As i said, chill mate 
and well i fully ditched windows this time. Sitting on cachyos kde with a mac inspired theme.
1 Like
I have to reinstall it sometimes because it just becomes uninstalled at some point on it’s own. So, I wouldn’t be surprised if it corrupted itself.
I think it happened when I accidentally opened the little latch/door for the interposer while the computer was running.
Yeah, it doesn’t like it when you do that. Speaking from experience.
Hi folks. I’ve been reading this thread for a while, and have a few questions:
- My CPU is routinely stuck at 100C at full power and comparing my results to some that people posted here it looks like my thermal management sucks. Here’s s-tui below running stress test maxing out at 35W, 3.4GHz at 100C with full fans on 140W adapter.
I assume it’s a reason enough to look into thermal management optimization?
- I have Framework 16 Batch 14 (shipped May 9, 2024). So I assume I have liquid metal on both CPU and dGPU. I’ve read people reflowed/resoldered the heatsink contact pad and/or added an additional copper shim/pad with PTM creating a sandwich.
Should I just get new versions of both heatsinks for both CPU and dGPU instead of trying to mess with existing ones or are my heatsinks expected to be good enough given my late batch?
Should I mess with the shims/headsink reflow at all or just get new heat sinks and PTMs?
If I am reflowing should I get shims as well or reflowing is expected to be enough?
If I get shims how likely is it that I’m going to overstress and crack the board/chip due to additional tension from the shim thickness?
If I get shims should I try soldering them to the pad instead of creating a PTM sandwich?
Whoever reflowed/resoldered their heatsink pad, what was the temperature/solder you used?
Thanks for your help!
1 Like
Yeah you do have one of the boards affected with the liquid metal issue. If you don’t want to put in a ton of effort into fixing the problem you can just buy some PTM7950 or the PTM7958 from Framework and call it good. That would work well if you’re only trying to get rid of the liquid metal issue. I’m not really sure how much of a difference there is between PTM7950 and PTM7958 but I think it’s really minor. Also you don’t need to buy any new heatsinks, the ones you have are the same as the ones in the store. The one for the dGPU never changed and never had liquid metal so if you’re having thermal issues with it I’d just replace the thermal paste or use more PTM7950. If you do want to go through the modding route for that extra peak performance then you’ll just have to make the PTM7950 sandwich. There’s almost no gain to soldering the new shim on and just adds a risk of ruining the heatsink you have. I don’t think you should worry about cracking the die with the .8mm shim either. I haven’t heard anyone crack theirs and I’ve been using the shim mod for about 6 months with no worries on durability. I’d only worry about cracking if it’s unevenly installed but that’s easily avoidable. Removing the old shim wasn’t hard too, just popped off with even lifting with a razor blade. You just gotta make sure to sand evenly.
Also if you do have the dGPU I recommend getting a higher power charger since you are leaving a lot of performance on the table. But if you don’t really care then it doesn’t matter.
1 Like
Thank so much for your answers. I do have the 240W bricks that I use when I’m using dGPU (it’s on my dock). May I bug you further?
Removing the old shim wasn’t hard too, just popped off with even lifting with a razor blade. You just gotta make sure to sand evenly.
So I’m supposed to replace the shim, not add one on top of the existing one? I assume if the existing one can be popped with razor blade it’s not soldered to the heatsink itself? Is it glued or something?
You just gotta make sure to sand evenly.
What am I sanding/lapping? The heatsink and the shim on both sides? Are you aware of any guides/case reports I can read to reproduce the process?
Thanks again!
Yeah you’d be replacing the old shim since it’s wasn’t properly attached/glued from the factory and has poor contact with the vapor chamber that it’s attached to. I lapped the copper shim with sandpaper on glass. The copper shim I used was 20mmx20mmx0.8mm. Here are the steps I took to do the mod. Uneven CPU thermals! - #1401 by zachary_stout
1 Like
I would also like emphasize making the two score lines on the heatsink before removing the old shim. It really helped with aligning the new shim.
1 Like
Thanks a lot! Looks like the old shim was soldered in and not glued. Anybody tried reflowing it instead to your knowledge?
I don’t think the original shim is soldered on. It didn’t take much force to peel it off and it took no heat at all. Those signs show it’s attached with a thermal glue of some kind. The shiny uneven area around what seems to be a dull rectangle is old liquid metal that seeped under the shim.
1 Like
Understood, thanks!
1 Like
Glad I could help 
1 Like
I just wanted to post an update on the state of the PTM sandwich copper shim heatsink mod.
I did this well over a year ago, I just installed and ran the cinebench benchmark again, to verify the state of the mod. After being in place for more that a year (can’t remember, it should be in this thread also way back) As you can see, I am getting 16k+ monitored the temps, during the benchmark, its moves between 92-96 with a few peaks at 98 but usually sitting at 92-93c.
So of anyone questioning the durability of the modification, it is still going strong.
1 Like
How long did the test run for? Was this a 10 minute burn test, or a single run? That’s a remarkable score.
The first cycle ended on 16440, I do have a 240W PSU now compared the 180W I had before, not sure if that helped, I didn’t log the power draw, but it could explain it.
That’s all the info I was looking for, thanks!
Optimization may make your computer perform more computing per W, but it has nothing to do with temperature at a given (in this case 35W) wattage. The reason your computer is overheating at only 35W is poor heat dissipation caused by liquid metal, which the PTM7958 can fix
Well, presumably, I will get more Hz per watt, which will potentially reduce peak temperature under non-sustained load as well. I’m not following your argument - better cooling will give me more GHz, more Watt and higher performance. A lot of people above also reported their systems don’t actually reach 100C at full sustained load so it may altogether reduce peak temperature overall as well.
Sorry I didn’t mean optimization is unnecessary. Optimization can reduce temperature at medium load, however, at full load you should be able to get 54W without heating the CPU to 100C, which is a heat dissipation problem.
1 Like