help with razer blade pro 17 (2019) w/ i7-9750H

[okan]

it seems like my pc was running slow so i decided to run a cinebench and compare my score to other 9750H's and it seems i am under performing quite a bit. i am getting different results each time as i fiddle with things, but the current best i have done in R23 is 5935 multicore. when i first was looking into why things felt slow i was getting low 4000's.

so far i have taken my laptop apart and cleaned the fans and heatsinks, ensured the fans are all working, repasted with some thermal pad shit i found on amazon that i saw had a better thermal conductivity than ptm7950 (12.8W/mK vs 8.5), but that didn't help at all. i then realized that stuff wasn't a phase change material like i thought it was and ordered something that specifically says ptm7950. so far the ptm stuff seems to be working better but that is how i got to the R23 5935 score.

i have gone and fiddled with random things according to other threads i've found, so my settings might all be pretty off right now. i reached out to unclewebb on reddit and he pointed me in this direction. i've attached the things he mentioned to post. i'm hoping to get this thing running like new or even better, with the undervolt. the current undervolt i achieved before applying the new ptm7950 stuff.. i'm not sure if i should try to go further now. i just got 12.037 on the 120M TS bench, and 108.305 on the 960M without errors, both of which are the best i've seen since starting this journey to get this laptop working correctly again.

i've seen the limits be different things throughout all the changes i've made. i followed some steps rom other posts and was able to not have pl1 or pl2 but was hitting thermal. eventually i deleted my ini and tried to start fresh again but i'm pretty sure it's all a mess.

edit: i just did a timespy and added the results screenshot as well as a TS log.

 

[PHVM_BR]

I had an i7-9750H and I believe I can help you.

First check the MMIO Lock box in TPL.

Return your Ratios (FIVR) to default as this CPU doesn't support overclocking.

Then, if your undervoltage is really stable (no errors in TSbench is a great start), lower the core voltage to -190mV.

Maintain -115mV in cache.

Take a long test on TSbench and check if any errors appear.

If everything is ok, run a test on Cinebench and compare the results.

Without changing the Windows priority, my i7 was doing about 8000 to 8100 on the CBR23.

Time Spy's CPU score was 7100 to 7200.

This is after several BIOS updates with security fixes from Intel.

Initially (2020) my i7 could do +8300 (CBR23)...

It was completely stable with -125mV / -210mV in cache/core on Windows balanced power plan (Speedshift 84) and -135mV / -225mV on high performance plan ( Speedshift 0).

 

[okan]

so it seems i am definitely able to get a better undervolt with the PTM stuff. i was getting errors at -120mV core / -115mV cache before, but i'm currently at -140 / -115 and still going.

i set the turbo ratio limits to 45/44/43/42/41/40, and checked the lock next to MMIO. what does the lock do? i see lock buttons all over the place... does something else try to change the values?

i keep getting PL1 around the 28 second mark which is the turbo time limit... is that how it's intended to work? what should my PL1 and PL2 be set to? what does the clamp option mean? i'm playing with the PL1 number to see if it will make that limit go away but it doesn't seem to.

i'm going to keep working on the undervolt, then i'll upload a new log and screenshots.

edit: i have added screenshots and my log. i need to stop for the night and continue tomorrow hopefully.

i tried undervolting the cache to -120.1 and saw an error so i put it back to -115.2 and worked on the core. -5 at a time i tested all the way to -219.7mV. at -174.8 i saw errors. stopped the test and ran it again without changing anything and it finished without errors. i ran it a third time and got no errors again, so i continued going -5 at a time. i saw another error at -210, stopped the test and ran again twice, as i did previously, and passed again without errors. i continued -5 at a time again to -219.7 and then i got impatient and jumped to -235.4. it passed. wtf... so i jumped all the way to -250 and it passed again... i decided to stop there for the night on the core, and switched back to cache to see if i could get any more out of it. i did -1 at a time. 116.2, 117.2, and 118.2 all passed (core still at -250). all tests are the 960M one. where i see a score between 101.5-104 seconds. that is ultimately where i decided to stop for the night, and ran a CBR23. in the attached log file, the cinebench was the last thing a did. i forgot to stop the log when i came to write this so theres some extra idle data at the bottom.

what else can i use to test the undervolt stability? or is TS bench good enough?

some observations:
-PL1 seems to flicker to yellow every so often, then back to red.
-PL1 doesn't hit at 28 seconds like i thought. i've seen it come on as early as 25 seconds into the TS bench, and as late as 40 seconds or so (wasn't always paying attention to the timer)
-PKG power drops to around 55w when PL1 hits.
-C0% sitting here typing this is boucning around 13-16% on the main TS screen

 

[PHVM_BR]

Checking MMIO Lock is to lock the power limits you have set so that the original setting does not change the values dynamically, but it appears this is not possible on your machine.

Did your system have performance mode selected in the Razer control software?

Are there performance profile options?

Your CBR23 log shows a PL2 of ~65W at ~95°C and a PL1 of 55W at ~90°C.

Was this with PTM 7950?

Take a look at this post I made about undervolt on the i7-9750H:

i7-9750H: undervolt x power ratio

I have a Precision 7540 for more than 3 years and with undervolt it always delivered easily the maximum multicore performance of the i7. At the beginning of the year I changed its thermal paste for the Honeywell PTM 7950 phase change pad and decided to test its maximum cooling capacity and took...

www.techpowerup.com