I thought the IMON offset indicates the offset in power draw but It seems like I was wrong. I set my IMON offset at 15000 and I got max power (registered by Aida) similiar to yours at max boost, i.e. ~42W.
During stress test my CPU power is registered as ~36w, which is 6w less than the max power draw. Yours stablized at 28-29w, which is lower than mine, probably due to your lower prochot limit point.
Edit: below link is my power drawn (TS measured) after clock speed stablized.
http://forum.notebookreview.com/attachments/capture-png.161804/
-
Offset 0 and IMON 5 = Offset 31000 and IMON 50 in terms of power consumption pattern.
Edit: GOing for dinner now... will come back and take screenshots of stress test @ 95C -
I wont be using time window when testing @ 95C... its scary as individual temp spikes are not catered for.. only avg of all temps registered within time window..
-
Ok, done 30mins of Aida, first 15mins was stressing everything, second 15mins excluded GPU.
Stablized @ 3.6GHz, min freq was 3.3GHz
Power stablized @ 36-37W, max was 44W (this is just for reference of power difference)
Voltage max 1.1V, min 0.9V, average of 0.995 (I assume voltage value is measured correctly, hence can be used for comparision between systems, unlike power drawn which is modified by BIOS)
Below are screenshot after turn off GPU Stress.
Power now went up and stayed at 41-42W, full turbo @ 3.9GHz. Temp went down and seemed to stable at 86C.
@sush33king
P/s: Fan is at Cooler Boost, ambient is ~24C. I also try to change the IMON offset value and observe no difference in power measure. Not sure how all the value related. -
@sush33king , @Le Quan
You have your answer here:
You need to test again with GPU ticked, because the thermal throttling limits your clocks, which in turn limits your power draw.
If you want to test if your power draw is limited in combined loads, you have to test with a higher thermal throttling limit (don't forget to use max fans).Last edited: Aug 1, 2018 -
Also what I am confused about is the exactly power drawn of CPU. Different value of IMON offset translates to same CPU power measured by TS and Aida, so I'm not sure how to calc the exact CPU power.sush33king likes this. -
Actual Power Consumption (Watts) = IMON Scale * Slope / 1.673 ? -
-
Ask @Falkentyne, @raz8020 , dear Prophets, how do we calculate the actual CPU power draw based on the IMON settings? My laptop is lying to me, i don't trust what she is telling me. -
The IMON slope offset works ONLY WHEN the CPU is under stress, otherwise you wouldn't be able to sustain 3.9Ghz in aida with CPU+FPU.
When your load is combined (CPU+GPU) the imon offset is most probably ignored and the CPU power draw and the total power draw is controlled elsewhere (most probably by the EC, which can override any other power limit).Papusan and sush33king like this. -
-
-
Did a test at thermal limit 95C.
CPU = 38.3 watts
CPU+GPU = 38.35
Its the same.
CPU Only power
CPU Only Clock
CPU+GPU Power
CPU+GPU Clock
raz8020 likes this. -
Falkentyne Notebook Prophet
Imon SLOPE is a multiplicative factor. e.g. 100 = 1.0x. So 50=0.5x. So this should be lower than 100.
IMON OFFSET is a positive or negative increase of the Imon with an amount that doesn't make sense unless you work at Intel. But increasing this slowly makes the CPU increase its reported power consumption (but not real power consumption or amps draw) and decreasing this (NEGATIVE) makes it report less. -31999 and 50 give the most reduction.
The Embedded Controller can read the amps directly from the PSU line, so changing these settings has no effect if the EC is trying to throttle you or increase its "battery boost" because the CPU is using too much power or throttling the CPU because you exceeded a total AC power draw limit,and so on. This is only for preventing "basic" TDP throttling. The EC can override anything it chooses.Papusan, sush33king and raz8020 like this. -
raz8020 likes this.
-
raz8020 likes this.
-
-
Falkentyne Notebook Prophet
It's based on CPU VID and amps. Or rather, vcore and amps.
Power consumption also seems to 'ignore' the "IA DC loadline" setting as this only droops the reported VID (vid drop), but the actual VID going into the CPU itself, while the IA AC setting actually boosts the voltage the CPU is actually USING and then reports that as a "baseline" VID, before the IA DC setting is applied to that.
Actual power draw is (Amps * volts) = Watts.
Good luck finding out how many amps you're using....Papusan, raz8020 and sush33king like this. -
-
Here you go (start from page 38):
https://www.intel.com/content/www/u...e-voltage-regulator-down-11-1-guidelines.htmlPapusan and sush33king like this. -
raz8020 likes this.
-
Sush33king, on your "CPU only clock" graph you left the gpu box ticked so the gpu is also under load which will alter your results.Hence why your 1st and 3rd graphs show identical power usage (38.3watts vs 38.35watts) as its the same test setup in both. Also if you view the "unified" tab in AIda64 it will show clocks, temps, wattage and volts all together making it easier to narrow down the issue.
To get 3.9ghz on all cores and no thermal throttling you will need to do two things, set the IMON offset to between -10000 and -15000, and undervolt your cpu cores + cpu cache. The offset will give the cpu more wattage legroom as this cpu needs around 60watts to maintain that speed on all cores. The voltage offset will lower the core temperatures meaning higher clocks can be achieved. Also due to the relationship between volts/amps and watts (volts*amps=watts), lowering the voltage will allow a higher amp amount to be used whilst sticking to the same watt limit.
On my laptop I've got the offset on -13000 and the undervolt set to -170mv. I still get a bit of throttling (less than 10%) and need to tweak it a bit more to get 3.9ghz and no throttling. -
@Chipmunk642
Oh, sorry. Heres the correct single cpu screenshots. I dont have the unfied tab, i think coz its a trial version.
-
-
Edit: My undervolting cant go any lower that 156 mv. Thats the best so far.Last edited: Aug 1, 2018 -
3.9ghz, sits at 93C, -11000 imon offset and -170mv, yet it's hitting 41watts power usage and 5% throttling now. -
imon slope = 60
Imon offset = 31000
Really cold aircond room
laptop cooler. -
Whats your IMON offset value when you're doing the test? I wanna compare the power comsumption between our system. @sush33king @Chipmunk642
-
-
-
-
Edit: I'll change the slope to 60 and revert back to -31000 and update this later.Last edited: Aug 1, 2018 -
-
-
-
I was ready to make it a cold day at home today coz of the tests @ 95C. Otherwise it will surely throttle.
-
Okay so, -31000 Imon offset + 60 imon slope, fans on auto, cpu package is at 90C but the cpu is spending more time at 3.8ghz than 3.9ghz even though throttling is very low @2%. How come the throttling has decreased with a higher negative imon offset? I imagined it would be the opposite as -31000 offset allows the cpu to pull more power over a -9000 offset.
-
Falkentyne Notebook Prophet
There is no vcore sensor on these laptops so its impossible to know EXACTLY how much voltage is going into the CPU.
VID isn't voltage. VID is what voltage the CPU *thinks* it's getting or is requesting from the motherboard.
The IA AC setting affects the voltage lines coming directly from the voltage regulators from the AC adapter, and this value is "boosted" up by a "gain" based on resistance (current) of 2.10 mohms on the Z370 chipset (although MSI uses 1.79 or 179 in bios)--the higher the current the more the voltage is boosted internally. This has a direct impact on "true" vcore but there is no vcore sensor on these laptops. Setting this value to 0.01 mOhms (1 in the bios) prevents the voltage from being boosted at load like this.
The IA DC setting is an intel calibrated setting to measure "VOLTAGE DROOP" (which is the amount the voltage is "supposed" to drop at high load). The droop is supposed occur on the voltage regulator side, and the VID you see is based on the result after voltage droop. The CPU doesn't droop the voltage, this is supposed to be done by the mainboard. The IA DC setting, which is also reference value of 2.10 mOhms on Z370 chipsets (but MSI uses 1.79 mOhms or 179 in the bios), shows how much vdroop is supposed to be applied.
The problem is, when you guys are using the 'stock' bios settings, at full load, the VID you see is substantially LOWER than the actual voltage going into the CPU. Often by about 100mv. The "high" voltage spikes you wind up seeing (like 1.4v or something) is the actual real voltage you see at maximum load, but the current VID is much lower. The reason is, MSI uses something called "loadline calibration", which functions on *VCORE* (remember these laptops dont have vcore sensors) to remove voltage droop on vcore. VID is not vcore.
So the "droop" that you are supposed to be seeing (e.g. the VID you see at full load) is LOWER than what the real voltage is, because MSI is REMOVING the voltage droop on their onboard voltage regulators themslves (This is NOT the IA DC setting. this is "loadline calibration" which only desktop boards allow you to toggle). However IA AC and IA DC are designed to function with Loadline calibration *DISABLED*, not enabled. So the IA DC setting, which is causing the VID to droop, is giving you what the voltage "would" be if there were no Loadline calibration at all. The IA DC has no way of knowing if LLC is in existence or not.
That's why on these laptops, setting IA AC loadline to 1 and DC Loadline to 1 works so well. Since there is no vdroop (or very little), there is no need for IA AC to boost the voltage by 1.80 mOhms at full load (which is why you need to undervolt these laptops--undervolting is simply removing the effect of the 1.79 mOhms of IA AC voltage boost).
Clevo and eVGA laptops don't use loadline calibration at all, so setting IA AC loadline to 1 on these simply makes them crash (because the voltage droops too low). -
-
-
@Chipmunk642 I set IMON slope to 50 and no more power limited. CPU stablizes at 86C max @ 3.9GHz without GPU load, power ~42W. I got -215.8mV core, which is quite fortunate. Test all done with Cooler boost fan. -
Image below is -31000 offset with slope set at 50 this time. TS showing Thermal and EDP flashing red again, Aida is showing more thermal throttle than before.
Edit: I thought -170mv was lowest undervolt possible because -180mv crashed instantly, yet I jumped to -200 and Aida hasn't crashed yet.....Last edited: Aug 1, 2018 -
-
Anyone has problem that Speedshift EPP value doesn't stick when laptop is on battery? For em EPP value in TS only applies when plugged in and with best performance option in Windows.
-
My freq remains max after pulling the plug at epp 0.
-
-
-
sush33king likes this.
-
@Chipmunk642
Did you manage to stop the throttling?
I think this guy may have the solution for your problem.
Last edited: Aug 2, 2018
*** The Official MSI GP63 Leopard Owners and Discussions Lounge ***
Discussion in 'MSI Reviews & Owners' Lounges' started by Spartan@HIDevolution, Apr 3, 2018.