On a Z170-A, after BIOS 3703, RAM frequency affects bw/power to onboard audio and USB

[JDA81]

My motherboard is the Z170-A. For over a year I was using BIOS 2202 with no problems. I could run my RAM at 3000MHz without issue, perfectly stable. I used the XMP profile which set the DRAM voltage to 1.35v. I also overclock my CPU but that has no part in this story.

When I updated to 3703 because of the recent Spectre flaw, I noticed several issues right away. If I set my RAM to 3000MHz I get static in my audio. That's weird right? It's 100% reproducible. If I lower my voltage to 1.2v from 1.35v at 3000MHz, I can't even POST. (I confirmed this had no relation to the CPU by running that at stock speed and voltage)

The 2nd (related) issue is even weirder. So I played around a bit with various frequencies and voltages. I found I could run with perfect stability at 2666MHz at 1.2v. I did a Prime95 blend test and a full Windows Memory Diagnostic to confirm stability.

HOWEVER, if I play VR (Oculus, so I have 4 devices plugged in which use a lot of bandwidth and probably also a fair amount of power) within sometimes as short as 5 minutes, I'll get SEVERE interruptions on the USB devices, leading to major gameplay issues or even crashes. I can't confirm if it's power or bandwidth related, but it's almost certainly one of the two because the system is otherwise perfectly stable.

If I set my RAM down to 2100MHz it is perfectly stable in VR. I can play for hours, no problems at all.

Note, this all happened IMMEDIATELY after updating to the newest BIOS, so I know it's not the fault of anything else but Asus's BIOS. I also have eliminated all other factors because the issues are perfectly reproducible by setting my RAM to a higher frequency and changing no other factors. If it was DRAM voltage related, I'd understand, but I can have my DRAM at 1.2v and 2666 and it breaks and 1.2v at 2100 and it's perfect.

I'd love to hear if anyone has any idea why this is happening. If anyone knows a workaround for this issue, too, that'd be great. If there is a better part of this forum or a better community to ask this question, feel free to mention. Also, does anyone know how to contact an Asus tech and make them aware of this issue? Thanks!

(Remember, I did a Prime95 blend test and a full Windows Memory Diagnostic to confirm stability, yet VR is still affected)

[coret3x]

The RAM speed issue with Z170-A is well documented here https://forums.oculusvr.com/community/discussion/51187/ram-frequency-affects-rift-tracking/p2

UPDATE: Asus has fixed this now!

[JDA81]

coret3x wrote:
The RAM speed issue with Z170-A is well documented here https://forums.oculusvr.com/community/discussion/51187/ram-frequency-affects-rift-tracking/p2

Seems like Asus is just ignoring this.

I think that's definitely true, but I'd at least like to know why it would affect the onboard sound regardless of Oculus functionality.

I imagine they have some terrible voltage regulator or some other really bad power management setup by the BIOS, which is leading to RF interference under certain conditions. Too bad they can't reinstate the things that made BIOS 2202 stable.

[JDA81]

A new BIOS came out a few days ago. I updated. Same issue.

What I have found, though, is if I up my PCH voltage I can mitigate the problems caused by the RAM frequency, which at least tells me this is a power issue and not something even more arcane.

Again my problems are tied to RAM FREQUENCY, not RAM voltage. Unless more voltages go into RAM frequency than just DRAM voltage (someone could correct me on this if I am wrong).

Still doesn't completely resolve my problem because I need absolute stability out of my USB buses.

Is it safe to overvolt PCH?

[Kronos35]

I thought I was going crazy but I'm glad I'm not alone. Starting with BIOS 3703 I've had nothing but trouble with RAM on the Z170-A as well but I was able to solve mine. Prior to the BIOS update, everything worked flawlessly after simply enabling XMP. This is using a G.SKILL 3200 CL14 kit. After the BIOS update though, my Vive became completely unusable. I had similar symptoms to what you describe with your Rift where I would be fine for a few minutes and then my HMD display would go completely gray and I'd lose all tracking. I finally narrowed this down to RAM and confirmed everything worked when disabling XMP and reverting to stock 2133.

After lots of tweaking and troubleshooting, I found that when the XMP profile for my RAM was enabled on BIOS 3703, the DRAM, VCCIO and VCCSA voltages were all drastically overvolted compared to what they were on earlier BIOS versions. The biggest culprits definitely seemed to be VCCIO and VCCSA. I disabled XMP, manually punched in all RAM speed and timings, and then lowered the DRAM, VCCIO, and VCCSA voltages to sane levels. I'm now back running the rated 3200/CL14 and I've been 100% stable and it solved all of my Vive tracking issues.

I noticed a new BIOS (3802) as well but haven't bothered to update yet since I'm now hesitant to even try it after my experiences with 3703. I would say that before you overvolt anything else, maybe make sure that the motherboard isn't already pushing TOO much voltage and causing power/stability issues as it was in my case.

EDIT: To clarify your question regarding voltages and RAM, from what I understand, there is more to it than just the DRAM frequency and VCCIO/VCCSA can affect RAM overclock and stability. See more here: https://www.tweaktown.com/guides/7481/tweaktowns-ultimate-intel-skylake-overclocking-guide/index5.ht....

[JDA81]

Kronos35 wrote:
I thought I was going crazy but I'm glad I'm not alone. Starting with BIOS 3703 I've had nothing but trouble with RAM on the Z170-A as well but I was able to solve mine. Prior to the BIOS update, everything worked flawlessly after simply enabling XMP. This is using a G.SKILL 3200 CL14 kit. After the BIOS update though, my Vive became completely unusable. I had similar symptoms to what you describe with your Rift where I would be fine for a few minutes and then my HMD display would go completely gray and I'd lose all tracking. I finally narrowed this down to RAM and confirmed everything worked when disabling XMP and reverting to stock 2133.

After lots of tweaking and troubleshooting, I found that when the XMP profile for my RAM was enabled on BIOS 3703, the DRAM, VCCIO and VCCSA voltages were all drastically overvolted compared to what they were on earlier BIOS versions. The biggest culprits definitely seemed to be VCCIO and VCCSA. I disabled XMP, manually punched in all RAM speed and timings, and then lowered the DRAM, VCCIO, and VCCSA voltages to sane levels. I'm now back running the rated 3200/CL14 and I've been 100% stable and it solved all of my Vive tracking issues.

I noticed a new BIOS (3802) as well but haven't bothered to update yet since I'm now hesitant to even try it after my experiences with 3703. I would say that before you overvolt anything else, maybe make sure that the motherboard isn't already pushing TOO much voltage and causing power/stability issues as it was in my case.

EDIT: To clarify your question regarding voltages and RAM, from what I understand, there is more to it than just the DRAM frequency and VCCIO/VCCSA can affect RAM overclock and stability. See more here: https://www.tweaktown.com/guides/7481/tweaktowns-ultimate-intel-skylake-overclocking-guide/index5.ht....

Thank you for the brilliant reply. This is exactly the sort of info I wanted, and at the very least this seems to absolutely be the right track! I was not aware that VCCIO and VCCSA were tied to RAM frequency at all. I've just tried clamping my VCCIO & VCCSA to what they were 'auto' at during my successful overclock on the old BIOS, and it makes a really significant difference - bigger than overvolting PCH!

Under Prime95 load I'm still getting some static as I'm testing at my RAM's spec frequency, but it's very infrequent. I suspect with some tweaking I'll find some good stable numbers - stable for RAM as well as PCH!

I suspect you're exactly right, that before they were ramping up like crazy in the new BIOS whereas they did not in the old BIOS. I'm sure this was then draining the PCH in turn. I hope no permanent damage was caused, but it's still great to learn more about how all this works.

Which monitoring software do you use? HWMonitor doesn't seem to display VCCIO & VCCSA properly.

[clovitox]

JDA81 wrote:
Thank you for the brilliant reply. This is exactly the sort of info I wanted, and at the very least this seems to absolutely be the right track! I was not aware that VCCIO and VCCSA were tied to RAM frequency at all. I've just tried clamping my VCCIO & VCCSA to what they were 'auto' at during my successful overclock on the old BIOS, and it makes a really significant difference - bigger than overvolting PCH!

Under Prime95 load I'm still getting some static as I'm testing at my RAM's spec frequency, but it's very infrequent. I suspect with some tweaking I'll find some good stable numbers - stable for RAM as well as PCH!

I suspect you're exactly right, that before they were ramping up like crazy in the new BIOS whereas they did not in the old BIOS. I'm sure this was then draining the PCH in turn. I hope no permanent damage was caused, but it's still great to learn more about how all this works.

Which monitoring software do you use? HWMonitor doesn't seem to display VCCIO & VCCSA properly.

What values did you have in your old Overclock (Before BIOS update)? i have the same issue as you all guys, and i lowered the VCCIO and VCCSA manually to around 1.1V and the audio problems got fixed (now i dont have to overvoltage my PCH as well, i had it at 1.1V and didnt fix all the issues). But i wanna know what were the ´´safe and working´´ values for the VCCIO and VCCSA, i dont really remember and since we have the same Z170-A i could give it a try.

[JDA81]

Followup, I think you were right about everything Kronos35.

I once again have my system stable on the latest BIOS and with RAM at DDR4-3000! VR plays just fine - maybe better than ever. Got my system clocked as before.

Long story short - I downvolted my VCCIO & VCCSA a bit from what I had recorded them at on BIOS 2202, and kept the same DRAM voltage. This allowed the RAM to be stable in various tests (mostly Prime 95), and my PCH is stable so no stuttering audio and Oculus Rift works perfectly!

I found another thing while testing Prime95. There was a test I did not usually run, and my system was not actually stable there. I'd get a fault after several minutes (this was something that was stable on 2202). So I had to volt my CPU up a bit to resolve that stability issue. I never had any unusual crashes nor BSOD, but the system was not perfectly stable I assume since January.

So now it's all resolved, my temps are fine, none of my voltages are crazy, and VR works great. Thanks again, Kronos35!