Any help for improving the situation? Another 8750h throttlestop optimization attempt

[Quack991ˇ]

Hi all,

Finally decided to make an account here after couple years of tinkering and lurking around the forum following a lot of posts by 8750h users and the notorious Throttlestop uncle :=)

History/current setup:
I'm using MSI GE75 8SF I7 8750h model laptop since ~2019/20.
Reason for all this tinkering is the poor factory default thermals, dissapointing performance, poor airflow design for a gaming laptop and paste quality/method used by MSI as most of us have encountered.
I Have done a re-paste by some random pc shop in Australia in the past because while I attempted a re-paste myself, I ended up bending one of the heat-sink pipes slightly by accident(happened very easily due to VRAM/CPU components being stuck hard on the board from excessive ammount of paste they used. Mby I should have used a heat gun...) which then led me to buy a complete new cooling unit(heatsink + fans) because i wouldn't take a chance that even a minor bend would affect my output performance.

Still havent figured out whats the best method for best cooling method of CPU/VRAM components(pads, K5 pro, MX6, Kryonaut). Im considering an attempt with MX6 because manufacturer claims that its suitable for these situations. Quess i have to start testing a bit more because somehow i doubt that K5 Pro is actually the best option for low temps(like some people on youtube claim it to be).
Not a big fan of liquid metal btw, would like to point that out as well.

Current Idle temps(while under light load like just browsing the web/writing this article) with below Throttlestop config:
*48(avg) 71(max)
*38(avg)-50(max) max fan

My goal is to get 90c or less on the CPU while playing CS2(Counter Strike 2) getting around 250fps(or more if lucky because my monitor goes up to 360hz).
With current settings im hitting around 96-98c staying around 200-300fps which isn't really ideal in my mind while hoping to have this unit around for at-least couple more years...

Laptop sits on a glass table in cool ambient settings(currently located in Northern Europe and its pretty much already winter).
Not using external cooler or cooling pad currently(worth getting both of em? Like this one here?)

Ive also stocked up on 0.5mm/1mm/1.5mm pads + Arctic MX-6(To consider another re-paste cause last time was done over a year ago)

Anyways, heres my current Throttlestop settings while running R15(max fan):

Gyazo

gyazo.com

Gyazo

gyazo.com

Gyazo

gyazo.com

Gyazo

gyazo.com

Is there anything i could try to improve this at all?
I used to have older version of Throttlestop, so unfortunately im not quite sure if these are optimal/ideal settings

Thanks !

edit: Also would like to point out that im running cs2 1080p on lowest settings for competitive and ive also installed process lasso probalance enabled

 

[unclewebb]

ThrottleStop shows a 20C difference in the maximum core temperatures. Either the heatsink is not flat and not making even contact with the CPU cores or the thermal paste needs to be replaced. Have a good look at the heatsink when you redo the thermal paste.

That seems to be the biggest problem. Your settings look OK. Does the TS Bench 960M test report any errors? Most 8750H ate not 100% stable with the cache offset set to -170 mV.

You can reduce the turbo power limits to reduce heat but this will reduce maximum performance.

 

[Lewzke]

I have the same CPU in my notebook * HP Omen 17 inch. I had the same issue: overheating and throttling.
How I solved:
- Honeywell PTM7950 thermal interface, probably the best non-conductive material for cooling notebooks and GPU-s * no pumpout effect and performs almost equal to liquid metals without any dangers (this was the biggest change in the thermals) (this is not a pad but a paste with a pad state installation, this pad will melt above 45C) (Lenovo uses the same stuff in the Legion notebooks, PTM7958 paste version - performs almost the same as the PTM7950 pad)
- Throttlestop program * -0.125V for core voltage offset and -0.125V for cpu cache voltage offset (big change in thermals and power consumption) (if stable you can run it with windows scheduler every startup)
- Lapping the heatsink * this was too much work for a minuscule gain
- for me MX4 paste was a disaster compared to PTM7950 (I had throttling even with Throttlestop), the problem with regular pastes that the direct die contacts in notebooks will pump out the paste from the contact surface, the thermal conductivity number is not the whole story ... with regular PC CPU MX4 is good, but I will never use it for direct die cooling where there is no CPU cover.

 

[Quack991ˇ]

ThrottleStop shows a 20C difference in the maximum core temperatures. Either the heatsink is not flat and not making even contact with the CPU cores or the thermal paste needs to be replaced. Have a good look at the heatsink when you redo the thermal paste.

That seems to be the biggest problem. Your settings look OK. Does the TS Bench 960M test report any errors? Most 8750H ate not 100% stable with the cache offset set to -170 mV.

You can reduce the turbo power limits to reduce heat but this will reduce maximum performance.

Jeez that's not good at all, i didn't even notice the difference myself before. Thanks a lot for pointing that out mate!


total 224 errors, ref attached. Im assuming that target level is 0 here?(never played around with it too much)

I have the same CPU in my notebook * HP Omen 17 inch. I had the same issue: overheating and throttling.
How I solved:
- Honeywell PTM7950 thermal interface, probably the best non-conductive material for cooling notebooks and GPU-s * no pumpout effect and performs almost equal to liquid metals without any dangers (this was the biggest change in the thermals) (this is not a pad but a paste with a pad state installation, this pad will melt above 45C) (Lenovo uses the same stuff in the Legion notebooks, PTM7958 paste version - performs almost the same as the PTM7950 pad)
- Throttlestop program * -0.125V for core voltage offset and -0.125V for cpu cache voltage offset (big change in thermals and power consumption) (if stable you can run it with windows scheduler every startup)
- Lapping the heatsink * this was too much work for a minuscule gain
- for me MX4 paste was a disaster compared to PTM7950 (I had throttling even with Throttlestop), the problem with regular pastes that the direct die contacts in notebooks will pump out the paste from the contact surface, the thermal conductivity number is not the whole story ... with regular PC CPU MX4 is good, but I will never use it for direct die cooling where there is no CPU cover.

Cheers brother, il defenetly look into it and appreciate the suggestions

Out of curiosity, what kind of thermals are you getting under heavy load with PTM7950 ?

 

[unclewebb]

I like to be as close to 100% stable as possible. One error during the TS Bench test usually indicates that the cache needs more voltage. It is up to you. Game stable might be good enough for you. I would decrease your cache offset voltage until the errors go away.

The core and cache undervolts do not need to be set equally. Do some Cinebench testing. Many users get their best results when the core offset is set to a larger number compared to the cache offset.

PTM7950

I second that. Genuine Honeywell PTM 7950 has been getting a lot of recommendations lately for laptop CPUs.

 

[Quack991ˇ]

I like to be as close to 100% stable as possible. One error during the TS Bench test usually indicates that the cache needs more voltage. It is up to you. Game stable might be good enough for you. I would decrease your cache offset voltage until the errors go away.

The core and cache undervolts do not need to be set equally. Do some Cinebench testing. Many users get their best results when the core offset is set to a larger number compared to the cache offset.


I second that. Genuine Honeywell PTM 7950 has been getting a lot of recommendations lately for laptop CPUs.

"whea uncorrectable error" bsod appearing now pretty soon after booting into windows(throttlestop not opened). Which is more likely, could this be related to undervolts in throttlestop or am I having some sort of a hardware failure now or just corrupted system files from unstable undervolt?

Is undervolt still active even though throttlestop is not running?

Only thing I changed before going to sleep was setting battery setting to balanced from performance mode in windows and installed hwinfo to further attempt checking the package temps. (Didn't update any drivers, didnt change any throttlestop settings- it's still as seen on above previous pics uploaded). Left the laptop idle and woke up to PC being shut down(to my surprise).

Another thing I noticed was that battery level has dropped from usual 99%-98% to 93% even though it's been plugged in and charging the whole time. Would it be worth trying to remove battery and try rebooting then?(even though it's not designed like that)

Also wanted to kindly ask/clarify - by decreasing cache offset, do you mean - or + from -170mv

 

[Lewzke]

I have in between 82-90 Celsius under CPU heavy games (Counter Strike 2, Hell Let Loose), no throttling to 94C. With MX4 the temp was on the thermal limits. 94-97C.
With PTM7950 i have throttling only on Cinebench and syntethic benchmarks when I load the CPU + GPU Furmark. But no throttling in real world applications and games. The 8750H is definetely a heat generator with 6 cores all at 3.9Ghz. These notebooks are designed in throttling in mind... The cooler is undersized for 8750H boost TDP + GPU. It is designed for 45W TDP limit, but the boost is a lot more...
It depends now how I set the Throttlestop, I can get better results with -0.13, but i made to -0.125 for stability headroom. I had stable operation with -0.13 too. At -0.14 I have crashes. I use windows scheduler to run Throttlestop in the background every startup. In rare ocasions Throttlestop stops working and the offset voltage drops back to 0, in that case I have to restart the PC.
I was thinking first to use liquid metal, but the risk is too high and I have this notebook for personal projects to run, I sleep better with the PTM.

 

[kingsback]

there is no hope mate. gave up

 

[Quack991ˇ]

there is no hope mate. gave up

Just made an account for this haha? Great to have you around!

Managed to successfully restore the system. Lucky no components are damaged after all the troubleshooting... Still not quite sure though what may have caused the bsod. Not gonna give up though

 

[kingsback]

Just made an account for this haha? Great to have you around!

Managed to successfully restore the system. Lucky no components are damaged after all the troubleshooting... Still not quite sure though what may have caused the bsod. Not gonna give up though

yes mate, i did, just 4 you....

ps. its good you overcome the "problem"