon, so long as you have sufficient CPU and case cooling.
Still, stability is JUST as important as thermal compliance. It doesn't make a heck of a lot of difference if you're thermally compliant but as unstable as a pitbull trying to pass a peach pit.
I find it humorous that I said a short run on prime95 was enough to prove stability of a system. Of course this was in another thread. Then you said to me that 24 hours is needed to prove stability. Now your saying the same thing I did basically by saying 15 or 20 minutes is enough time, since you will not reach close to those stress levels while gaming.
When I said this about my "ridiculous 5.0ghz overclock", you said that was not proof it was stable. Oh ya I don't crash, my games don't crash, nothing crashes my system, and all my games run around 70c.
Oh ya I do agree with whoever said that 95c is way to high in prime95.
I did not make that statement. I mentioned thermal testing specifically, not stability testing. These are two distinct activities that can be performed with the same tool. Also, I wasn't referring to anything about achieving stress levels while gaming. What I emphasized is the importance of a stable system to prevent silent corruption from impacting operating systems or application files over time. I believe you may have misinterpreted something else being said. I would never make such a claim.
For thermals I rely on p95 26.6 small fft because it's all CPU work and doesn't require AVX or anything like that. To maintain stability, I favor Asus RealBench as it penalizes the whole system at once, testing CPU, GPU, and RAM together in typical usage scenarios with tools such as Handbrake. No optimization is 100% stable—everything is just a guess. The goal is around 99% stability, but since most users only use 50-60% of their CPU power, you could theoretically have a fully stable machine.
Testing several parts together is not a helpful approach. (EDIT: Originally presented as a key technique for assessing stability. It is not entirely pointless overall, though it has limitations.)
It provides no understanding of the reasons behind failures, making it irrelevant for meaningful analysis. Memory checks should come first, under normal settings, unless you need to verify performance at specific clock speeds without overclocking the processor. Occasionally, a slight increase in speed is necessary to achieve desired performance. Typically, begin with memory testing so any later instability can be traced more accurately. If you consistently run multiple passes of memtest86, it suggests memory is unlikely to be the main cause of crashes or issues.
(ADDITIONAL NOTE: This process can also be reversed—starting with default CPU settings and adjusting memory configurations as needed. It might be preferable in certain scenarios if you anticipate needing CPU overclocking for optimal memory performance.)
If all components are tested simultaneously and the system fails, pinpointing the issue becomes difficult. You usually lack clear guidance and may have to rely on trial and error, adjusting clock speeds or voltage, or make assumptions until stability is achieved. This approach deviates from standard practices and should be used cautiously.)