Asus Z170I Pro Gaming and ram

[mackerel]

Today I built a mini-ITX box for VR, in part from components I had already, and some new bits due to the form factor. The mobo I chose was the Asus Z170I since I like the VIII Hero, but wasn't feeling like paying for the VIII Impact. The ram that went in it was Corsair Vengeance LPX 3000 2x4GB. This had been running fine in MSI Z170A Gaming Pro simply by setting XMP previously, so thought I could simply do the same in the Asus. Nope. Everything seems fine in SPD mode. If I turn on XMP, no boot. I'm on latest bios so it should include all the ambiguous "improved stability" stuff. If I turn on XMP and set ram to 2800, it boots. So... marginal instability? Would this be a VCCSA/VCCIO tweak or something else?

I'm running a bit low on time, and am thinking of swapping in some Crucial Ballistix something or other 2666 instead and see how that does. The 3000 can go back in the MSI board which currently has the 2666.

 

[mackerel]

More by guessing than any solid reasoning, I quickly tried 3000 XMP + SA 1.08 and IO 1.1. Actually not a total guess, 1.1 for both seemed best for some G.skill ram on MSI board, but no guarantee it'll carry over. Also the Asus put SA into the red zone at 1.1, so I backed off a little to 1.08. This was no boot. I put them back to auto, and then set Vram to 1.375, again the last step before it went red. I know it has been said here I could go higher, but I wouldn't run higher long term anyway. No boot.

I didn't swap in the Crucial 2666 after all, since there is no practical way for me to extract the ram safely without taking the system apart again. I wanted to reconfigure the cooling anyway so I'll do that as part of a bigger job another day.

I can still run SPD stably for setting up software until then.

 

[SalemSaberhagen]

More by guessing than any solid reasoning, I quickly tried 3000 XMP + SA 1.08 and IO 1.1. Actually not a total guess, 1.1 for both seemed best for some G.skill ram on MSI board, but no guarantee it'll carry over. Also the Asus put SA into the red zone at 1.1, so I backed off a little to 1.08. This was no boot. I put them back to auto, and then set Vram to 1.375, again the last step before it went red. I know it has been said here I could go higher, but I wouldn't run higher long term anyway. No boot.

I didn't swap in the Crucial 2666 after all, since there is no practical way for me to extract the ram safely without taking the system apart again. I wanted to reconfigure the cooling anyway so I'll do that as part of a bigger job another day.

I can still run SPD stably for setting up software until then.

My RAM didn't like going directly to 2667 XMP?

I set it downclocked to 2367XMP or something (100/133), then after that boot I corrected it and it stayed fine even with a nudge to 103fsb/2678.

 

[mackerel]

http://forum.corsair.com/v3/showthread.php?t=156580

Found the above on Corsair's forums. Could that be my limiting case here? I don't mind swapping to the Crucial 2666 and return the Corsair 3000 to the MSI board where it works fine.

 

[mackerel]

Did a mini-rebuild on the case today while diagnosing some heat problems, and have now swapped over to the Crucial 2666. Seems to be working ok.

 

[SalemSaberhagen]

Glad I was cheap like a gamer lol

 

[mackerel]

I had another look at this today...

So, same Asus Z170I Pro Gaming mobo, currently bios 1809 which isn't newest, but the one newer only lists future CPU support (presumably Kaby Lake) so I haven't bothered.

It has been recently running Kingston khx2666c15d4/8g just fine. I put the Corsair cmk8gx4m2b3000c15 back in, and just like before, no boot at 3000. I tried both 100x30 (3000) and 133x22 (2933) and neither would boot. I did get 100x29 (2900) and 133x21 (2800) to boot ok. I noticed the mobo auto set vccio and vccsa were lower than that when the same ram was put in a MSI Z170A Gaming Pro mobo. I tried increasing both of those to match MSI at 3000, no boot. I read elsewhere someone with the same ram on different Asus mobos also couldn't boot at XMP. They found reducing the ram voltage to 1.29 to allow it to boot for them. I tried that and... no boot.

Just for laughs, I dug out my G.Skill f4-3333c16-4grrd from another system. This ram I've had a fight with in the past as it wasn't 100% stable at 3333 in either of my VIII Hero or MSI Z170A Gaming Pro mobos. So I put it in with SPD, and it booted. I turned on XMP, and to my surprise, it booted. So far it has been running Prime95 blend for over 15 minutes ok. Regardless, it does show the mobo itself isn't limiting at 3000, and could it be some other setting or limitation?

As an observation, I found it curious the MSI board doesn't seem to allow the 100 strap for ram. It only offers 133 so when I set the ram to 3000, it ran 102.5x1.33x22 in that case.

I'll let Prime95 soak a bit longer to see how stable it is, but I'll revert back to the dual rank Kingston anyway for long term use.

As another kick to the Corsair 3000, I just did an accidental overclock. After removing the G.Skill 3333 I put the Kingston 2666 back in. Didn't change settings, assuming it would detect and reset them on boot, and it reached the desktop. When I fired up CPU-Z I was surprised to see it still running the same settings as the G.Skill did! Is it stable? I did a quick Prime95 4M benchmark, and got 246.77 iter/s, compared to 204.40 with the G.Skill. Same settings, only difference is the Kingston is dual rank which I knew from past experience can help a lot. That's 20.7% worth of "a lot". No idea if it is stable... now thinking should I revert to 2666 to get a baseline benchmark, then check stability at 3333? For comparison, my main desktop got 246.17 which has Ripjaws 5 at 3200 slightly tighter timings than the Ripjaws 4 3333. Both systems have 6700k @4.2.

Edit: note the parts I gave previously are for the module. The kit part for the Kingston is hx426c15fbk2/16.

 

[Woomack]

Many Vengeance kits are from old series ( but were collecting dust as were overpriced comparing to competition ) and they had not the best SPD/XMP profiles. You can probably make it work playing with timings manually. Your asus board should run up to 3600 without issues.
Up to ~3200 there is no need to change VCCIO/SA unless there are some BIOS issues and I doubt.

 

[mackerel]

Woomack when you say no need to change VCCIO/SA, do you mean from the lower values as they are at 2133/SPD, or the higher mobo auto values when running at higher speeds? I don't have the exact numbers on me but from memory it is something like 0.95 and 1.0xx at 2133, up to 1.1+ and 1.2+ at higher speeds. I observed both Asus and MSI boards increased those voltages at higher speed, but the MSI was far more generous.

 

[Woomack]

Skylake is strong enough to run at stock voltages up to at least 3000-3200 but some motherboards are bumping voltage when you leave it at auto. At 3333+ and auto settings I saw 1.35V SA on some boards what is more than enough to pass 4200.
In most cases issues are causing SPD/XMP profiles as older memory kits were designed for early X99 while 100 series have slightly different sub timings. There are also new Hynix IC which require new BIOS to work good. I had issues at higher frequency with Kingston and Team Group memory till I moved to new BIOS. Problem is only that not all motherboards have new BIOS release prepared for new memory IC. I had no problems with any Samsung IC so far. On MSI board I got new beta BIOS about 3 weeks ago but ASRock released it last week.
On MSI Z170I Gaming Pro AC I can't set more than 1.25V SA and I'm able to run memory up to 4200+. For ~3200 shouldn't be required more than ~1.05V ( stock is 0.95V ), the same for VCCIO. However what you set is not always equal to real voltage ( you can check it in hardware monitor tab but it's also not always guaranteed that read is correct ).

 

[Theocnoob]

Skylake is strong enough to run at stock voltages up to at least 3000-3200 but some motherboards are bumping voltage when you leave it at auto. At 3333+ and auto settings I saw 1.35V SA on some boards what is more than enough to pass 4200.
In most cases issues are causing SPD/XMP profiles as older memory kits were designed for early X99 while 100 series have slightly different sub timings. There are also new Hynix IC which require new BIOS to work good. I had issues at higher frequency with Kingston and Team Group memory till I moved to new BIOS. Problem is only that not all motherboards have new BIOS release prepared for new memory IC. I had no problems with any Samsung IC so far. On MSI board I got new beta BIOS about 3 weeks ago but ASRock released it last week.
On MSI Z170I Gaming Pro AC I can't set more than 1.25V SA and I'm able to run memory up to 4200+. For ~3200 shouldn't be required more than ~1.05V ( stock is 0.95V ), the same for VCCIO. However what you set is not always equal to real voltage ( you can check it in hardware monitor tab but it's also not always guaranteed that read is correct ).

Hey Woomack, I missed the information that comes from your brain. Always so informative. Cheers man.

 

[wingman99]

Post a screenshot of Vccio Vccsa in XMP mode with HWiNFO64.