Originally Posted by SuperZan Warning: Spoiler! (Click to show)
To elaborate on the above, the stress test in Y-Cruncher should by default reserve as much memory as is available so when it runs the various instruction-sets and workloads, your system will be tested for stability in the CPU as well as interaction with memory and cache. One run will generally show any glaring instability and five runs will generally reveal any subtle instability which may come back to haunt you later.
I've yet to pass five or more runs of Y-Cruncher stress test and fail any subsequent stability test or suffer from system instability or data corruption later. This has been true of AM3+, FM2+, Z170, Z97, and X79 so I expect nothing different from AM4.
Originally Posted by mus1mus Warning: Spoiler! (Click to show)
Just to add, X99. It's overkill for the platform, but hey, Stable.
Just open the app, press 1 to cue the stability suite, press 0 to start.
The test will run each subtest for 2 minutes, 5 sets will be around an hour or so. 3 sets will actually be enough as this is rather very quick to detect instability. Only FFT doesn't stress Ryzen.
For reference, minute memory instability may cause the app to finish one set but not the 2nd or the 3rd. Even long term Memory stress testing can be omitted if you pass 5 sets of y-cruncher.
That's just my experience.
Edit: 5 sets can also detect Vcore deficiency on the stability front. This causes Ryzen systems to restart though testing passes short duration runs.
. I went nuts with Y-Cruncher yesterday evening
. I found same settings as x264 for sufficient for Y-Cruncher. Runs ~5°C hotter than x264/RB/f@h, but still super quiet fan profile
. RB for me needs less voltage than x264/Y-Cruncher and shows instability only after longer time than x264/Y-Cruncher.Warning: Spoiler! (Click to show)
Only thing is I still get core crash on f@h
. In chronological order.
My base 3.8GHz OC 2133MHz used +106.25mV (~1.319V on DMM) for x264, etc. Repeatedly tested in x264/RB/f@h and no issues. I upped VCORE to +112.5V when I went 2400MHz RAM and left SOC [AUTO] it shot from ~0.838V on DMM to ~1.050V.Warning: Spoiler! (Click to show)
f@h 16/03/17 evening, (3.8GHz OC 2400MHz VCORE +112.5mV (~1.330V on DMM) SOC: [AUTO]).
f@h 17/03/17 morning (3.8GHz OC 2400MHz VCORE +112.5mV (~1.330V on DMM) SOC: [AUTO]).
Due to the errors on L0 cache I went and lowered SOC to 0.950V, upped VCORE to +118.75mV. As thought perhaps CPU doesn't like high SOC and need some more VCORE. Through out the day no issues on f@h. Tested with Y-Cruncher as shown above screenies.
This morning (3.8GHz OC 2400MHz VCORE +118.75mV SOC: 0.950V).Warning: Spoiler! (Click to show)
So what you reckon guys, SOC or VCORE need increase? I'm currently testing lowered SOC again. 3.8GHz OC 2400MHz VCORE +118.75mV SOC: 0.925V.Edited by gupsterg - 3/18/17 at 4:40am