Stability rating 95/100
Stability rating 95/100
Hello. I just assembled my first PC and was currently working on a conservative overclock of the CPU at around 4.8ghz. I was curious about how long it would take to run Prime95 for stability testing. While opinions differ, I haven’t found very recent updates beyond the 6700k discussion. My Prime95 session has been about four hours so far, with temperatures staying between 66-67°C. I plan to run it for roughly six hours total and then perform a full stress test for eight hours while sleeping. Any advice or insights would be greatly appreciated. Also, my CPU is an Intel i5 9600K.
You're not supposed to use Prime95 as anything other than "are my cpu cores working at load" really. It does not test memory, it does not test all the various instructions found on cpu's. If the computer BSOD's, or Prime95 shows errors, then you need to lower the settings or return it to stock. What's dangerous, is if the thermal sensor is damaged or not working, and/or the heatsink isn't on the CPU. You can incinerate a CPU without a heatsink. It's not recommended to run a CPU without at least the heatsink, and any damage the CPU incurs will not be noticed for possibly years. Short story: So I was originally sold a Pentium II 300 as a "450", it had no fan on it, it was just one massive heatsink with fins over an inch long. Anyway, before I ever looked up the numbers on it, it was fine. Then Diablo II came out. Diablo II would BSOD crash the system, on the dot every single time. Likewise the new Xvid/H323 codec's at the time ALSO would do this. So around the time Windows XP came out I was forced to replace the CPU with a Pentium III 500, and lol, it now worked fine. This PC is still under my desk as a last-resort backup as it has the 5.25", 3.5" and CD-ROM burner in it. >_< The point is, what doesn't crash it now, may crash it in the future after the damage has long been done. All it takes is a change in the compiler optimizations and the CPU goes from useful OC to not functional at any speed. OC'd GPU's have the same issue, which is why you really shouldn't buy factory OC'd cards, or if you do, don't OC them further. When new games come out, they may use new instructions or features from newer versions of the API, and then suddenly that card you OC'd becomes completely worthless. Neir Automata and FFXIV (DX10 version) did this to a lot of people, and it's like "no, it's not an optimization problem, it's your crappy hardware that you don't want to admit is crappy." Likewise that's also why a lot of games just kinda never get past DX9 feature sets, because a lot of people play games that work on potato PC's and if they push the requirements too high, they don't run on ANY laptop.
others claim prime95 creates excessive demands on your CPU and that AVX instructions don't matter. they mention crashes after P95 tests but claim normal operation otherwise. they warn it could damage components and turn your space into a dangerous area. while these points hold some truth, I regularly run stress and stability tests on all CPUs I own. It proves highly effective for checking overclocks and memory performance. You can also use mfactor86 to stress-test your GPU, which is closely linked to P95. Today these are the main tools I rely on for thorough testing, as they consistently uncover hardware weaknesses. Of course, I can push overclocks beyond these tools, but I prefer a reliable overclock with no stability concerns. (This usually means my monitor and Windows installation...) If you need a bit more speed instead of rock-solid reliability, prime95 won't be the best choice. If not, I'm happy to walk you through my testing process.
I'm not saying tests are bad, but I'm a bit worried about running something that heavily uses the CPU and GPU for 24 hours. I'm curious what kind of stress test is best to check stability. My PC is essential for work, and my laptop is being repaired. I'd appreciate knowing your approach to stress testing. I tried Prime95 for around 5 hours and 30 minutes, ran Cinebench R20, an 8-hour realbench, and two MemTest86 runs. If I could have fixed XMP, I'd go further—currently it's stuck at 2133MHz with two passes before the CPU OC. It didn't crash, no warnings, temperatures were normal except for a peak of 69°C in P95 (the highest was 56°C). Usually I only play games for about 2 hours at a time, maybe up to 4 if I have more time; otherwise it's just regular work tasks.
the chance of adverse events is very minimal with current parts; they shut down before lasting harm. remember that any ram oc raises cpu temperatures slightly, so consider this when planning. xmp plays a key role in speed. i perform prp evaluations for my primary stability check after an 8-10 hour small ffts session overnight. if you have spare time, you can schedule a few hours of large ffts or small ffts. once everything is optimized, you can run prps via the worker menu, selecting "first time prp on mersenne cofactors." i use just one worker window for simplicity. for a 9600k, aim for six threads. it resembles the standard stress test but uses self-checking methods to detect minor memory issues. i also run extended mfactc sessions; gimps has a github repo for support. it helps resolve GPU overclock quirks and monitors core clocks. you can find more info at https://www.majorgeeks.com/files/details..._test.html. i regularly contribute to gimps, so prps and mfactc runs act as continuous stress tests. for long-term power consumption, i limit 4 cores on prps and set mfactc to 60% power on my 1080ti, which still warms the room noticeably. all this might seem excessive, but if you achieve zero errors or crashes, hardware problems are unlikely. you could potentially reach 4.9-5.0 ghz on the 9600k without overheating, using avx offsets to keep p95 AVX temps below 90°C. the default is -2, though you can adjust it for your setup. what cooling method are you using?