Uneven CPU thermals!

Yes remove the shielding to get to the liquid metal. Be careful, if its really cured to the capacitors its not able to be removed without a defect.

You could possibly remove it with a heating gun. You mist get it above 56C Material Temp to get it loose. And when its in liquid Form swab it up.

3 Likes

Bro upgraded to 24H2
Some how score decreased from 15500to14900, power decrease from 45w to 40w.

Wtf is going on here😭

For the best possible performance you need to check the following:
Power Profile, set to highest performance. Linux and Windows both have settings for this.
Battery: The battery near full charge rate (what ever you set in bios like 80%)
External power: If you have the framework charger use that, is is capable of 180w power delivery or a comparable power supply connected to the laptop.

When this is all lined up you should get the best results, as you will get the max performance possible, tempature allowing (throttling).

That is exactly what I did, And the score WAS 15515 (power around 45). After ~a week, or change to windows 24H2 version, it decreased to 14900 and 38w. That is miserable.

1 Like

Are you still on the stock configuration?

Thermal solution done by Framework/Coolermaster seems to fail over time due to the liquid metal not being contained enough and or insufficient amount used.

Options:

  • Open a ticket with Framework and see if you are eligible for some kind of replacement or fix.
  • Start tinkering yourself, remove the LM and replace it with your own preferred solution.
    ( Recommendation would be to use PTM7950 by Honeywell or rebranded suppliers of the same material.)
  • Leave it as is and accept it if you’re not often running your machine to the max.
2 Likes

I just ordered the FW16 earlier today and now I’m reading up on this thread. Has framework commented at all on this officially yet?
I’m guessing when my unit arrives I’ll have to keep an eye on the thermals and just apply some of the ptm7950 I have sitting around if I see an issue.
I’m also curious now what thermal solution does the GPU use? Is there possibly a similar issue or potential benefit to changing the TIM there as well? The only device I’ve never found benefit with improving the thermals on is the laptop the framework is replacing (zephyrus M16) so I’m curious to see how well it’s been done here.

2 Likes

The GPU uses a conventional VaporChamber + Thermal Paste + Heatpipes with no added Shim Setup. The GPU is never hot. The highest i oberserved was just above 70C.
I am curious to hear how your Fresh Out of the Box Device performs and if its degrades how bad over the Time.

2 Likes

I’ll definitely follow up when I have it in hand. I’ll try to remember to get a day one benchmark to compare to in the following weeks and months as well to see if there is any performance degradation over time. It looks like cinebench is the standard used in the discussion so far so I’ll use that and OCCT stress testing to try to obtain useful data.

2 Likes

Hello everyone. I stumbled across this thread the other day and after seeing all of the issues with the uneven thermals I thought I should give it a go…

I’m using a Batch 1 Framework 16 with a Ryzen 9 7940HS.
My first test was in Ubuntu using Lutris and cinebench R23. I saw I got a score of about 12500 and figured maybe I should install windows and see what hwinfo can show.

These are my temperatures during the run, as well as the package power. I’m using windows 11 24H2 with the framework driver pack installed. The power settings are currently on Performance…

I got to the end of the test and it appears that the test froze, with the system still responsive. The score shown when I hover over my system in the ranking is 12821.

I saw that people had good success with the Honeywell pad. At this point should I find some time to try it out? :sweat_smile:

2 Likes

Well yes definitly you should do something. Either Getting a RMA on Framework so they know they hit the “shit” Jackpot or yes you could change the Thermal Interface Material on your own. We try to Spread awareness and somehow “force” Framework to do something.

3 Likes

New thermal pad installed, first run CB23 14965, Windows 11, max CPU package power 47.508W, a few days ago was 11891 and 28 Watts (with the LM).

3 Likes

There’s a pattern here…

It doesn’t bode well for the current offered thermal solution…

5 Likes

This is the same as the RTC Battery issue they had.

I really hope that the framework team will look into it.

2 Likes

PTM7950, day 10 update.
Test after one hour on idle.

54.674 W CPU Package Power after 9 minutes and 10 seconds

Result: 16122 pts

CPU Package Power

  • Average: 53.016 W
  • Max: 61.569

I’ll try to have a new test after 30 days. So far, it looks like nothing changed.

2 Likes

I have just tried to replace my liquid metal as my CPU Package Power wouldn’t go over 36W using the 7940HS (and 180W charger).

After replacing the liquid metal with 2 layers of Thermal Grizzly PhaseSheet PTM (I felt like 1 layer was too thin), the CPU package power holds a steady 53-54W with a maximum of 60.052W and average of 53.439W after 10 mins of cinebench. I’ll keep doing tests over the next few weeks to see if things get worse or stay the same.

Just an important note about the PhaseSheet PTM, I initially hit 48W on average but after a few thermal cycles it stabilized at the above 53-54W.

2 Likes

did you do anything with the shim, or was this simply removing the previous interface material, and substituting your own?

Nope, I did nothing. Just took off the heatsink, which was a bit of work as it wouldn’t unstick, and then scraped all the liquid metal out, cleaned the surfaces with 99% isopropyl and applied the PTM.

I am a bit worried about my CPU having black patches on it that couldn’t be cleaned no matter how much I rubbed isopropyl on it, but it seems to not affect any functionality, so I hope it is nothing bad. I attached some pictures of how my CPU and heatsink looked like after I got the liquid metal out. I then just rubbed the surfaces with isopropyl and applied the PhaseSheet PTM.


The black patches on the CPU here couldn’t be removed no matter how much isopropyl I used:

On the heatsink I managed to remove the black patch by rubbing and scraping with the plastic end of the FW screwdriver

1 Like

that’s kinda scary looking :o
looks like something cooked in there. which is not ideal

1 Like

Yeah I am still slightly worried about it, but it doesn’t seem to affect the functionality nor thermals.

here’s hoping. it’s really good to know, though, that this kind of performance can come from such a simple swap. i was watching everyone else do surgery to their thermal assembly and getting really nervous. if simply swapping to some ptm gets results like this, i’m much more confident about it

1 Like