Soldato
What about running Prime95 on large FFT's? have you done any proper stress testing?
I'm just experimenting now and I'm at almost 62C on the socket temp just by running 4.4ghz at 1.39V, 4.6ghz seems to need considerably more voltage to be stable (~1.45V) and at that the socket temperature goes through the stratosphere, it just seems to be completely unmanageable.
Okay, I've just tried a fan blowing on the back of the motherboard socket while writing this and the CPU temp goes down 2C straight away, it's not a lot but at least I know it's not directly on the CPU and hence my waterblock is okay, but still if socket temp is linked to throttling how can I pump another 0.06V and 200mhz through the chip without the socket getting hotter?
How are people are getting these amazing temps at high clocks/voltages? are you guys just not testing with Prime95?
edit: I also tried fan blowing on VRM's and temps didn't noticeably change. my core temps are sub-50C under Prime btw for reference.
I don't use Prime as I much prefer OCCT. It's automated so I can set it to stop the test if temps reach a set point. It also records all the temps and voltages in graph form for the duration of the test. This cuts down time spent trying to find out why OCs aren't stable. It's currently running on my 4.6ghz OC and core temps are 49C max (actual is 44C), CPU is 55C max (actual is 55C). That's with 1.464v under load.
If you get no benefit from cooling VRMs (can you monitor their temps in something like I can in the Asus AI Suite?) then you're either cooling them well already, or the heatsink doesn't transfer heat well for some reason.
Also check your VDDA voltage, and CPU/NB voltage. as they are directly linked to CPU temps. My VDDA is 2.7v (quite high AFAIK) and the CPU/NB is 1.35v I think (that's what the board set it to on auto and I believe this is also high. Higher VDDA can help to run lower vcore and CPU/NB helps with higher clocked RAM.