©ASUSTEK COMPUTER INC. ALL RIGHTS RESERVED.

Hayte · ‎01-19-2017

System specs out the way first:

i7 6700K
ASUS Z170-A
16GB RAM (2x 8GB G.Skill Trident Z 3200MHZ, 14, 14, 14, 34)
MSI GTX 1070 Gaming X

CPU overclocking is currently work in progress but on hold while I figure out memory overclocking. Its currently running with adaptive vcore and a -0.060V offset, with core ratio at 43 where its stable on 4.3GHZ @ 1.200V after 2 hours of Prime 95 26.6 small FFT. Beyond this, I've been using the computer for normal use (gaming, music production, idling, web browsing) for over a week without problem.

Whacking up the ratio to 44 results in a clock watchdog timeout BSOD within minutes so I know vcore has to come back up to make progress on this front, but for now its being put to one side so I have a stable system to jump into memory overclocking.

I couldn't get my kit to XMP for any reason so I manually set the following in the BIOS:

DRAM frequency from auto to 3200 mhz
DRAM voltage from 1.200V to 1.353V
DRAM CAS# Latency from auto to 14
DRAM RAS# to CAS# delay from auto to 14
DRAM RAS# ACT time from auto to 34

This works fine, boots into Windows and mem tests for hours no problem. I did notice that whacking up DRAM voltage from 1.200V to 1.353V while leaving VCCIO and VCCSA on auto will result in both sky-rocketing.

VCCIO went from 0.960V to 1.180V
VCCSA went from 1.072V to 1.280V

I don't know much about these two variables or what can happen if they are too high or too low, making it difficult to hold everything constant, change a single variable and observe the effect. In this respect, CPU overclocking by holding vcore constant and raising core ratio or CPU undervolting by holding core ratio constant and lowering vcore is much easier to tell when something is going wrong, why and whats causing it.

Jumping into memory overclocking has been difficult to figure out when something has gone wrong, why and whats causing it. I don't know what to expect if VCCIO and VCCSA are too high but I have seen some warnings against it on this forum.

I have held DRAM voltage at 1.353V and taken down VCCSA and VCCIO closer to stock. It mem tests fine for an hour with VCCIO = 1.100V and VCCSA = 1.100V respectively but below that I started to get mem test errors and then I was hit with a spate of oddities which may or may not be related. The point is I can't tell because I don't know what the possible effects of too low VCCIO and VCCSA are.

Down at VCCIO = 1.05V and VCCSA = 1.07V, I was getting mem test errors within minutes, had a bad_pool_error BSOD and my USB wifi adaptor disconnected 3 times and could not be re-enabled.

So what have your experiences been with manually varying VCCIO and VCCSA? When overclocking memory both have to come up I think. At least, I am unable to get my RAM to 3200MHZ stable without them coming up.

How high can they or should they go and has anything bad happened to you when you set them too high? Has anyone set them too low? What happened? I'd love to know because I hate flipping switches and putting in numbers without knowing what its doing.

(This is my first system where I overclocked anything btw).

Edit: mem test error at 1.100V and 1.100V. Took a while but it happened. Trying again at 1.100V and 1.125V

erickpk · ‎01-20-2017

It's been a while, but I did a lot of reading about the 3200MHz kits of G.Skill cl 14 ram before I bought one. The general consensus at the time was that they don't clock past their rated speed very well, if at all. I don't have the Trident kit (heatspreaders are too high), I got the RipjawsV instead, but they have the same timings.

To get 3200MHz to run at cl14 with an overclock on Skylake (6600k) with Z170 (pro gaming) I had to manually enter vccio 1.8v and system agent 1.23v. It was advised to not exceed vccio of 2.0v and sa of 1.25v. That said, I do recall that at 4.2 and 4.3GHz I didn't have an issue just going with the xmp; it wasn't until 4.5GHz that I had to increase the voltages... mileage may vary though?

Hayte · ‎01-20-2017

Ok yesterday I tested:

VCCIO = 1.1000V
VCCSA = 1.1250V

Actual voltages were slightly higher. This mem tested fine for several hours while playing Dishonoured 2 with Opera tabs open in the background and eventually threw up an error. Still had wifi issues. Possibly not related but worth documenting. Failed to detect wireless network multiple times on restart. Failed to uninstall driver (progress bar goes nowhere). Failed to reinstall driver (again, progress bar goes nowhere). Eventually managed to get it uninstalled/reinstalled. Currently working.

Commuted to/from work. 10 hours later...

From cold boot, system posts but does not reach Windows login before shutting down unexpectedly. Turning on again, it reaches windows login screen and BSODs immediately with the bugcheck name: ATTEMPTED_WRITE_TO_READONLY_MEMORY. Back to the BIOS we go...

Tested:

VCCIO = 1.1250V
VCCSA = 1.1375V

But mem test returned an error almost immediately.

Currently testing:

VCCIO = 1.1250V
VCCSA = 1.1500V

Trying to find the baseline for rated speed. Those VCCIO and VCCSA voltages you posted seem pretty high. VVCIO seems really high (much higher than leaving it on auto for me).

erickpk · ‎01-20-2017

Hayte wrote:
Those VCCIO and VCCSA voltages you posted seem pretty high. VVCIO seems really high (much higher than leaving it on auto for me).

holy s#$... my bad. vccio was 1.18, not 1.8.

Menthol · ‎01-20-2017

I was running high memory speeds on Z-170 from day one, 3200 to 400mhz, I found on my M8Extreme VCIO 1.25 and VCSSA 1.275 worked for me for almost every memory kit I used.
In my mind you could possibly run those modules at 3600mhz 16-16-36, or the same timings as a Gskill 3600 CL16-16-16-36 kit, you may need up to 1.4 volts vdimm but that is not high at all

First foray into RAM overclocking (advice welcome)