Aurora R7, Overclocking

I've actually gotten it lower than 1.1v with a -0.185v offset and 46x multiplier, but Windows started doing weird things on reboot.  Strange because it'll run Prime95 like that for an hour without issue but as soon as you reboot, all bets are off.  Peak temps with a -0.140v offset for 1.110v across the cores running the Small FFT torture test was 72C.  That's a 20C drop compared to running on auto voltage settings using AWCC.

Decided to tweak it a bit again.  Settled with the following tonight.

So I had to tweak things a bit more.  Apparently I wasn't 100% stable running Small FFTs on Prime95 with one worker usually halting after a few minutes.  But now I think I've got everything sorted out.  4.6 GHz on all cores. Peak temps at 79-80C running small FFTs.  Core voltage is variable 1.110 with a negative offset of 0.155v.  HW Monitor reports IA at 0.978V. AVX is set to 0 offset so all 6 cores are running at 100% with AVX instructions.

@GTS81haha... there's the ADATA 1TB NVMe, then a 500GB Corsair MX500 SSD, a 480GB ADATA U63 SSD, and an internal 3.5" HD.  There's also one external USB drive connected.

@amstel78: I just looked at the pic you posted earlier. How many storage devices do you have in your R7??? 

Decided to go for a full 1 GHz above base.  4.7 seems to run steady at 1.110v variable with a -0.135v offset and AVX to 0.  Peak temps at 84C, but had to bump TDP up to 118W to prevent power limit throttling. 

Very surprised at how little voltage this particular 8700k needs though.  Must have hit the silicon jackpot with this chip.

I'll probably just leave it here and call it a night. The poor 120 rad really needs to be a 240 for me to want to push further.

Well, couldn't leave well enough alone. 

Finally have a stable OC at 5 GHz. Core voltage at 1.200v with a negative offset of -0.065V.  TDP raised to 140W. Runs stable on Prime95 with Small FFTs.  Core temp peaks at 92C.  Idle drops to high 30's.  AVX is set to 0 offset and remains there even after a reboot.

@GTS81yep, thanks for that link. I've seen it before.  The whole purpose of my experiment was to see how low I could go in terms of voltage while trying to achieve the fastest stable multiplier. 

I've had my 8700k at a static 1.35v before and off course, was rock solid in terms of stability.  But, with AVX at 0 offset, would heatsoak the 120mm rad very quickly. Even with it at 3, it would get very hot.  Playing with variable core voltage at 1.2v with a small negative offset seems to work pretty well.

I'm starting to think I've got a dud of a CPU here with my 9600K. AVX kills everything. AWCC auto configures me to -3 by default. CPU reaches boiling point running Prime95 FFT @ 4.6 GHz with 0V offset. Using Dell's AIO CPU cooler with Corsair ML120 Pro running at 2000 RPM.  

Suspecting the way cores are stacked in the die which gives some leeway to 8-core configuration in terms of thermals and electrical characteristics.

Do you control fan curve using AWCC? I thought that needs the OC Control software?

I just feel it is riskier to leave the fan control entirely to BIOS. It seemed to work fine with stock Dell fans but not with the ML120 Pro. Without a custom curve, it almost never turns on until its too late.

Oh waita sec, you have an R8 right? If so, AWCC is completely different. The R7 version has the OC functions as a separate add in that can be uninstalled so it doesn't conflict with other tuning programs like XTU. I can still control fan curves because that's managed by the thermal control module in AWCC instead.

You could still try XTU if you're feeling adventurous but I'd make sure to have a backup image of your windows partition first in the event of weird software conflicts. 

There's something really borked with the way Dell's BIOS implements voltage control.  No matter what I do, vcore voltage will not remain static.  Even with SpeedStep turned off and voltage set to override, VID will fluctuate past what's been defined.

Also with regards to offset, since it affects the entire VID stepping tables, if set too low, idle voltages will drop to ~0.650v.  You'd think the offset would only affect peak draw with a static setting, but that's not how it's working here.  That's how I discovered that I still had instability. Even if Linpack would run without fault for 8 hours, the moment one of the C states kicks in, voltage drops below 0.750v and the system becomes unstable.  As a result, Windows would randomly reboot when idling.  

Unless I've missed something, it's pointless messing with vcore voltage in XTU because it always gets overridden. This really sucks because now you're at the mercy of having the VRM dictate voltages and often it's far more than what the processor actually needs.  

TL;DR: Aurora's aren't for those who want to really get the most out of hardware.  Yes, you can OC but it won't ever be as efficient as it could be with different hardware.

Yep, I've never gotten Vcore to be static using the knobs you mentioned. However, I'm not sure if the fluctuation is due to the intense CMOS switching or the VRM supply side like you said. If you forced the system to be in high performance power state, do you still see Vcore fluctuation while idle? I can try that tonight with my R8.

I suppose most of us here are resigned to the fact that we do not have a true gaming rig where we can tweak everything and we're the minority that isn't Alienware's target demographic.