Uneven CPU thermals!

Yeah, if you are confident that you can do the swap, by all means do it. I just did 2 layers of PTM mainly due to those black patches, as I did not want to risk applying too little of it.

It would be nice if Framework does anything about the issue, because even if I went through support I doubt a new mainboard would be a long term solution, since I did use to get 45W on the CPU when the laptop first arrived, but keep in mind, that was still worse than the 53-54W I get with PTM currently.

1 Like

my ideal would have been fw providing pre-cut ptm of some variety so i don’t have to deal with dancing around providers (i’m fond of the results for genuine honeywell PTM7950 but it’s so hard to actually get), but i’m not getting the impression they’re interested in doing that as of yet.

that the grizzly ptm performs that well (and is so much easier to source) is heartening, and i might go for that in a near future.

2 Likes

Only Framework could know how big this problem really is.

We can only see the people coming here on the forum looking for information and sharing experiences and findings.

In case it’s just a small percentage they might just keep it to in warranty replacements. I and a couple of others had replacements only to suffer the same problem later on with the replacement. Bad luck, coincidence or systemic problem, only Framework and/or Coolermaster knows.

I think only a redesign of the whole thermal solution could fix the problems.
A work around or patch is using PTM7950 which will give you around 99% of the performance what Liquid Metal gives you initially before it starts to degrade.
~99% is based on my initial benchmarks with LM and the replacement pad installed after seeing it degrade.

2 Likes

the concern i have is even framework might not know how many people are impacted, if most customers don’t know what to look for, or don’t think to. i know i went “wow, this is performing great” because i was comparing it to the machines i had previously. it was only when i started asking questions about “why is it always so hot even when in power-saving” that i found this thread and had more stuff to investigate.

2 Likes

Do you have access to a hot air rework station? If yes, I’d try to warm it up to 55/56C and see if it can be removed afterwards.

Just make sure that it won’t go in a different place.

If you don’t have access to a hot air rework station you can still use a hair dryer to warm up the heatsink and see if the black patch can be removed.

I’d say that you shouldn’t apply more than a layer, but, in your case, it can help because of the black patch.
Normally, most of the thermal paste will be pushed to the side.

See the attached picture from one of my tests.

Photo

That’s a really good result! :smiley:

4 Likes

I can confirm 1 layer is enough (stock), I have done it now 3 times and had 2 times to see the results when removing it, some of the pad gets pushed out to the sides.

I ordered an extra heatsink, I think the original one that came with my board performed a little bit better so I witch back.
I have some copper shims incoming, once I have them I will mod the 2nd heatsink.

2 Likes

I might try to heat it up using my rework station if I open it up again in the future. I still got plenty of PhaseSheet PTM left for a bunch more applications. Will post here when I do it.

2 Likes

I loved the results of the Phasesheet PTM when I redid the GPU and my steam deck. Waiting a little longer on the CPU as I am still holding 48+ on my PPT when doing R23 on Fedora.

Dosnt LTT sell that stuff?

you can try to undervolt your CPU. I have 7940HS too and with undervolt -20v I got 16405 pts but package power stays at 45W. I’m using universal x86 tuning utility.
in my case I have not modified a CPU cooling.

result

2 Likes

Is that score from cold, idle or after using the laptop for a while?
That’s quite a nice score and the delta is just perfect.

I’ll need to have a go and try to undervolt.

1 Like

It is score after idle. start temperature CPU Core = 40C and CPU(Tctl/Tdie) 42C. Ambient temperature in room 21C.

Without undervolt and starting from cold system. Peak package power is 60W and then drops almost immediately to 50W and after 1,5 minutes reduces package power to 48W. At the end package power is around 46,5W

result

2 Likes

When I first got the laptop I tried using x86 tuning utility, but even with a tiny undervolt I experienced issues with stability. The cooling was still new at this point too, so I don’t think it was with thermals

I am using the thermal grizzly brand. Amazon like… $15

Update:
My intergrated GPU can have full power comsumption (45w) for 20 min, but cpu could only cosume 38w (hit the temperature wall) and has a poor performance. uneven not just in CPU, but also between CPU and GPU?

What program are you using to undervolt? I’d like to give this a try with my 7940HS.

I’m using “universal x86 tuning utility”, and I set only undervolt on all cores using custom presets. I’ve started from -5mV and I was increasing by -5mv. at -25mV system behave unstable. That’s why I have -20mV. You need to test system under stress and in idle.
image

1 Like

Well I Have 2 new Heatsinks and i decided to test the Heatsink with the Higher Serialnumber. Worked like a Charme.
Thats my First Run, without a Burnin Cycle. Just One 10Minute Run i aborted and than this Single Run.
I am amazed. As it seems Framework did something as the new Heatsink unmodified performs better than my modified one.
i didn’t see 100C or even touched a 100C. 98C is the Highest Number.

edit:

Still no Cooling Cycle but with no Background Software and Monitoring:

I Do now hit 100C but only when i apply the X86 Universal Tuning Utility “Extreme” Setup and its pretty even.

6 Likes

That’s a nice result!
I don’t even think that you’ll need to modify it.

Out of curiosity, what serial number do you have?
So far, for the ones I received, I have:

  • 1405100AE
  • 140410304

The only problem is that I can’t remember which one is the one installed :joy:

Is this with the LM TIM that comes with the FW16? Or PTM / Paste?