OC settings after installing Win 10

[BonzaiDuck]

The 2700K in my sig is as rock-stable as you would want.

When I added another memory kit under Win 7, however, I had to boost the VCORE by 12 mV. No problem, and I moved forward to capture my free Win 10.

The OSes co-exist in a dual-boot configuration, and with longer sessions in Windows 10, an instability occurred during gaming. Further tests with LinX and OCCT confirmed that with stop-codes 124 and 101, I needed to increase the VCORE again by 8mV, and some fine adjustment may be in order. But that 8mV bump in the voltage allowed LinX to run past 30 iterations.

I also notice that affinitizing LinX in Win 10 shows too much variation in GFLOPS during the usual test. That is, the variation was noticeably greater as either a range, standard error or variance around the mean for affinitized LinX under Win 10.

Has anyone noticed this? It can't be my imagination. Any comments or insights would be appreciated.

 

[Ketchup]

I can see the added memory sticks requiring more voltage, maybe a little less overclock. There are more points to get that data to, and some are a tiny bit further away. It all add up (even if it is in microseconds). Are you sure the stability issues from the added memory weren't there before Windows 10? And just, perhaps, not noticed at the time?

On the other hand, drivers are changing, and Asus doesn't even mention Windows 10 on the page for this board. So, you are stuck with whatever Intel sent MS for this chipset. And as you and I know all too well, a good driver can make all the difference.

 

[BonzaiDuck]

I can see the added memory sticks requiring more voltage, maybe a little less overclock. There are more points to get that data to, and some are a tiny bit further away. It all add up (even if it is in microseconds). Are you sure the stability issues from the added memory weren't there before Windows 10? And just, perhaps, not noticed at the time?

On the other hand, drivers are changing, and Asus doesn't even mention Windows 10 on the page for this board. So, you are stuck with whatever Intel sent MS for this chipset. And as you and I know all too well, a good driver can make all the difference.

After the memory upgrade, the instability arose and I chose to bump up VCORE based on the stop-codes. I then ran OCCT for three hours and 25 or 30 iterations of LinX -- no problem.

So the additional shortfall of >= 8mV could only be explained by the OS or the drivers. And I still may have more tweaking to do: 8mV may resolve the instability, but it may be still be short of optimal. Of course, there's my dilemma with LinX: I used the variability in GFLOPS of a 10 to 25 iteration LinX run to assess how much extra to give it after reaching a "stability" threshold. Perhaps I should add another 8mV to see if that indicator shows an attenuation in the spread of GFLOPS.

Also, I'm wondering if the new OS version, if touted for being more efficient, would push the processor harder. That's just a thought. But there shouldn't be a compatibility problem between Win 10 and the last Win 7 chipset driver for the board. I don't even remember if I reinstalled the chipset driver from ASUS under Win 10, which should have included it anyway for the Z68 chipset.

The only drawback for me is the extra voltage, but the extra voltage doesn't put me much beyond my self-appointed limit of 1.38V under stress. It only means that I shouldn't bother with trying to reach a 4.8 OC. Or, it might mean that I want to enable PLL Overvoltage, but I'll want to cool the motherboard better if I do that. Which -- I can do . . . I've posted my solutions in the Cases&Cooling forum recently, still promising to come up with "pictures."