No V_Core Sensor?

HWiNFO64, CPUz throttlestop - none of them show a VCORE sensor. They only show VID values. I'm trying to OC with a static voltage, but without a VCORE sensor to confirm the resultant voltage makes the job difficlut. Could it be that my machine doesn't have a vcore sensor? (model below)

 

Yep you're correct all the LGA Clevos lack Vcore sensors

If you're really desperate to know I guess you could theoretically read it with a DMM off the capacitors on the back side of the socket (the other usual probe spots, the FETs, would be inaccessible underneath the CPU heatsink) but that would require disassembling the unit, somehow reattaching the heatsinks and the fans (since the case with mounting points is absent), and running it at 100% load while upside down outside the case.

 

So since I have no Vcore sensor is it unsafe to OC with static/override voltage?

 

Not really. The only way Vcore would be higher than VID under load is if there's load line calicration operating, which the Clevo LGA boards also lack. So you can be safe in assumption that the VID will always be higher than actual Vcore under load.

If you keep things under 1.4V it'll be ok. Then all you have to worry about is temps.

Generally, static is better for the benching / high end overclocks squeezing out that last 100-200mhz, but for everyday, adaptive with voltage offset is better for reducing voltages on idle/low load as it leads to improved power consumption, temps, fan noise, and long term component longievity.

 

Ok, so since VID is the rquested voltage, if I was to use adaptive offset, say -40mV, does that mean Vcore is 40mV less than VID?

 

When I was overclocking my desktop few years ago Vcore was just the number: a) to remember at what point computer will crash
b) to remember to keep bellow certain max level, or CPU will burn/shorten it's life. At the end I found out, the most stable overclock was with small undervolting, not overvolting, since heat was my biggest problem.
Usually set voltage is little higher than actual vcore due to losses and load voltage drop, so if you don't go crazy and test little by little you should be fine, since computer will crash to let you know settings are getting wrong, but overclocking always have some risks vcore or no vcore reading, since you're exceeding maufacturer's specs. BTW I'm just curious, what vcore readings would you be looking out for: some particular max do not exceed voltage? voltage drop/fluctuation under heavy load? or something totally different?

 

Initially I was trying to find the lowest stable static voltage for a particular overclock speed, because thermal throttling kicks in at 98C. But if I want to overclock the CPU for everyday use in mind, then I should be stress testing with adaptive offset setting right? With the goal of finding a stable, useable offset value.

 

No and No.
I have been doing some extensive testing these last few days, and I determined a few things.
On these laptops, there is no vcore sensor. I believe a few may have them, but I don't know which.
These laptops use VID instead.

Voltage overrides (manual voltages) also work differently on these laptops than they do on desktops.

On desktops:
Manual voltage overrides override the VRM voltage going to the CPU, with a custom user set voltage amount going to the CPU, which is then manipulated by mOhms of Loadline Calibration voltage droop. VID is completely ignored, but is used for power measurements (CPU Package Power) on most desktops, unless the MSR is remapped to VRM power measurements instead.

Auto / Offset/ Adaptive (if available) voltages function similarly to laptops, with the exception that Loadline Calibration is still available to manipulate the voltage signal after it is set by the AC loadlines, boosted from the default VID for that core/cache frequency.

On laptops:
Manual voltage simply reprograms a new CPU VID into the processor and VRM directly, overwriting the original hardwired CPU VID. This cannot usually be done on desktops.
Since this is VID and not vcore (core voltage), the AC and DC loadlines still function on this manual VID. (DC loadline will affect vdroop by the value of DC mOhms of resistance * Amps current draw= vdroop in mv subtracted from the VID after AC loadline boosts it. Then this voltage is affected by VRM vdroop, with whatever hardwired value the ODM decided to use, assuming they didn't deviate from Intel defaults. The VRM usually ignores the DC loadline however, and receives the voltage after AC loadline bias).

Adaptive/Offset voltages are the same as on desktops (original CPU VID is preserved; AC loadline affects the VID before it goes to the VRM as final voltage, with whatever vdroop is hardwired into the VRM).

DC loadline is used only for power measurements, but high values of DC loadline for some reason *DO* raise the voltage and temps slightly (it's possible a very high DC loadline value will boost how much AC loadline goes into the VRM, but the difference between a 0.01 mOhms DC loadline and a 1.60, 1.80 or 2.10 mOHms DC loadline seem to be about 5C only (maybe 15mv-20mv), while the difference between a 0.01 mOhms AC loadline and a 1.80 mOhms AC loadline is massive.

In most cases, if you want to see the original unaltered CPU VID for your processor, set AC and DC loadline to 0.01 mOhms. However the load VID will then be completely off (higher than expected) since the VRM will still have its own loadline.

If the ODM followed Intel's specifications, setting AC loadline to 0.01 mOhms and setting DC loadline to the SKU for that processor (1.6, 2.1 or 1.8 mOhms) will make the VID drop at full load, depending on current, and the idle VID should be close to the original pre-programmed VID for your CPU (which if set manually via an override is now your base voltage instead)

This works extremely well for manual voltages since you are writing a new VID to your processor anyway.

Note that "Default" Vids for auto/adaptive voltages stop scaling at the highest official 1 core turbo multiplier.

Having the AC loadline value kept at default mOhms will guarantee your CPU will be stable up to 100C up to the max turbo ratio, but will pump far more volts than the CPU may actually need to be stable.

*Unfortunately*, changing DC loadline has NO EFFECT on CPU vdroop and has NO link whatsoever to loadline calibration (which is missing). AC loadline is NOT loadline calibration whatsoever, but boosts the original "default or custom" VID higher up by a certain amount, depending on different loads and other bizarre criteria.

 

I don't have your particular laptop, but most laptops I came across have two major problems at high loads: heat and reliable voltage limit. Both can be improved by undervolting (not overvolting), repasting cpu, gpu and possibly upgrading fans, other cooling components. If I was you, first I would find reliable undervolting at stock speeds and test extensively over few days under heavy loads. If I wasn't thermal throttling at all and had some thermal headroom, then I would start experimenting with overclocking, still undervolted but little less, than running stock. There should be sweet spot, where undervolting is low enough for computer not to overheat and throttle but high enough to be overclocked without crashing. Just to give you idea my i7-8750 stock can be undervolted by up to -170mV, but it will crash about every 3-4 days even at idle, but at -125mV is rock solid, no crash in months. BTW how strong is your power supply? I just noticed your signature, GPU 185W plus CPU at least 70-80W, you better have at least 300W power supply.

 

Power supply is 300w. Did the whole delid+ liq metal+ copper ihs job so more thermal headroom is avilable. But I guess the only thing left would be to get a cooling pad. I found 4.4ghz is good at no less than -65mv, 4.3ghz at -125mv. 4.2ghz @ -135mv. Using x264 16T stability test. 4.6 ghz is fine at +45mv but the fans get kinda loud for my liking esp. in summer.

 

If we have temperature limit throttling then is it necessary to have power limit throttling in place too? Isn't the goal to throttle the speed when temp limits are reached regardless of power consumption?

 

Thermal throttling happens automatically, if you have a super custom BIOS with the setting exposed you can alter TJmax, but whatever it's set to you can't stop the thermal throttle at that point.

So yes, you could set a preemptive throttle by limiting power, but that would always be active so would unnecessarily constrain burst performance, so basically nobody does it.

I set a power limit slightly higher than what I observe is required, to protect against a sudden FMA load whacking the CPU harder than it can cool and be stable with

 

So from what I understand PL are there to prevent the CPU from reaching the PROCHOT throttle temperature, and this would ensure more consistent clock speed as opposed to clock cycling up and down due to temp throttle?