P170EM and 3940XM very high temps ! Help !

[hoani]

Hi !

I ordered a 3940XM for my P170EM. At stock, I get more than 50% throttle with prime95 with 3 cores reaching +102°C in seconds, note even approaching 3.5GHz... Please note that I'm living in Tahiti and room temperature here is at 28°C min... when using the air conditionner.

Even if I switch back to my original 3610QM I get not so cool temps : I can get up to 100°C while stressing the CPU in 5 mins

I (re)pasted my CPUs with Coolermaster E2 Essential thermal paste.

What do you think I should do to improve those temps ? I remember that before repasting the 3610QM, I maxed at 90°C max. Do I put too much pressure (tight the screws to MAX) ?

I also plan to lap my CPU heatsink and to apply Collaboratory Liquid Ultra. Already done that on my 970M with great results (-10/12°C).

Please help !

[Tomaszz1975]

Hi!

Try Cooling IC Diamond 7 I think is one of best , see in google "cooling mod p170em" and https://forum.techinferno.com/clevo/2808-%5Bhardware-mod%5D-p170hm-cpu-cooling-2920xm-overclocking.html

Last option is the best for you but require good skills and is easy to damage your cpu heatsink.

In my P170hm I had copper pads on cpu and gpu heatsink's + fans on bottom cover like on p170em mod. I dont know what normal temp is for your CPU but I had 2960xm on P170hm and in OC idle was aprox 92C, room temp 22C.

Hope this help

Regards

Tomasz

[hoani]

Hi!

Try Cooling IC Diamond 7 I think is one of best , see in google "cooling mod p170em" and https://forum.techinferno.com/clevo/2808-%5Bhardware-mod%5D-p170hm-cpu-cooling-2920xm-overclocking.html

Last option is the best for you but require good skills and is easy to damage your cpu heatsink.

In my P170hm I had copper pads on cpu and gpu heatsink's + fans on bottom cover like on p170em mod. I dont know what normal temp is for your CPU but I had 2960xm on P170hm and in OC idle was aprox 92C, room temp 22C.

Hope this help

Regards

Tomasz

Thank you for you reply !

I already look at every hardware mod for inspiration but I want to understand the boost of my original 3610QM temperature.

Finally I tight evenly to the max every screw of the CPU heatsink and I got stabilized temp under 90°C. I was just missing 1/8 turn on one screw !

I will try again with my 3940XM !

[hoani]

I switched back my 3940XM in my laptop, properly tighten screws... and still getting awful temps... However I can get similar temperatures to my 3610QM at 3.1GHz when I set the multiplier to x31.

core0 is OK and never go more than 94°C while core1 to core3 can bump to 103°C in a second.

Is it normal to have one core with such difference on other cores ?

There are some cosmetic wears on the CPU die and I wonder if I have a damage one... It is a SR0US anyway.

I still have today to decide whether my CPU is defective or if my capacity to dissipate temperature is too bad... and if I chose the first option, I'll have to request a refund.

[Khenglish]

It sounds like You just aren't getting the heatsink all the way down to the CPU. The CPU is fine, or else it would not work. You could try thin shims between the heatsink screws and the heatsink to get a little more pressure. You'll have to pop the screws off the heatsink to do this though.

Also if you replaced the thermal pad between the heatsink and the powerFETs with something thicker than the original, that could be tilting your heatsink so it does not touch the CPU properly.

[hoani]

It sounds like You just aren't getting the heatsink all the way down to the CPU. The CPU is fine, or else it would not work. You could try thin shims between the heatsink screws and the heatsink to get a little more pressure. You'll have to pop the screws off the heatsink to do this though.

It seems to be the issue I'm facing, agreed with that. Do you thing the original heatsink is enought to dissipate 55W or more of power ? I'm also wondering if popping the CPU heatsink screws off it is hard or not. How to you proceed ?

Also if you replaced the thermal pad between the heatsink and the powerFETs with something thicker than the original, that could be tilting your heatsink so it does not touch the CPU properly.

I have some more K5 Pro available and I can try with it. I still have many few thermal pad left from previous installs. Which one would you advice me to use.

FYI, I saw that XTU set the Turbo Boost Powermax to 68W and the Turbo Boost Short Power Max to 98W. I lower them to 55W/63W and can stabilize at 3.4GHz with temps below 99°C, not ideal but suitable situation until I find I better solution.

Anyway, thank you for your directions Khenglish !

[Doctor JO]

Do you have any overclock?

Replace thermal compound, thermal pads and clean all went system.

In case if your EC have any issues with you current CPU compactly is HWINFO64 and set yours fan speed.

For its help achieve best performance and freez my gpu and cpu in Alienware m18x R2

Sent from my SM-G928F using Tapatalk

[Khenglish]

It seems to be the issue I'm facing, agreed with that. Do you thing the original heatsink is enought to dissipate 55W or more of power ? I'm also wondering if popping the CPU heatsink screws off it is hard or not. How to you proceed ?

I have some more K5 Pro available and I can try with it. I still have many few thermal pad left from previous installs. Which one would you advice me to use.

FYI, I saw that XTU set the Turbo Boost Powermax to 68W and the Turbo Boost Short Power Max to 98W. I lower them to 55W/63W and can stabilize at 3.4GHz with temps below 99°C, not ideal but suitable situation until I find I better solution.

Anyway, thank you for your directions Khenglish !

Popping the heatsink screws off is not hard. You just sqeeze the plastic on the bottom side with pliers. Removing the screws also allows you to lap the heatsink, which you said you desired to do in the first post.

Yes, if mounted properly the P170EM heatsink can easily handle 55W. Older clevo systems like the EM series have problems with warped heatsink plates though, which affects mounting. They are never warped enough though to have temps as bad as you are having though. Lapping the heatsink will fix the warping.

It sounds like you have some other mounting issue besides a warped heatsink contact plate. When removing the heatsink, look for where the paste seems thicker. This means that part of the die does not have enough pressure. You said your core0 is cooler (it's still too hot). IIRC core0 is on the end of the CPU away from the fan. Try to see what can be preventing your heatsink from tightening down near the fan side.

[hoani]

Hi,

I did lapped my CPU heatsink and used washers to enhance pressure on my CPU and I'm still getting crazy temperatures :

Here are some photos of the lapping process :

1. After using 600 grit for 15 mins :

2. After using 1000 grit for 15 mins :

3. After using 1500 grit for 15 mins :

4. Here is the re-assembled heasink with washers inserted to enhance pressure :

5. I was quite afraid to crack my CPU so the screws are to fully tighten... So maybe the results are so bad because of uneven "tighteness" of my screws ? Do you think it is safe to screw to the max with those washers ?

I'm still wondering if my CPU is OK or not

[hoani]

Tried to tigh the screw to get much more pressure (I saw the arms of the heatsink starting to bend) but no luck, my temps still horrible :

It seems to be a very precise science to just tighten those screw at the right level evenly... Do you have a special technique to get into it ?

[Khenglish]

What paste are you using?

[hoani]

What paste are you using?

Not the best one, but not the worst : Coolermaster E2 IC Essential. 5.5°C difference with IC Diamond 7, nearly 10°C with liquid pro

I have two seringe of Collaboratory Liquid Ultra coming to my PO BOX... Ordered them one month ago, should received them this week (I hope).

[hoani]

I have also to mention the fact that the 3940XM I received has "fossilized" thermal grease on top of die that I wasn't able to remove. I can feel with my nail that there is an difference between mirror parts (that slides very well) to the old paste (that makes my nail grip on it). Maybe this is why I'm facing such difficulties making this CPU working properly at stock and I'm observing core0 lower temp...

Here are some pictures of the guilty CPU :

[hoani]

I put my laptop on a small room with air conditioner set to 22°C, I tried to screw evenly all the four screws (3 turns) and I managed to stabilized the temps :

I don't know if regarding the conditions (room temp and my paste quality) my temps seems correct... I'll try at home in a less optimal situation (27°C room temp) if I'm still reaching the red zone.

I tried to push to x39 on all core : some cores still bump over 100°C and I can't make further x37 due to throttling.

I hope that with coolaboratory liquid ultra (should be -10°c) I'll get suitable conditions to have some slight OC possible !

[Khenglish]

Have you tried using any chemicals to get the old paste off?

[hoani]

Have you tried using any chemicals to get the old paste off?

I used alcohol and acétone with no luck.

Envoyé de mon A0001 en utilisant Tapatalk

[chap]

i d advice liquid metal, as its better than this ic diamond stuff and its easyer to apply and that paste normally comes with some sponge which is pretty hard so im sure u can polish off the old paste.

i have a 3920xm and i only reach 100° with prime95 and 4,3 oder 4,4 ghz

why are there 2 pieces of thermal pad on the cpu heatsink for the VRM, right now im pretty sure tey might be to high so the heatsink is not propperly touching the cpu. maybe remove thos tinyer therma pads and try it again. that would explain why your temps get better when u tighten the screws as much as u can

[hoani]

i d advice liquid metal, as its better than this ic diamond stuff and its easyer to apply and that paste normally comes with some sponge which is pretty hard so im sure u can polish off the old paste.

i have a 3920xm and i only reach 100° with prime95 and 4,3 oder 4,4 ghz

why are there 2 pieces of thermal pad on the cpu heatsink for the VRM, right now im pretty sure tey might be to high so the heatsink is not propperly touching the cpu. maybe remove thos tinyer therma pads and try it again. that would explain why your temps get better when u tighten the screws as much as u can

Thanks for the tip, as I mentioned, I'm waiting for two seringue of Collaboratory Liquid Ultra to come to my POBOX. Already used one on my 970M GPU with impressive results. I still have the sponge and will try to use it to remove the old paste as mentioned.

I effectively removed the additional pad and get better results (the last screenshot has been dont WITHOUT the additional thermal pad). I tried this to follow an advice from Khenglish.

I'll try to tight more and more my screws and see the difference between... I don't want to break anything :/

[hoani]

that would explain why your temps get better when u tighten the screws as much as u can

With those additional pads removed, if I tighten my screws one more turn or half more turn, I get worse results... Seems I found the coolest area.

[chap]

thats strange, i always turn the screws as hard is i can turn them before the screwdriver pops out of the screw. u cant destroy anything there. it than can only be the old paste.

can u look up the cpu vcore with cpu z?

your using prema mod right? do you test with full fan mode?

cheers

[hoani]

thats strange, i always turn the screws as hard is i can turn them before the screwdriver pops out of the screw. u cant destroy anything there. it than can only be the old paste.

can u look up the cpu vcore with cpu z?

your using prema mod right? do you test with full fan mode?

cheers

The vcore jumps up to 1.261V. I'm testing with full fan mode on prema mod.

Don't forget I inserted spacers between the heatsink and the screws so I potentially can break the CPU if I screw has hard as I can.

[Khenglish]

With those additional pads removed, if I tighten my screws one more turn or half more turn, I get worse results... Seems I found the coolest area.

Don't run the CPU hard for long with those pads removed. I'd say keep testing under a minute. There is temperature reading and throttling if necessary for the powerFETs, but it's not very good (just one sensor off to the side) and you can blow them easily.

[hoani]

Don't run the CPU hard for long with those pads removed. I'd say keep testing under a minute. There is temperature reading and throttling if necessary for the powerFETs, but it's not very good (just one sensor off to the side) and you can blow them easily.

I never removed ALL the pads, just the ones I added in addition to the original ones just to test if there is a difference. There was a difference but it was a hotter one.

[hoani]

Finaly I decided to go to the no return point and start lapping my CPU die to remove this awfully stuck old thermal paste :

after the furious lapping session, I got this (you can see the tool I used - from an used Coolaboratory Liquid Ultra kit - to lap my CPU die) :

ready to screw the heatsink (you can notice that I kept the original pads for the powerFETs) :

And the result was... disapointing The same, awful temps, reaching 103°C on some cores with more than 45% of throttling....

I decided to tune the tighteness of my screen LIVE, using a Viltrox LCD screen I have for my camera and HDMI output as mirror of the screen. After many tries, it appears that the upper right screw needs to be particulary more tighten than the other screws to keep on "normal" temps (below 95°C).

Here are my measures of the temperature with the optimal tightening configuration I've found :

I can't wait getting my liquid ultra !!!

[chap]

so ok, i checked my 3920xm now and with 4,2 ghz and +50mv i get throttling 2, i think i have to renew the liquid metal.

but, prime95 generates a ridicules amount of heat and power draw, my tdp easily hits 100w while gaming bf3 i barely get 40w tdp, so its not that big deal with the high temperatures as u will never hit those temps with normale use. even programs with high cpu utilisation will not generate such a heat.

in 3d mark11 i can run my cpu with 4,5ghz or more as is also doesnt generate much heat

so id say dont give too much on the temperatures under prime as long as u can run other programs without throttling.

i also never test if the cpui is stable with prime because it takes way too long to run unit there is a error

bf3 is pretty good for stability testing it crashes very fast if the mv or clock is too high or low

prime95 will sooner or later destroy the vrms if u keep it running for hours, also keep in mind hat lenovo never ment the extreme cpus to overclock so the vrms arent designed to provider 100w over a long timespan