Uneven CPU thermals!

It’s the same issue with the soldering process as before.
It sounds like the test was a “quick run” to see if it would work and that’s it.

So, no long-term tests, or, at least multiple tests in 24-48h…

But, even if it passed the quick test, I wonder why nobody looked into the fact that every single board has the same issue.

Core #4 is hotter than every other core.

2 Likes

I’ll receive my new heatsink tomorrow, and, because I’ll open the laptop I also decided to try some “Thermal Grizzly Kryonaut” and a 0.5mm Copper Shim.

4 Likes

Where are you sourcing these plates from?

I decided to also further my guest to get more performance / less fan noise.

I ordered a spare heatsink, see if I can win the heatsink lottery…
If I don’t win, start working on modifications and testing.

3 Likes

Well you will let us know. You will test your already deshimmed Heatsink vs the new one i suspect?

My Personal Goal on the Framework is to get a sustainable 55-60w with even Temps to give it some extra juice in the Shortburst Territory. Like the Notebookcheck Press Review, which comfortably gets a 80w Burst and sustained 55w with 93C

3 Likes

Amazon.

Just bought this: https://www.amazon.co.uk/gp/product/B09STJRCMC/

3 Likes

Yes, I’ll test the one I already have on my laptop.

I plan to keep the new one as it is without any modifications.

3 Likes

Pardon my ignorance here - but how are you planning to implement this?

Will you be removing the shims already in place on the heat sink and placing these there instead?

I have had my replacement mainboard for ~3 weeks now and already my average R23 score is down from 15,500 to about 14,800.

At this point I am considering a move to PTM 7950 - without modifying the heatsink - but would like to see the performance from someone after a month or so before I rip things apart

2 Likes

FWIW Ive had my new board about 3 weeks - in that time my R23 score has dropped from 15,500 to about 14,800

2 Likes

Just to throw my hat in the ring, I’m having issues as well. I ended up with one core peaking at 100C, then two, then three, and after getting the mainboard and heatsink replaced for an unrelated reason, temps got much better… for a while. Now they’re worse again. I’ve literally slowly lost TDP as the days pass, down to low 40s at best at this point.

I really hope Framework finds a solution to this. These thermals are really disappointing. Has anyone considered adding more liquid metal, if not enough LM is the problem?

2 Likes

After seeing that I was not the only person who noticed a drop in expected performance and an increase in temperatures - I decided to email framework support again.

I figured I would share it with yall in the event others wanted to push for upper-level management to address this issue publicly.

Considering the fact that I did pay ~2400 for this laptop (and have a second on the way for the spouse) - I think it should only be fair for FW to address that there is an issue, and what they are doing to fix this.

Me again!

I wanted to give you an update on my situation. I have successfully swapped out the main board with the new one.

At first - the performance and temperatures were much more in line with expectations.

Pulling mid to high 15k runs in cinebench. After 3-4weeks of normal use, however, it seems to be dropping.

I am now back down to the high 14k range and my temps ramp up to 100c very quickly.

I likely will have to pursue modifying the mainboard just to get expected performance from the chip - but I wanted to let you know.

My hope is that Framework leadership can be made aware of this issue and provide a resolution that will not cost consumers more money after spending such an amount for these laptops.

As you may know, this is not a one off issue from me - there is a thread with close to 200 comments with users having similar issues. It seems as soon as someone stress tests their laptop - they are finding that they are not getting the advertised performance. Users have found that just simply replacing the LM pad with an alternative can increase performance significantly - however that costs more money, time, and risk.

Here is the thread:
https://community.frame.work/t/uneven-cpu-thermals/55614/189

Can you escalate this to someone internal to Framework to make a statement on if this issue is being reviewed or not? There are many people willing to help support this company, who want to see it succeed, who are frustrated by the silence of this issue.

Thanks, 
2 Likes

However, it does seem that the description for the heatsink may actually confirm that we are at the company’s planned thermal limit :expressionless:

The Framework Laptop 16 Heatsink is developed with a liquid metal thermal interface and three heatpipes meant to handle up to 45W continuous processor power, keeping CPU performance stable and quiet.

2 Likes

Well we answered all your questions in the last two days in this Thread. And 45w is the Thermal Limit they slap on their Description to be safe in a Legal Battle. Press Review Units did go 55w constant and up to 80w in Burst.

Edit i already escalated it with the Support and did link to this Thread Weeks ago, but its somewhere in the Liop internally.

2 Likes

Well I can say about thermal grizzly’s pad (PTM 7950) that after a week I don’t see any degradation yet.
Actually, as it has been settled in, it is performing a little bit better than initial tests.
It’s not a month but it is looking good.
If the liquid metal wasn’t seeping into the spaces besides the die and plate it would perform just a little bit better than the thermal pad I think, looking back at the first benchmarks I did when the replacement came in.

I will soon have a 2nd thermal solution coming, and I have thermal grizzly’s liquid metal ready to use with it. And will test with that setup as well.
I think if I can install more LM than Framework does, it might solve the degradation over time.
Having a second, also gives me the option to test if there’s a difference and sacrifice one for testing with copper plates removing the one glued or soldered on plate.

4 Likes

Have a look at this: Uneven CPU thermals! - #156 by Alex_Uta

perfect - i thought it was that solution but had fallen behind this thread.

Thanks for linking it.

1 Like

Yup… I missed where the shims and their play was mentioned - been a busy week and 50+comments between the initial shim theory was provided… my bad…

Regarding:
Thermal Limit they slap on their Description to be safe in a Legal Battle. Press Review Units did go 55w constant and up to 80w in Burst.

Unfortunately that also protects them in saying that the heating solution is “unsufficient”. If you get 45 watts… you are within spec. Mine has performance degredation since the mobo replacement… but Im still getting 47-51 watts… so… withing their expected results.

That being said - providing the findings you are - about these modifications, could provide a more robust cooling solution going forward.

I received the 20mm ones and because of the 20x20, it will get too close to the sides.
Unfortunately, I think that we need a 20mm x 14-17mm.

The 20x20 could still work, but, it feels a bit too risky to use it :slightly_frowning_face:

Tomorrow ( if I receive it ), I’ll install the new heatsink and see if I’m lucky or not.

If the shims prove too big, you can easily take a file and reduce the size until they fit. Just make sure to not bend them, as that would greatly reduce their function. Also, how are the tolerances and surface quality of the shims?

Not the best quality.
Instead of 0.3mm, I received 0.24-0.25mm.

Just gotta find the time to reduce the size, but yes, I could try that.

I ordered from Amazon too and especially 0,8mm ones so i can lap them and improve their Surface Finish. For the Size i don’t mind that much, i will put a 0,5mm Thermal Pad ontop of the Diodes and Resistors next to the CPU Die, so it will definitly not touch them. But if we Recall the Framework Deep Dive they already coated the Parts next to the Die as they suspected there could be leakage from the LM to improve its security.

1 Like