*** The Official MSI GP63 Leopard Owners and Discussions Lounge ***

How about just changing the epp value?
I keep it at 100-120 for normal uses
240 for downloading or movies.
Btw
What should be the values for the maximum battery life?

 

255 is max savings...

 

I suspect speed shift could be the problem but cannot be certain, can you try disabling the speed shift and then try your aida test just to be certain... i am trying settings that used to crash in warframe without speed shift and it doesnt crash anymore... maybe thats why notebook manufacturers avoid it???

 

@Phoenix what are your maximum temps?

 

Ok, after lots and lots of testing, i still cannot confirm it, as it still crashes randomly. As such, I am back to speedshift.

 

I couldn't solve the issue nither. However, the crash should never happen during realistic usage so it doesn't matter to me.

P/s: If I manually untick the boxes (GPU/CPU/FPU/Cache) before stopping Aida stress test, it doesn't crash anymore. So I was correct about the cause (sudden power drop), still no way to solve it.

 

This is one of the problems with undervolting.
When the load stops, the voltage drops down to an unstable point, if you are using ADAPTIVE voltage with cpu downclocking if you are undervolting too far for a certain voltage / mhz point to be stable anymore.

With a static voltage and no downclocking, maybe you would set 3.9 ghz on 6 cores with a -150mv undervolt over whatever "stock"voltage is (finding the true stock voltage is rather difficult without an unlocked Bios). For example if your stock voltage at 3.9 ghz on all 6 cores was 1.15v for example, you could then undervolt by -100mv (in this case, you are simply setting a STATIC manual voltage override rather than an "undervolt") for 1.05v. If the CPU then never downclocks, you get 1.05v at idle and 1.05v at full load. If you are fully 100% stable at full load (no WHEA Errors or crashes) you will always be 100% stable at stock.

It "is" possible to downclock to 800 mhz while maintaining voltage without having the voltage drop but I forgot how (This ONLY applies to static voltage overrides), but that uses more power.

For adaptive voltage, it gets tricky here.
If the default for 3.9 ghz is 1.15v, your CPU will be 1.15v at 3.9 ghz and lower voltage (based on the pre-programmed default VID) at every speed bin lower than 3.9 ghz. For for example, 800 mhz might be 0.700v, 3000 mhz might be 0.925v and 3.9 ghz might be 1.15v.

If you did a -150mv undervolt, this would make 800 mhz->0.550v, 3000 mhz 0.775v and 3.9 ghz 1.0v.
Any point in this chain (and there are OTHER speeds and preset voltages too that scale up with each step) might have an unstable point. If 0.550v is not stable enough for 800 mhz, you will error out.
if 0.775v is too low for 3000 mhz, then you may randomly crash at "light" load, but not at full load or at full idle. Every CPU is different.
Frankly I don't even bother with this because it's too much work trying to find where you are stable, unless you are on battery power and trying to maximize battery life. It's just easier to set it to full speed, find the voltage you need and keep it there. But those are for the AC power only people.

 

So what you mean is that because of undevolting, at a certain voltage/freq point, the system is unstable, right? However, the laptop only freezes when stopping stress test, whereras it's stable doing everthing else.

 

@Papusan

 

Is it stable and without freezes if you don't let the Cpu downclock when you Stopping stress test? If you haven't tried maybe time to test it?

Spoiler: HOW TO UNDERSTAND THE DIFFERENT VOLTAGE MODES

CPU Core/Cache Voltage: Sets the voltage control mode for CPU Vcore and the Uncore:

Manual Mode: Allows setting of a single value for Vcore that is applied across all Core ratios, irrespective of application load.

Offset Mode: In Offset Mode, we can add or subtract voltage from the CPU’s default voltage for a given CPU core ratio. The default voltage scales according to the active multiplier ratio. This provides power saving when application loading is light. The side effect to using offset mode is that any offset value we select will be applied to all core ratios. This can result in too much or too little voltage being applied for a given ratio, which leads to instability.

If you wish to use Offset Mode, then bear in mind that the Vcore displayed in the UEFI is simply a snapshot of the offset voltage stack; the firmware interface only places a partial load on the CPU. The full-load voltage in the operating system will be different, so you will need to check the voltage by running a suitable application within the OS. Use Ai Suite to monitor the voltage when the system is under full load. Also, bear in mind that the default voltage receiving the offset changes with the applied CPU ratio.

Adaptive Mode: Adaptive Mode was developed to account for the inadequacies of Offset Mode for overclocking. We use it to specify the voltage used when the CPU is faced with a heavy application load. The voltage we set is the maximum voltage the PCU is allowed to apply, which takes all the load-related guesswork hampering Offset Mode out of the equation. The other boon of Adaptive Mode is that it does not alter voltages for non-Turbo CPU ratios, allowing us to enjoy the benefits of power saving without the voltage adjustment range issues presented by the Offset Mode function. We recommend Adaptive Mode for all normal overclocking.

To use Adaptive Mode, simply enter the full load voltage you wish to use in the Additional Turbo Mode CPU Core Voltage box. So, if you wish to set 1.20V for full load, just type 1.20 into the box. The target full-load voltage is shown in the Total Adaptive Mode CPU Core Voltage area.

When using Adaptive Mode, configure the following settings within the Internal CPU Power Management sub-section:


Note that the Adaptive voltage target works on the Turbo ratios only. So, if you use a non-Turbo CPU ratio, the value in the Adaptive voltage setting box will not be applied. In such instances, use Offset Mode or Manual Mode for CPU Core/Cache Voltage.

The major caveat of Adaptive Mode is that the minimum possible voltage for a given ratio is pre-programmed into the CPU. If you happen to have a very good CPU that can run at a lower voltage than the minimum adaptive voltage for a given ratio, there are only two ways to lower the value. The first method is to apply an offset. That’s why there is the option to apply an offset when in Adaptive mode. The offset value is added or subtracted from the Additional Turbo Mode CPU Core Voltage box, and the total is displayed in the Total Adaptive Mode CPU Core Voltage pane. The side effect of applying an offset is that it affects the entire voltage stack – from idle to Turbo ratios, which can limit the usable offset voltage range.The second method is to use the CPU Load-line Calibration setting in the External DIGI+ Power Control section. Using a lower value will lead to more sag under load, resulting in a lower voltage. Again, the issue with this is that it will affect how much voltage the CPU receives under all loading conditions, which can lead to instability when it is too low for a given load state, or when the CPU transitions from idle to load state.

Even with those caveats, we still recommend using Adaptive Mode for all normal overclocking, unless your processor can run at voltage levels that fall substantially below the minimum adaptive voltage for the applied CPU core ratio.

 

Hi, new member here, I got my Leopard 8re about 2 months ago and I've been lurking this thread for a few days trying out some undervolt values etc and would like to share my findings.

First thing I did was run cinebench and this is where I saw that 3.9Ghz wasn't possible with the 45 Watt limit, only with a quick 60-70watt burst was 3.9Ghz/6 cores actually happening. Even though temps were in the mid 70's. I've come from a 3rd gen i5 which always had thermal throttling, it was rated at 35watts but never went above 29watts on full load. So this i7 is new territory for me.

I downloaded Aida64 and found it settled to 2.8-2.9Ghz @0.95v and 76°. I used ThrottleStop to gradually lower the core/cache offset whilst running Aida64 to test stability. The lowest voltage offset I found is -170mv, with this the cpu runs at 3.3Ghz @0.9v and 76°, the temps are great it's just the 45 watts that is still the main issue.

Next thing I did was combine the IMON values from page 5 (50 / -31000) with my undervolt findings and the cpu will run at 3.9Ghz on 6 cores however the temperature jumps to 95° and stays there. Aida64 reports throttling upto 15%. I tried changing the IMON offset to -5000 but it doesn't change much.

Reason I'm posting this is because I was gaming earlier saw some strange things;

-Whilst all 6 cores are active I saw 4 of them at 4.1Ghz at the same time, CPUID and CoreTemp both stated this.
-In the same gaming session I saw all cores sat at 3.9Ghz for a long time (10-15 minutes) whilst my voltages and Imon Bios offsets were back at stock values. Even the core temps were in the 80's.

Are 3.9Ghz at high load and 3.9Ghz in mild load two different things? I thought 3.9Ghz wasn't possible @45W but maybe the load plays a part?
I plan to use this laptop for video rendering so my initial aim was to have all 6 cores running at 3.9Ghz as advertised, anyone got a way to do this whilst staying away from the 90° zone?

Sorry for the long post.

 

Temps jump to 95C because the processor isnt throttling anymore. But you need to put on some decent thermal compound repaste like Coolermaster Mastergel Maker Nano *AND* change the thermal pads on the *CPU* (NOT THE GPU, the CPU) from the stock 1mm pads to 0.5mm Arctic pads (only $10 for a 145x145mm sheet on Amazon). This will give you a nice temp decrease without having to use risky Liquid Metal. But repasting these laptops require disassembly, unlike the Titan series, where all you do is remove the fans, and the heatsinks and bam: easily done.

 

What was your speed shift value during the test?

 

I tested with speedshift at 0 and 64, both caused crash.

 

I worry about the 90C zone too. I didn't want to do a repaste, so i had to compromise. I set my thermal limit to 88C. When stress testing in aida, the freq drops to about 3600Ghz. Depending on the ambient temps, it can drop to 3200Ghz. I am not sure if this can cause any problems...

 

Or just play games at epp value 150.
Temps may become less than 80 too.

 

You are better off by being power limited than thermal throttling, because you won't have so much clock fluctuations.

You can play with the imon offset and set a value that will limit your power draw just before you reach throttling temps.


@Le Quan , @sush33king , @Gursimran82956 , @Chipmunk642

With that imon offset, do you get the 95C temps in aida64 stress test with CPU+FPU+Cache+Mem and GPU? Or is it without GPU?

 

Look in the TS FIVR window and see if the speedshift value changes (assuming that you set it to 0 in the main TS window at speed shift epp) if you play with the undervolt.

If it changes, then there is a conflict between win power plan and TS epp.

With speed shift 0, does your CPU maintain max clocks all the time (even in idle)?

 

Under load, the voltage decreases, but the VRMs cant stabilize the voltage fast enough (especially if the difference between idle voltage and load voltage is v high) so there are fluctuations that aren't registered because they occur extremely fast.

The same thing happens when you cease the load on the CPU and it returns to idle. Those fluctuations can cause instabilities if they are too high.

https://www.anandtech.com/show/2404/5

 

Now you mentioned it, CPU clocks do not stay @ 3.9 all the time if I set speedshift at 0. Does it mean speedshift doesn't work properly?

Edit: I noticed the speedshift was disabled in BIOS, fixed it and now it works properly, max clocks if EPP @ 0

 

So it's just the VRMs aren't good enough hence I face such problems, and there's no way to fix it?

 

CPU temps reach 95C after 3-4mins of Aida64, everything is stressed, CPU+GPU+cache+mem+GPU.

Edit: GPU temps is never a problem, it's always around 70C at max load, it's the CPU reaching thermal throttling during stress test.

 

It is possible that there is a conflict for power management between win and TS. Did your speed shift value change from 0 in the FIVR window (under the voltage offsets) if you change the uv offset?

The VRMs are good, this is just the way they work and if you read that article I linked, you'll see how different settings (vdroop, voffset, load line calibration) affect those fluctuations.

I'll give you some links just in case you want to see in more detail how the power delivery works:





Due to the way that your voltage is managed at different loads and clocks, your UV might be borderline stable at max clocks, but not in all circumstances, so either you find the power management settings that can keep the max clocks all the time (you might need to disable all power saving features for this one, including c1e) or you reduce the UV value.

You can use two profiles if you want: the current one that might be stable at max clocks, but you'll need to set the power plan to use max clocks all the time with this one (see if it works by using win high perf. power plan and test without speed shift) and another one with a lower UV offset value that is stable under all circumstances (with speed shift, but you need to be sure that speed shift is properly controlled in TS and there isn't a conflict with win power plans or other management apps like dragon center for ex if you have it installed).

 

Thanks, I might hold off doing a repaste until I do a few more tests and decide if I really need the extra speed boost.

Just CPU+FPU+Cache for me, if I include the gpu (UHD 620) then the CPU clock speeds drop to make room for the gpu due to the 45watt power limit on the cpu package.

 

Thanks for all the info, I set win power plan to best performance, which now makes speedshift works properly. Doing Aida64 again to see if it still crashes, I'll report back soon.

 

With a unified cooling system, the GPU temps affect the CPU temps very much (reducing the available CPU thermal headroom under combined loads, because the heat from the GPU is shared with the one from the CPU).

Now, with the cooling design of the GPs and GEs, you don't have many shared heatpipes, but even one is enough and the fact that the CPU and GPU are in close proximity of each other, means that there is heat sharing between those two.

GPU stress in aida is not as intense as in some games or other GPU stress test, but it would be good to know how much does it impact the CPU temps in aida, so that is why I asked if those temps were under combined loads or just CPU loads.

At what temperature does the CPU stabilize (with your current imon offset) with just CPU+FPU+Cache+Mem?

EDIT:
And what is the sustained CPU power draw (total, by adding the imon offset) and the sustained cpu clocks (you can use the Unified tab instead of temperatures tab in aida stress test to see all the relevant values in real time)?

 

I doing Aida stress right now, 15mins in, cpu+gpu stress, cpu stablizes at 3.5GHz. Gonna turn off GPU stress and report back temps soon.

 

Your prochot temps is realy low, set it higher (97C should be default for most notebooks) so thermal throttling isn't interfering with the results.

Also, speedstep isn't used if speedshift is used and you can untick bdprochot.

You can even untick c1e and see if it behaves better (as in not freezing at the end of the test).

 

Here the result after turning off GPU stress for 10mins. CPU holds max boost @ 3.9GHz , Aida shows minimal throttling, power draw stablizes at 41-43w (with 15W offset, which means 56-58w), temps in high 80s and seems to settle around 85C (I set DB Prochot @ 89C and no thermal throttle seems to happen).

p/s: my TS setting, please comment if there's anything should be changed as well.

 

What does C1e do? It allows CPU to idle at lower power state right?

 

damn, I forgot to test if the system crashes upon exitting aida. Since I turned off GPU stress first the crash didn't happen, not sure if it's still stable otherwise.

 

Only the prochot temp higher and you can limit the max temps by allowing just the right amount of max power draw so it doesn't thermal throttle (you thermal throttling limit can depend on your preferences).

Thank you for the test! This means that even with stock paste, this notebook is a good substitute for a CPU rendering workstation, since it is able to maintain a high enough power draw to keep the max CPU boost with an undervolt and the temperatures are acceptable at that power draw (the situation should be even better after a repaste and with better thermal pad).

Regarding C states:

https://www.hardwaresecrets.com/eve...-about-the-cpu-c-states-power-saving-modes/2/

You should know that the CPU is constantly changing its clocks (extremely fast) depending on the energy saving settings, that is how a higher speed shift value makes the CPU to down-clock faster depending on the load (and the time it spent in various states).

 

I'll try and recreate this on mine as 3.9Ghz @85C is great, did you use a voltage offset on the cpu+cache in throttlestop by any chance? I've used similar settings (10W offset) and got 3.9Ghz but at 96C.

 

Yep, this is exactly why I made the purchase, best price/performance for 1.5K (i7 6core + GTX 1070)

With GPU included in the stress test, power draw was lower but temps are higher. That's why I decided to lower the prochot temp.

 

You can check my attached photo to see my TS undervolt setting. My laptop is stable @ -0.21mV, I also change values of iGPU and Analog I/O, not sure if they help.

 

It seems that the firmware is limiting the CPU power draw in combined loads (some of the reasons for this behavior are obvious and it shouldn't be a problem with gaming loads if the real cpu power draw doesn't fall below 45w, like it does with the razer blade 15 2018).

So is it still stable at the end of the stress test even if you tick GPU?

 

I wasn't doing Aida again, gonna test again tmr.

 

These are my latest results;

-170mv offset via ThrottleStop on Cpu +Cache
Dragon center set to Turbo + auto fans
IMON offset -15000

https://imgur.com/EUDkuRF

Going to try and tweak it a bit more as it is touching on thermal throttling territory and can't quite maintain 3.9Ghz....

 

What's your ambient temperatures? When running Aida I use cooler boost and my ambient should be ~24C

 

I think around 23C, I stuck with auto fans and not cooler boost because it's too noisy for long term use.

Edit: Does anyone know if the balanced power plan is different to high performance when doing a stress test? I tend to keep my laptop on balanced all the time and use dragon center to flick between turbo / eco.

 

I stick to auto fan and only use cooler boost for benchmarking or stress test. For me personally Speedshift only works properly with best performance in windows power setting. Also I uninstall Dragon Center, I use Silent Option for fan control instead.

 

I hit 95C with all activated, freq drops to 36xxGhz. With GPU only deactivated, temps stay below 90 with no throttle. Ambient temp about 23C with cooler boost.

I tried changing imon offset numerous times with diff slopes but it would not produce the desired results. Also, i'm not sure if it will adapt to ambient temps changes. Sometimes I have the air-cond on and sometimes not, temps ranging approximately from 23 - 30.

I decided to try using the thermal limit feature again, Tcc Activation Offset 12 (87C Limit), but this time I set the Tcc Offset Time Window to 500 ms. Now the frequency holds steady, not fluctuating, and somehow Aida does not identify the freq limiting as thermal throttle.



Below was what it looked like before activating the Tcc Offset Time Window. Here AIDA identifies the limiting as a throttle:



I am not sure if this is healthy for my system?

Edit: Fan Settings on Dragon Center Auto and ambient temp about 27 during Tcc Offset Time Window test

 

I keep mine balanced as well, i have yet to see the difference on my machine.

 

what does Tcc offset time windows do? With 500ms value does it mean it takes 500ms for the system to register the overheat, i.e. throttling?

 

It uses the average temp within the time window to determine the limit action. That's what i read it was.

 

If the system isnt thermal limit, why CPU doesn't clock higher? Does TS show thermal throttle during the test?

 

Nope. It does't identify it as thermal limit. TS also dont register. I am not sure why.

Edit: TS didn't register any limit at all.

 

What's the power draw? My CPU needs almost 60w to maintain max turbo. During stress test the CPU uses about 50w and clocks ~3.5-3.6GHz

 

These are all the screenshots of the test i did. I am not sure how to calculate the power from the imon figures.

 

Can you tell from the screenshots what my system's power consumption pattern is like?

Edit:
Oh, i didn't take screenshots when i tested with thermal limit set to 95C.

Ill do that after dinner.

I wouldn't use time window if the thermal limit is close to 100C as the limiting action do not take into consideration spikes in temps, as the average value is used.