Yup exactly that.
Im waiting for more concrete data and EFI updates to come out before I really update this thread. The release has been pretty crazy and rough around the edges
Yeah UEFI updates are needed.
Because pretty manny moboās seem to suffer from similar symptoms.
And that is the high voltage on idle.
1 Like
Yup especially with low end motherboards.
I mean digital pwmās on higher end boards are there for a reason.
1 Like
Early indication on a low end board (not disclosed, possibly one of their next videos)
1 Like
Yeah the VRM is gonna throttle. I would be curious if setting it to C.probe and running all phases on would balance the current loads and even the temps out but it might blow the circuits
Good Lord! I was feeling there is something funny with 1.486 Vcore ā¦
Iām on a X370 Krait from MSI that was feeding a lot of voltage to my ex 1600 Ryzen, but when I installed the 3600 (non X) I was amazed to see 1.486 ā¦ for a damn 6 core! After I read this post ā¦ I was thinking for a X550 when they arrive in Q4, but I think I need to go for a x570 asap. Leaning towards an Aorus Elite.
But until then, at least for a MSI board, I had to set the offset negative voltage to -0.1. The result is that now I see max Vcore 1.384 ā¦ should be fine until the X570 arrives.
Again, thanks for the post. It confirmed what I was thinking.
1 Like
Well which will be better for 3950X+O.C ? Aorus Master, Crosshair Hero or Meg Ace???
1 Like
Vrm wise those 3 boards are more then decent enough for a 3950X.
The Gigabyte X570 Aorus Master basically is the weakest of the three,
when it comes to max current capability.
However the Gigabyte does has the nicest control scheme of the lot.
Because itās a true 12+2 phase design.
But they use slightly weaker 50A IR3556 powerstages.
The Msi X570 MEG ACE is using a doubling scheme 6+2 phase doubled to 12+2.
But they use 60A IR3555 powerstages.
The Asus Crosshair 8 Hero basically has the most powerfull vrn,
it is only a 7+1 phase but with doubled up components per phase in paralel.
Which mean that they use 14 IR3555ās in the Vcore part of the vrm.
But in all honestly non this really matters at all.
Vrm wise you will not really notice any difference between those boards really.
Because they are all plenty powerfull enough.
It will be more of a matter how decent each brands bios is, and of course,
how great of a chip you get.
I would personally advice to just compare the boards based on features.
The boards that offers the best feature set for the best price,
is likely the board that i would go for.
Because Ryzen doesnāt have much OC headroom anyways.
4 Likes
And after reading this thread - donāt shit your pants when you boot it up and see 1.5 idle Vcore.
Itās what you are going to see on all working motherboards using AMD settingsā¦
The thread is dated to July before we were really aware of this. Thank you for the updates it will be updated when people have time.
this isnt really a problem on the current agesa.
1 Like
Im going to establish in the thread now that this is completely normal bagaget has been on it. Hey say @Bagaget since youve been so kindly active about it how about you make a thread on the New Ryzen load line and how voltages are suppose to look etc. 
@MisteryAngel this section is updated. Hopefully to Bagagets liking as well. If you are going to make a New Generation threadripper board wiki I would suggest that you state something along these lines. The thread-ripper processor will most likely behave similarly
1 Like
I was trying to get time to collect more info but most of it is covered by Robert hereā¦
1 Like
Link added to OP
1usmus (ryzen dram calculator guy) continues to explore windows scheduler and CPPC behavior.
What I take away from the article is that many BIOS still doesnāt default CnQ, CPPC, C-states to Enabled and they really should be for best behavior.
Check your settings!
Also interesting pertaining my small vcore crusade on this forum is the clock stretching explanation from HWInfo64 author.
The mechanic is that if the cpu receives lower vcore than it asks for, I.e vdroop OR too much negative offset. It skews the clock so that in effect it down clocks and you get lower performance. Depending on how frequency is measured this might not be āvisibleā unless you bench performance.
Iām not sure how this works if you set a manual VCore. The problem is that a low nice manual Vcore is lower than the max ryzen whatās to boost properly and you will probably get clock stretching.
And at the same time this ānice low VCoreā is still higher than what ryzen use when itās not trying to boost a light thread.
Soo for a typical use case the low manual Vcore will average higher than the Auto behavior.
The thing to take away is update windows and bios, enable CnQ, CPPC, CPPC proffered cores etc because auto might not be enabled.
Only use manual Vcore with manual OC.
While undervolting - check performance!
If it goes down and frequency looks normal you are introducing clock stretching.
3 Likes