Very intense, very power demanding, and VERY DAMN HOT stuff. If your cache overclock is even remotely unstable, you will find out in minutes, lol. Again, be very careful with it, as it's quite brutal.
#Ryzen 1700 undervolting download
Download LinX, and replace it's EXE with the one obtained from the MKL download. Changing ram capacity will automatically change problem size on a linear scale). (Use this GUI to set your work size to 75% of your ram's capacity. It's hard for any synthetic stress test to simulate power load switching and continuous switching of said loads, so a realistic stress test is always great to add in to the mix. Linpack (non-MKL, larger data sets) is actually good for short spurts of vcore testing, but you will definitely want a longer, more realistic load if you want to say "24/7 stable". After that, you are pretty much good to go. It's been years since I last used OCCT, so I cannot speak from experience as to how long it takes to force instability, but Prime95 with the above settings tends to take roughly 8 hours or so. In Prime95, the value would be 1344k-2688k. OCCT is usually the test that fails the quickest in bad OCs, most of the times way before temperatures are anything to worry about (it's not a big deal to cool the OCs I do anyway).įor OP's case, heat isn't going to be a problem since he's actually undervolting, so the most useful test is the one that detects a CPU no longer working properly due to too low a vocre the quickest.įor that, you do use larger data sets. However, I have the same experience as it comes to stability itself, meaning whether the CPU fails as in it corrupts data / miscalculate things.
So, as a test of your cooling solution, there may be no harsher test than whatever Linpack-based routine you are running as a test. So decreasing it down to 1 volt lowered heat output rather dramatically, and also lowered performance as well. You are talking about stressing the CPU in terms of temperature, though. Just don't attempt it while overclocked until you know how well your cooling solution performs. If you'd like to try it, I can share a compiled version of it. Both your cache and memory subsystem gets stressed, and your CPU will be as hot as it can ever be. Linpack MKL is scary because it performs both small and large, simultaneously. Large data set is done within ram, and is nowhere near as stressful as small sets performed in cache (AVX scales with bandwidth, cache bandwidth is magnitudes faster than ram).
0 kommentar(er)
