X99 - 6950x issue or no?

[alasdairvfr]

Hello, I'll start by saying that while I have been overclocking and monitoring processors for many years, I will admit that I don't know everything and need some help.

Different architectures/generations of CPU/chipset really do mean approach to overclocking and monitoring change really dramatically; a few common truths never change (keep temps, voltages as low as possible, etc)

I recently got an MSI x99a sli plus board and a 6950x on the cheap. Pretty happy with it so far, I had to flash a beta BIOS to get OC since the spec-ex vulnerabilities these chips are affected by caused Windows to bork OC for some reason. I got myself on a nice 4.3GHZ all-core on 1.255 volts. I do my workload and have hwinfo64 running on another screen as is my way for many PCs. Massive props to Martin, I never posted here or asked a question before but have lurked here for MANY years.

I look at VID, this tends to respect what I ask of it. Whatever voltage I set in the BIOS, it's there. That makes sense but VID doesn't tell me what my CPU voltage is. I look at the CPU itself and there are no numbers for vcore.

I'm looking at the "MSI X99A SLI PLUS(MS-7885) (Nuvoton NCT6792d)" section and it's suggesting my Vcore is 1.440v which is frighteningly high. It does seem to be creeping up over time

CPU-Z reports 1.249 v
AIDA64 reports 1.249 VID and 1.44 vcore
Throttlestop shows again the VID.

Despite what my BIOS says, I set the voltage in XTU to 1.25 in the OS (I don't use SW to OC but trying to get to the bottom of this) and the VIDs do respond but the vcore does not

I know the sensor reading in HWINFO is the socket voltage, so the actual CPU would be lower by a very tiny bit. When i set the OC, the Vcore was also at 1.25 but seems to have "grown" overnight

The temperatures are low, they are idling 31C with ambient temps maybe 18-20C so I'm really at a loss... I want to believe I have an erroneous/ inaccurate reading because the rest of the temps and the fact the system seems happy otherwise

edit - typos

 

[alasdairvfr]

30 mins later its crept up to 1.448v

These are my detected settings from hwinfo64

 

[Martin]

These CPUs feature a Fully Integrated Voltage Regulator (FIVR), which means that the mainboard's VRM supplies only VCCIN voltage (~1.8V) to the CPU, which then provides the respective Vcore internally.
So monitoring of Vcore here is not straight as the mainboard doesn't handle it and it seems to rely on a Vcore sense pin from CPU. The Vcore shown in HWiNFO is the reference Vcore voltage provided by MSI, but I wouldn't bet on accuracy of this value; many vendors report here the VID instead of Vcore (that's the case of CPU-Z for example).

 

[alasdairvfr]

So you think the reported Vcore is settling in as an average between VID (which reflects actual usage which is my static 1.25 or whatever i manually set) and the supplied Voltage @ VRM which is in my case VCCIN-1.904v. I have rebooted it a few times but I wonder if it would keep climbing to 1.575.

Still trying to identify if there is a risk of imminent failure due to this or my trust-vcore-not-vid mentality is causing me undue stress lol.

My observations thus far - the VCORE value starts off where I set it in BIOS and it creeps up over time. It's a line in a graph so easy enough to ignore so long as it's a red herring and not indicative of reality (I got this system cheap as chips but don't want to have it burn itself out)

My CPU @ 3-10% usage temps are still in the 30-33C range, whichever core spikes making total% hit 10% might be in the 40-50C range till it drops in usage.

I plan on rebooting, check the voltages, run a few CB20s then try manually upping the voltage to 1.325 or something higher-safe and do the same. I can then compare the temp ranges to see if my current temps/reboot temps/75mv higher temps have deltas. If nothing is amiss, my temps should increase measurably. Then decrease again as I'd bring the voltage back and to one last run.

You think this is logical as a sanity check?

 

[alasdairvfr]

Well did a few CB20 runs @ 1.25, 1.34 and yes 1.34 runs ~10C hotter as I expected. A simple reboot, the Vcore was still showing 1.49 or so but mucking with BIOS the Vcore is back to being aligned with VID. Now all there is to do I guess is let it be and see if the Vcore creeps back up past 1.34, do a CB20 and check the temps. Its peaking ~80C on hottest core @ 1.25, almost 90C at 1.34.

Thermodynamics, if this thing really is up like 100mv, a CB20 run would bring it 10C higher. Then if temps are the same I can feel comfortable to ignore the issue.

I also reduced VCCIN from Auto-1.9v to fixed 1.8 since I think 1.9 is overkill for a moderate OC