Bro I know that you have said couple time, I tried to find your conclusion but it’s too far to reach. Thank you so much anyway.
@Destroya I know you can’t give any information, which makes sense, but I do have a question.
Is this fix oriented towards unmodified systems or, is there a way to also cover us who replaced the LM?
Currently, the thermal solution works better than the LM but, this doesn’t mean that I wouldn’t like to have better thermals, mainly when I need to compile stuff or run Docker.
A possible issue I can see is with the Support Team when they ask for a video. Because of the current modification, they could say, “It works as expected, there is no need to do anything.”
And that can be true and false at the same time 
6 Likes
Having read some of this thread.
I have also seen this YouTube video
How can we be sure this is a heat spreader / paste problem and not cpu variability?
On Linux, one can limit the clock rate on a per core basis. If one limits the core that is hitting 100C, does it then let all the other cores be used more, and thus better performance overall with just a config change as maybe a temporary workaround?
In my observations, if I turn off the core that is at 100C, a different core peaks. I was expecting instead that they all rise equally. This problem might just be caused by slightly uneven scheduling algorithm.
Maybe a better test would be turning off all cpu clocks except one, and then see which core executes the work quickest.
One might then be able to separate the affects from cooling method, scheduler and cpu variability.
1 Like
Repasting making a big impact for one
1 Like
No way it can be just a variability issue when a repaste of my cpu put me from 13k to 17k in cinebench on the same system configuration.
My fans only turn on under load now as opposed to before when I did any little thing, which also points to a thermal issue as opposed to a scheduling issue.
3 Likes
Just curious, was PTM available back then?
Just wondering why they did not mention it there.
I have watched a few PTM7950 videos (I know, not very scientific!) and it is interesting that on paper PTM does not look that amazing compared to paste, but when actually applied it seems to perform better than paste.
The video link above, at the end, does say Liquid metal is still a little be better than PTM, but not much. So, it would appear that FW choice of LM was a good one, but maybe, as mentioned in this thread, it might not have been applied in the factory correctly.
That’s probably because most of the “paper” on pastes is beyond meaningless.
Though I definitely did not expect ptm to work anywhere near as well as it did when first trying it, it really is one of those things you gotta try to belie. Stuff’s closer to lm than paste.
lm may still be easier to machine apply, or at least that’s my hypothesis why it is not uses in laptops from factory and lm is.
Because people who have re-pasted their systems see a massive improvement? I don’t mean to sound rude, but did you actually read through the thread?
Also, the higher end processor should never be massively below the cheaper processor (by a massive $200 I might add) in performance.
1 Like
I was raising it because I did not see it discussed and specifically discounted as the problem.
It is probably a heat spreader / paste problem, but people might be more convinced by scientifically discounting some of the other things.
I guess my question would be, can the existing LM arrangement be fixed, so that it still uses the FW LM material, but so that it actually works better, as opposed to completely replacing it with a different material? LTT do suggest the LM is slightly better the PTM, but PTM is easier and safer(less likely to damage anything) to work with (re electric shorts etc.)
I believe the content of the thread is showing it’s not an issue of LM or PTM.
The core issue is a heatsink manufacturing issue, where poor soldering of the shim, combined with an unfortunate design pattern to prevent leakage on it’s face, results in poor thermal contact no matter what paste/LM is used.
Excess LM application or change to PTM shows some improvement, but this only fixed the face pattern. Complete removal of the Shim and replacement with alternatives shows the most improvement from the tests above, meaning the real issue is a heatsink design/manufacturing defect
5 Likes
I 100% agree this is a fundamental design issue with the heatsink and I am convinced to the point that I have even requested FW to put my RMA on hold. It’s a useless exercise swapping boards.
2 Likes
When you mention “Shim”, are you referring to removing the heat spreader, as shown in this video:
De-lidding a CPU:
These are laptop boards, there already is no heat spreader. They look the same as a GPU board which was shown near the end of the video, again because of height/cooling efficiency.
When psierra talks about the shim in this case, it’s the small raised portion of the heatsink itself, that’s contacting the bare cpu die. A similar problem to the heatspreader issue from the video but coming from the heatsink instead of the CPU assembly.
By removing the shim and applying a PTM pad on either side of it, he reported better CPU performance and less thermal throttling, indicating that the shim to heatsink/heatpipe interface itself is poorly conducting heat.
Framework’s decision to dimple pattern the shim face contacting the CPU itself is potentially causing the insufficient LM contact problem, but this decision is somewhat understandable from wanting to prevent LM runoff. The shim-heatsink soldering is on the manufacturer Coolermaster however, a laptop heatsink should not be improvable by simply desoldering the contact plate from the main block and slapping some PTM in between.
If a PTM sandwich is more efficient than the actual solder connection, Coolermaster seriously screwed up.
10 Likes
If there was enough LM - if it was calculated properly, anyway - the dimple pattern probably wouldn’t be as much of an issue. I’m most unhappy with the shim-heatsink soldering above all, THAT is the truly disappointing part of it all.
I’m going to do the PTM sandwich mod myself soon. I really hope I’m able to get out the LM without destroying my machine. It’s become very difficult to use, but Framework demands that I send it in instead of offering to send me a mainboard and I’m a dysgraphic college student with no alternative. This is all I’ve got.
I am myself tempted to try that but I am afraid I will get the LM onto some critical component and fry the board. It can be done as others have and full Kudos to them for bravely attempting with success. I paid over $2100 for the machine we should not be turning it into a Frankenstein project…not for that price. Cooler Master and FW need to work it out and provide us with a proper fix!
1 Like
As much as this issue bothers me, I still like the machine and the concept and would like to believe that FW will address this because if they want to be different they will think outside the box and figure it out how to provide a proper fix. I am hopeful and confident and not giving up on FW yet. I want them to be successful and I want them to do the right thing…which would be a proper fix.
As far as I’m aware, getting LM onto a component while the machine is off and unpowered probably won’t be so much of a problem, so long as all the LM is removed afterward.
The LM used by framework is solid at room temperature, what I did was run the laptop hot with Cinebench, and let it cool down bottom up to let it bond more to the heatsink than the cpu die.
When I removed the heatsink, the LM was like a piece of thick metal foil stuck to the heatsink for the most part. There were a few flakes stuck to the die but I could scrape those off easily with the back of the framework tool.
The replacement is much easier than anyone might think.
3 Likes
I also used tape to help pick up the flakes after scraping. It helped me quite a bit.
2 Likes
Well i tried soldering a shim to a Spareheatsink…
Preheating the Heatsink, well went u sideways 
I was able to solder a shim, but i think it wasn’t that good of a contact area, as i was only able to hit 14k in cbr23.
I reverted back to my ptm sandwich and i will try to solder it better, but enough for now.
As i had to sacrifice both spare heatsinks, i removed the shims on both and yeah. The soldered area was to small as expected.
6 Likes