Only my view and opinion based on what I've read and what I've experienced.
2x R7 1700, batch UA 1706PGT (1st CPU) and UA 1709PGT (2nd CPU). 3x RAM kits, CMK8GX4M2A2400C14 , BLS2C4G4D240FSE , F4-3200C14D-16GTZ. Same mobo and rest of HW.
The Corsair kit I could get upto 2666MHz, on 1st CPU I would need SOC 0.900V in UEFI. I would get errors in HCI MemTest, intermittently/variable time periods. Slacking timings, increasing VCORE/SOC/VBOOT&VDIMM did not resolve errors. G.Skill kit 2933MHz not an issue. 3200MHz even with SOC 1.050V in UEFI no go, VCORE upped no go. I have not tried to OC the Crucial kit yet.
2nd CPU I did not try any other RAM yet than G.Skill set. Again increase of SOC using my method of assessing default needed only +100mV.
So I would concur with Silent Scone that we may not need SOC as high as peeps going. IF SOC increase is stabilizing RAM for another perhaps it is masking a non optimal other setting or just the immaturity of platform. As RAM has an effect on Data Fabric clock this is another thing to consider. Perhaps some CPUs do not like this. Perhaps May update will "tweak" this aspect to enable higher RAM use.
The Stilt's share on SOC, link
, another post
to ref and this one
So the take away for me is SOC doesn't need to be as high as we may think we need. RAM training is an issue, immaturity of platform and as he states extremely complex training. From having read past posts of The Stilt he has various Ryzen boards and a fair number. He also has Intel, as I've seen his posts where he has said "I will test x and compare".
Then SOC may also be affecting other voltages we are unaware of, link [email protected]
Then his post
in context of say cold boot vs warm boot.
For me I get occasional cold boot issues on 3200MHz, standby/reboot not an issue. I have thrown Y-Cruncher / x264 / IBT AVX / RB / [email protected]
at system many hours and "back to back". One such run was 36hrs+ with these programs. So my failure to cold boot is either a setting I'm over looking or platform immaturity.
Today I plan to retest ProODT, this has been highlighted by Elmor, Praz and [email protected]
to aid RAM training.
As I use a PState 0 OC with Offset voltage to CPU on a failed memory training boot there is a "quirk". AMD CBS resets, Extreme Tweaker does not. Then as CPU is in "Default Mode" PB/XFR is active even when UEFI loaded. My R7 1700 uses ~1.35V in these higher states, throw in the +162mV offset I use = ~1.5V.
Screenshot of UEFI and getting ~1.5V is in this post
, it is not just UEFI reporting 1.5V I have checked with DMM on ProbeIt point. Solution for this at present, for me is this