Same as the first post, stock 2133 on my memory. I had to turn the offsets back on, here they are with both the Vcore and vSoC at a positive 0.0125.
Thanks for the info, it is probably going to help. Some surprises here.
First your VSoc looks dead on good, min and max.
Now your VCore at max is okay, but you have to ask yourself how often is it at max and how many cores are being dispatched when it is at max. You can get a feel for this in Ryzen Master if you expand the two CCDs.
Now lets talk about your GSkill Memory sticks. I understand that they currently are running at SPD and not overclocked. However most memory sticks running at SPD require 1.2V and yours is getting 1.196V at max. Now that's close. However when overclocked I believe GSkill allows you to push 1.35V through these sticks. So I think that while you are still running at SPD it would not hurt to push say 1.24v through the Dimms.
I have to hand it to GSkill though. The balls of them. Take some memory with an SPD rating of 1.2V, 2133MHz, tell the user to juice it up to 1.35V and expect it to run @ 3600MHz.
For some contrast I normally run Kingston ECC @ 1.2V - Their 2666MHz sticks I can bump up to 3200MHz with no additional voltage, and their 3200MHz sticks I can bump up to 3600MHz with no extra voltage. That is like 2 extra steps for each.
GSkill is asking you to push the voltage to 1.35V, so you can push 2133MHz right past 2400, 2666, 3000, 3200, 3400 and go to 3600! Like I said Balls. Take someone else's product and do what the user could have done himself. Oh, I'm sure they binned and cherry picked the best. LOL
Again for the time being keep the GSKill at SPD, you don't need the additional headache while trying to get stable.
Now your Title says the problem happens at idle. However, now looking at the latest figures you posted, it appears that I may have been looking at this all wrong. You are getting a tremendous boost. This is a product of the higher voltage. Do your minimum clocks ever fall below base? They should.
Ryzen is very funny, it can ramp up in a couple of milliseconds. It is very possible that your problem is not at idle but at boost frequencies. (I am tempted to tell you to take back the VCore +0.012 increment) Keep that on for now, but I think that particular adjustment wasn't necessary in this case). Your VSoc is fine with the .012 adjustment.
Lets trim you back from boosting to those highest frequencies.
PBO to advanced.
Limits to Manual
PPT change from 142W to 125W (As the system approaches 125W, PPT will discourage higher clocks)
Of course keep Thermal Throttle temp to whatever you set it before (I think I told you but I forget)
Oh, yeah. I found that a little odd, but I also use HWiNFO64 instead of HWMonitor. It shows the "Effective Clock" rate down quite low. I've never had to get this down and dirty with monitoring PC performance and details, but I assume this is it doing so? I also am aware of the HWiNFO64 crashes from earlier this year. I make sure mine is up to date and I don't have it run at startup, I run a stand-alone .exe and only run it when I'm actively checking something.
Here's a picture of it showing the effective clocking down. I've also just this morning set my DRAM voltage to 1.24, and set that cap on the PBO PPT to 125. I'll let you know if anything happens. Again, I appreciate the effort you're putting into helping me and others here. Way more than AMD's outsourced e-mail system has done.
