no difference after undervolting ? 10875H

hi guys

am new to this undervolting thing,

on my GS75 stealth i have a 10875H, so i thought id undervolt it to get lower temps, I KNOW this chipset cant get very low undervolt values so i was recommended to start with -0.50mV, and this is what i did, only i did Not see any difference in terms of temps, it feels like no changes happened !

b4 undervolt:
https://ibb.co/b3kPLwP

after:
https://ibb.co/DgRFPqV

also how to go about increasing the voltage value without bricking the laptop ?...

thanks,

 

If you are using Intel XTU for undervolting you will need to match Core and Cache undervolt. I do not believe you can have them at different values. -.50mV is a good place to start but may be only a degree difference if that. If you match core and cache and still don't see a difference try -.75 or a bit higher and watch voltage/temps.

If that is still not working I'd recommend uninstalling Intel XTU and give Throttle Stop a try. There is a support thread here on this forum that can guide you as to setup/use. http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

 

It is possible that your processor does not support undervolting. Try ThrottleStop, it will tell you if undervolting is available. Uninstall XTU and restart before installing ThrottleStop.

There is still hope to improve thermal performance without undervolting. You can apply better thermal paste and use ThrottleStop to tune the Turbo power limit (TPL) values.

Ask in the ThrottleStop thread if you need some help with it.

Charles

 

I unlocked the cpu in advanced bios settings and TS doesnt say its locked ANYMORE but for some reason core voltage is greyed out in XTU !... maybe u guys can help ?!...

 

Don't run both programs at once.

Better yet, uninstall XTU.

 

There is a difference, power went down by around 5W, its just that your CPU is still using 83W, if thats with -50mV, go down a bit more, but 83W on that thin chassis will lead to high temps..

 

Thanks senso, am only worried id brick my laptop if i lower the voltages any lower, also i still cant get xtu to work which is supposed to be safer than TS, anyways, how to go about lowering my voltages with TS ?, small decrements till it crashes or i get a BSOD then increase again ? And what will happen if i go lower tha what the cpu requires, will i brick it permanently or will bios restore last working voltage etc ?!...

Thanks again,,,

 

With TS, if it crashes, there is no risk of bricking, because it only applies the undervolt after windows boots, and when it detects a crash it will ask you if you want to keep the undervolt or discard it, TS is smart, XTU not so much...

Undervolt core and cache in steps, start with -80mV core, and -50mV cache, and leave cache untouched for now.

And if its an MSI laptop, press the power button for 40s and it resets the BIOS settings.

 

well, i dunno but something is not right here, i reduced the voltages all the way down to -0.125 mV, no temp difference whatsoever, in neither stress tests NOR in gaming !... not sure whats wrong, it the new temps show up in HW Monitor but 0 effect ! also i know for sure the 10th gen cant really go down to 0.125 mV
settings:
http://grabilla.com/0ab11-184eeab2-10ee-49b7-ad79-92c869425991.png

http://grabilla.com/0ab11-dec8df37-c2aa-4966-b1ec-ab10dee40842.html

http://grabilla.com/0ab11-82aa2eb6-fa60-465e-aec5-92d51fffc60e.html

i will simply assume that undervolting the 10th gen is just non existing !...

 

Exactly.

You already proved that undervolting a 10th Gen works correctly.

If you really think that your undervolt is not working then keep going. Try -150 mV for the core and cache offsets then -175 mV for both and then -200 mV. If your computer crashes, I would conclude that the undervolt really is working. Run a light load 1 or 2 Thread TS Bench test along the way. See if it reports any errors. It usually does when you go too hard on the cache undervolt.

When exploring the outer limits of what voltages your CPU can handle, open the FIVR window and select the option, OK - Do not save voltages. This is a good safety feature. If your computer crashes while testing, when you boot back up and run ThrottleStop, the voltages you were testing will have been forgotten.

https://ark.intel.com/content/www/u...0875h-processor-16m-cache-up-to-5-10-ghz.html

You are pulling 90W from a CPU that has a 45W TDP rating. What are your expectations? Any laptop heatsink and fan are going to struggle trying to continuously dissipate that much heat. It would be normal for your CPU temps to go up over time while gaming or running any stress test.

Instead of overwhelming your CPU with Prime95, try a more practical test like Cinebench R20 or R23.
https://www.maxon.net/en/cinebench

Do a baseline run with the core and cache set to +0.0000 and then do another run with the core and cache at -125 mV. If you are stable, keep adjusting only the core to -150 mV, -175 mV, -200 mV and -225 mV. All of the 8th and 9th Gen CPUs show an improvement in temps or performance when testing like this. If you lose stability or a light load TS Bench test shows errors, back the cache voltage off a little first. It tends to be the limiting factor.

It looks like your laptop is running great. You are maintaining the full 43.00 multiplier when fully loaded. That is as good as it gets for a 10875H.

 

>>> You are pulling 90W from a CPU that has a 45W TDP rating. What are your expectations? Any laptop heatsink and fan are going to struggle trying to continuously dissipate that much heat. It would be normal for your CPU temps to go up over time while gaming or running any stress test.
i didn't engineer this chipset, am simply saying i cant see an effect on temps when undervolting, but for the sake of this argument ill make a 30 minutes test Again and compare to my original results here and see IF i can hit over 4 Ghz (less throttling) with undervolting...

I ran RE2 Remake game, temps were identical to what i saw before as soon as launched even at -125 mV maybe am missing something... or maybe temps doesnt necessary have to reduce with TS ??, only will get less throttling and hence the name of the program ??!

i know for sure 10th Gen chipsets can't possibly go that low in voltages so am speculating, the lowest i read 10th gen chip can reach maybe 80 mV ! now ur asking me to go even lower up to -200 mV ?!...

thank you for your reply ...

 

Do you know for sure? How?

You win or lose by thinking you win or lose.

 

Reducing the CPU voltage can result in increased performance or decreased temperatures. You might see both or you might only see one of these. You should see something if the undervolt is working.

Most of the 10th Gen testing that I have seen is for the 10750H. Have you seen a wide range of user testing of the 10875H?

The 10875H should be a better binned CPU. Perhaps Intel is using cache that is 100% stable at -125 mV like many of their previous gen chips were stable at.

You are saying that the undervolt is not working on 10th Gen CPUs. Some desktop boards have an option in the BIOS that bypasses the CPU voltage settings that ThrottleStop lets you access. If your computer does not crash then it is possible that you are 100% correct and CPU voltage control in ThrottleStop is not doing anything. Best to do some testing and prove this one way or the other.

Run Cinebench R20 or R23. With an 8 core CPU it should not take you 30 minutes to run a test. Easy enough to run this test at default voltage and then run the test again after reducing the voltage.

 

How can you run a -125mV undervolt and at the same time say FOR SURE that it doesn't work below -80mV, you cant even maintain continuity in two sentences..

 

I was helping a guy today on the TechPowerUp forum with his 10th Gen 10750H. His cache undervolt is only stable at -73 mV. That is kind of pathetic but it seems typical for these 10th Gen 6 core CPUs.

He was doing some Cinebench R23 testing with his core offset at -150 mV. It took a long time to convince people that using different offset voltages for the core and cache is a good thing to do. Now there is an internet myth that a 2:1 ratio between core and cache is some sort of magic number and you should not undervolt the core any more than that. I suggested to press on. Ignore the myths and do some proper testing.

His laptop has a 65W hard turbo power limit being enforced. By increasing the core offset to -200 mV, he immediately observed that his CPU throttled less and ran faster. A bigger offset core undervolt has helped him run faster, longer before turbo throttling at 65W begins. ThrottleStop is definitely working for him on his 10th Gen CPU. Power consumption is exactly the same as before so there is no difference in temperatures. That is what one would expect.

https://www.techpowerup.com/forums/threads/why-doesent-cpu-go-back-into-turbo.274794/#post-4394928

One user followed my crazy advice today and now his 10750H is probably near the top of the Cinebench pile compared to all of the other laptops running this same CPU. Do some Cinebench testing. It is a good tool to help get a laptop dialed in.

@tilleroftheearth knows. Never say never.

Edit - @MagillaGorilla is using ThrottleStop to reach over 4400 points in Cinebench R20 with his 10875H.
http://forum.notebookreview.com/thr...l-to-unlock-all-hidden-options.830993/page-40

 

I finally found it, this:
http://grabilla.com/0ab12-40eb711d-41b5-4963-aa98-66b1032f965b.png

here:
https://forums.tomshardware.com/threads/cpu-i7-10875h-temperature.3623641/

the entire time i was under the impression i cant really go lower than ~0.80 mV, also a youtuber said the same thing to me, we are ALL reading similar reports on google saying 10th Gen CPUs dont undervolt well, lets just admit it ! but so far am at 0.125mV for BOTH core and cache

funny thing one of my cores hitted higher temp 100c (momentarily only) but yes i saw LESS throttling (cant say for sure yet), but have u seen such behavior (if any) ever b4 with TS (ie. higher clock rates, but also higher temps ??)

you said:

now am at -.125 mV for both core and cache, do u want me to work on the core for now ? why do i need back the cache voltage if i loose stability not the core ?!?...

anyways will try cinebench and keep u posted, thanks developer

 

i saw no temp difference so i thought it wasn't doing nothing ! also am new to this area and dont know what to expect from undervolting nor did anyone suggest here (until I finally asked) and Uncle gladly replied to me and confirmed my assumption saying:
"Reducing the CPU voltage can result in increased performance or decreased temperatures. You might see both or you might only see one of these. You should see something if the undervolt is working."

such a shame though i wanted to lower the temps by undervolting...

 

If temperature was limiting your performance, undervolting would give you more performance at the same max temperature. You could, of course, dial back on SpeedShift and get back to the performance you had before, and temperatures would drop. Note, however, that these processors are designed to tolerate temperatures up to 100 degrees, so temps in the 90's should not be worrisome.

 

Tom's Hardware / reddit are only testing when the core and cache offset voltages are set equally. Many users on Notebook Review have learned that their best results are achieved by setting the core offset much higher than the cache offset.

The cache voltage is always the limiting factor. When you lose stability it is usually the cache that is causing the problem. When I suggest using different voltages, most people start at a point where the cache is already set too high. You either have a fantastic, well binned CPU or at -125 mV, your cache offset might already be too high.

Do some Cinebench testing. The goal is to run the entire benchmark at full speed with zero throttling. Open up Limit Reasons in ThrottleStop so you can watch if power limits or thermal throttling is holding you back. Bump only the core up in steps of -20 mV or -25 mV and test again.

There are users on Notebook Review running some of the fastest 10875H CPUs in the world. Follow their settings and advice.

 

i got 9000+, how is that possible ?!...
http://grabilla.com/0ab13-e31e3a58-84a1-4db8-9eb7-a6da609f936e.png

 

@jojonono - I am not familiar with R23 scores. The other guys were still running R20.
Scores in R23 are much higher compared to R20.

https://www.techpowerup.com/download/maxon-cinebench/

Is your CPU throttling during testing or are you running at full speed?

 

well, with TS running i got this result:
http://grabilla.com/0ab13-174e5421-6539-4543-8e0d-89121c481c87.png

before we were at 3.6 GHz. with TS during the 10 minutes test started out at around 4.3 Ghz, but dropper to 3.99 Ghz (probably due to cooling) so i guess TS is doing something at least, currently am at -125 mV for BOTH cache and CPU, No Crashes, should i work on the CPU for now or keep adjusting both ?...

anyways i will redo the tests with R20, thank you for the advise...

 

I would leave the cache at -125 mV and yes, I would start working on only the core. You are heading in the right direction.

 

NO TS:
http://grabilla.com/0ab13-e424756a-6bd0-4aff-a3c4-579ab7746c63.png

With TS:
http://grabilla.com/0ab13-2c15d5b0-44b7-4d6c-89b3-3322e2bbec60.png

 

how far should i go ? would i reach a stage of feeding less juice to the CPU and hence get less performance ? or it would simply crash if i go beyond a certain number ?...

 

Let your Cinebench scores guide you. You are the only person with your CPU. What your CPU is capable of can only be discovered by doing lots of trial and error testing. If bumping the core voltage to a bigger number does not increase performance or decrease temperatures then you have gone too far.

 

will keep trying to reach the best setting possible with TS, just one last thing, i tried the built in stress test and immediately got this:
http://grabilla.com/0ab13-589f2c8e-b816-4c67-96fe-906a3d220b13.png

does it mean anything ?...

 

You have been telling me that your CPU cache is stable at -125 mV. The TS Bench shows that you might be wrong. Reduce your cache undervolt and test again. Do some 1 Thread and 2 Thread TS Bench tests as well. The light load tests use a higher CPU speed so might uncover errors.

 

i never got any error, bsod, crash whatsoever and i did several so far with Prime95 and Cine 20 and 23, anyways, am at -125 mV for BOTH cache and cores, does this mean am fine ?
http://grabilla.com/0ab13-76c1ea2c-ec13-4e22-bf67-1ac06173d916.png

 

If the TS Bench was reporting errors before then no, you are not fine. You are right on the edge of stability. I would reduce the cache offset to -110 mV before trying to go higher with the core offset. You do not want the cache to limit your results. A setting of -110 mV is better than what most 10th Gen CPUs can achieve.

Once you are finished with your core testing, then you can go back and try again with the cache at -115 mV or -120 mV.

People use the simple TS Bench because it can warn you when you are on the edge. No need to run it for 12 hours like Prime95.

Edit - Here is a good TS Bench vs Prime95 story.
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-1020#post-10919625

 

to sum this up, TS Bench can mean either the cache or core voltage is not set properly right ?

what is the best TS bench setting to completely eliminate any instability ? 1 Thread @ 6144 MB ??...

thanks,

 

There is not best anything. Use a variety of tests. The 1 and 2 Thread TS Bench tests are both good tests to uncover stability.

I keep saying the same thing. When you run into a stability issue, stop and reduce your cache offset voltage first. The TS Bench is telling you that you have gone too far.

If you leave your cache with 5 mV or 10 mV of headroom, you will probably be able to undervolt your CPU core a lot more. This will ultimately get you the best temps and best performance. Follow this advice and you might clear 4400 in R20, just like other users are doing.

 

thanks a lot for your help, i was finally to get the bench tests All stable after spending overnight testing, funny i never saw a single issue with Cine bench, but you said errors would mean stability issues, so i kept reducing my cache and came down stable to -.80 (maybe even .90 would work) i was finally stable ! as for core was left on .125v, i got emotionally depressed after using the TS Bench lol ! since i had to do several tests manually and 1-2 cores took really long time, so i think i will stick to my results and wont waste more time playing with core voltages as i didnt see tangible difference in cinebench results above .125 mV

thanks,