Recommended CPU and GPU undervolt settings for Clevo NH55RCQ?

Hi, I am a user of Clevo NH55RCQ

My specs:

CPU i7-9750H
GPU GTX 1660ti
RAM 16GB
Windows 10 64 bit

Since I have no plan to overclock, stock power + low temperature is my goal so I'm trying to undevolt both CPU and GPU.

I'm totally new to undervolt here, and trying to use MSI afterburner and throttle stop to undevolt.

I heard that even same CPU/GPU will need different settings base on the system. Worried that the machine might become unstable, can somebody recommend the best settings for both CPU and GPU?

This is what I wanna try (base on some tut videos I found on youtube):
---------------------------------------------------------------------------------------------------------
*CPU:
- Speed Shift - EPP is on and set to 0
- BD PROCHOP is disabled
FIVR settings:
- CPU core and CPU cache both set to -125.0 mV
- Turbo Ratio limits: all cores set to 34
TPL settings:
- Turbo Boost long power Max/ Turbo Boost long power Max all set to 24
- Speed Shift: Min set to 1, Max set to 34

*GPU:
Drag all the points in the coreclock curve to a flat line equal to 1365 (the lowest available in default settings)
---------------------------------------------------------------------------------------------------------

With this settings will it be safe? Or is there any better settings?

Thanks in advance!

 

Hi. You are using the correct products (MSI Afterburner and ThrottleStop) for this. As you heard, the lowest possible voltage value will vary for the CPU and GPU from one system to the next. There are general settings (like -125mV on the CPU) that are going to be modest enough to be stable across the vast majority of systems. Some might go lower, and some might tolerate a little less voltage reduction. Temperatures also affect undervolting. Higher temperatures require more voltage. As temperatures climb, so do the voltage requirements. Cooler temperatures generally allow you to run stable using less voltage.

Since you are using Windows software, no permanent changes are being made at a firmware level. Nothing you set for undervolting using software will be "unsafe" per se. The only thing that would happen is if you undervolt too much it won't be stable. You will experience problems like BSOD, freezing, etc.

To determine the maximum stable undervolt for the CPU and GPU requires trial and error testing of each component. You can use something like Cinebench R20 to test CPU stability and run a Fire Strike stress test for the GPU. Find the best setting for the CPU first, then move on to the GPU.

For the GPU, you don't really need to move all of the points on the graph. Some people do it that way. It's more time consuming and tedious, and I have not found any advantage to doing it that way. Here is a video from my YouTube channel that might simplify the GPU undervolting process for you.

 

Note cooling the gpu could also speed it up as well so while temperatures might remain the same performance would go up. Depends where on the power band it sits.

 

Hi Mr Fox, Thank you so much for your reply! I watched your video, it makes GPU undervolt work so much easy now!

Guess I have to test a round to find a stable one.

There are 2 questions I also wanna ask if you don't mind:

1. There are many stress test software around, I'm using AIDA64 trial to test both the CPU and GPU. Will it work correctly?

2. Usually when running stress test, how long is proper to know that "it's fine" for the hardware?

If it's possible I really wanna use stress test software that generates results within 30 minutes (the machine either keeps running or got BSOD)

Oh that's great! More reason to undervolt then

 

Hi. Glad my video made it easier.

AIDA64 is very good. I use it.

I am not a fan of torturing parts for hours. I think doing so is silly and unnecessary. Even thirty minutes is too long for me. If my CPU passes Cinebench and GPU passes Time Spy Stress Test then I am done checking stability.

 

Thank you so much for your help, you just spoke out my mind! Even my machine is new, I hate to have my CPU / GPU stress out too long for result.

Since you are experienced so I really wanna follow your way to undervolt my CPU and GPU.

This is the work flow that I'm gonna follow. Do you think it's fine?

Undervolt CPU
→ test CPU on Cinebench
→ If CPU passes the test the undervolt GPU
→ test GPU on 3DMark Time Spy test
→ If GPU passes the test then DONE!

 

You're welcome. And, yes... you got it. Simple, safe, and effective.

 

Thank you for confirming! Just finish downloading 3d mark and Cinebench, now I can begin

One more question just in case...

For throttlestop, I heard that it has be run all the time in background to keep the CPU undervolted.
But for MSI After Burner, is it the same? Will my GPU's voltage return to normal the moment I close it (totally from system tray)?

 

As far as I know, once both ThrottleStop and MSI Afterburner have set the voltage values those settings will last until you reboot. I am not aware of an example of that not happening. The exception would be if your GPU driver crashed or there was a TDR, and that would reset the GPU clocks and voltage to their firmware defaults. In that case, you would need to reapply them using MSI Afterburner.

 

Or you hit the GPU reset combo on the keyboard lol. Will be set until overridden or rebooted.

 

Thank you Mr.Fox and Meaker!

Now I'm encounter some problems, need some advices though.

I finished undervolting my CPU and CPU. But now BSOD occurs randomly when NOT RUNNING BENCHMARK TEST, does it mean even if they all pass the test, I'm still undervolting them too much?

I got BSOD of something like store data corruption. This just happened after undervolting so I guess it is the reason?

My setings: -140.6mV for 9750H and 900mV (stock 1050mv) for GTX 1660ti

With these settings Cinebench and 3Dmark ran without problems until it finished (I tested a couple of times)

So I guess going back some voltage is the solution?

EDIT: one more big problem is, MSI after burner keep moving the curve on it own after I hit apply!! Some times the clock moves up, sometimes it moves down to 50mHz!

Will this cause any trouble? It make things hard for me to know which voltage/clock is the best if it keep moving like this.

EDIT2: after a couple times of trial and error, now I know that -140.6mV IS UNSTABLE, and -138.7mV IS STABLE (at least until now no unsual thing happen with this voltage). -For 140.6mV I will have a BSOD AFTER the benchmark test (almost every time), and -138.7mV runs with no problem after continous test on 3D mark.

For the GPU, MSI afterburner show different clock speed everytime in the curve so I have to search a bit about GTX 1660ti, now I run with 1905mHz (maximum is 1920 mHz but I wanna leave some room just in case) along with 887mV.

This settings is the most stable until now with no BSOD or unsual things happen. I hope this can keep on stable because after this until the day this machine die I probably won't do any benchmark test.

The temperature maximum for CPU while benchmark is 91C at peak (2 cores and the package exceed 90C at 90~91C) and I didn't use any cooling pad or laptop stand, just placing the laptop on a flat table, and for GPU it is 72C.

And this is the score on 3D mark with the above "stable settings"

 

Undervolting by dynamic offset will apply under all circumstances, often the limiting factor can be the idle voltage. My 8086k goes down to under 0.6v at full idle.