Uneven CPU thermals!

Wishful thinking…it is a fundamental design flaw! I don’t think people realize it and Framework doesn’t seem to be bothered to do anything about it other than swap another flawed board for what you have.

2 Likes

They just offered me another board replacement, a week into this one, lol.
Guess we need to wait for Ryzen 9000 series Framework 16 mainboards for an official solution?

2 Likes


The acpitz 3 seems not to be core 3. If you stress all cores only acpitz 3 heats up, other acpitzs remains the same chassis temp

wait nononono this is on linux

How do you get all of the temp zones?
lm-sensor literally does nothing, and digging around files I can only find GPU data

thats not “per core” temp - you would need amd-top to get that to show.

The values you have are the acpi temp zones.

Where did you get amd-top? I can only find AMDGPU_top and radeontop, both are about GPU not CPU

Sorry - thats the one.

It shows “GPU Metrics” but this is the CPU
It gets the core temps:

1 Like

I got the amdgpu_top and I can now read the temperature of individual cores, thank you very much. However I can’t find how to show the graph and the clock speed never below 1300MHz when idling, maybe it calculates differently compared to s-tui

@Charlie_6 The acpitz 3 seems not to be core 3. If you stress all cores only acpitz 3 heats up, other acpitzs remains the same chassis temp

ACPI is the norm that define thermal sensors. s-tui display core temp as adcpitz and hw info display core temp as core. It’s just a formatting decision.
As for why only core 3 heat up, you seem to have the same as many of us (your cores stay at 3.5 Ghz which is too low for this cpu

@Xavier_Jiang wait nononono this is on linux
How do you get all of the temp zones?
lm-sensor literally does nothing, and digging around files I can only find GPU data

cpu sensors for cpu temps are under acpitz-acpi-0

@knipp30 thats not “per core” temp - you would need amd-top to get that to show.
The values you have are the acpi temp zones.

amd_top show the same temp as s-tui, it read from the same sensors

1 Like

That’s what it seems to be the way forward most likely since nothing is being done other than the useless exercise of RMA.

3 Likes

Hello everyone! It’s been a bit over two weeks since I last gave an update regarding my system’s performance.

I last updated you all letting you know I replaced the CPU’s Liquid Metal and that I had lapped the heatsink. As of now, my CPU is still getting great performance with significantly less fan noise.

Below are the temperatures and power during a cold-boot cinebench R23 run. I have windows 11 24H2 (Build 26100.2033) with the framework driver pack installed.

I ended up scoring 16471 with the modification I did, and it has consistently been scoring similarly across my periodic testing.

I am using the Delta 240W charger and a DGPU in my laptop.

As far as core clocks go, each one has been boosting well, but I would like to get some input from others if their CPUs are all the same regarding the hardware-fused order numbers.

As you can see in the screenshot above, Core 5 and Core 1 are the preferred cores in my system. In my desktop computer’s configuration, these preferred cores are the ones that run hotter overall and can be harder to undervolt due to the power/performance curve used by the cpu.

Have others been able to get their pref #1 cores undervolted significantly using software? Does anyone have cores other than 5 and 1 as the pref cores?

5 Likes

For my machine it’s core3 and core4. I have the Ryzen 9 7940HS but don’t have the dGPU.

1 Like

The temps are much even now. Did you replaced with another liquid metal or replaced with thermal paste?

1 Like

I replaced my laptop’s Liquid Metal with copper shim/PTM sandwich (Using thermal Grizzly PhaseSheet) . I removed the stock heat spreader from the vapor chamber and lapped the area where it used to be.

2 Likes

Nice you used my Technique? Thanks for trusting this and it proves my Results :slight_smile:

My Next Mod i try is modifying one of my ecxess Heatsinks. I will try to solder a new Shim to the Vaporchamber and use Thermal Grizzly Conductonaut. I want to get 80w Peak TDP and above 60w sustained. That is my Personal Goal for my Modifications.

3 Likes

Last night I lapped my spare heatsink. It is smooth but in the reflection of the light I can see lower spots and going over with my finger I can feel lower spots, it’s minuscule but they are there.
Did your have that still after lapping, or did you go much further?
I have not had a chance to install it yet and test so I have no idea yet if it helped smoothing the surface.

1 Like

i still had some dimples after lapping the Vaporchamber. i wouldn’t lap the Shim again as it wasnt worth it. Replacing it and slapping the new Shim on the Vaporchamber (with little dimples) had the best effect.

1 Like

Yeah. I get Framework is also a victim of this, but this situation and the silence about it has so far sealed the deal that my next laptop won’t be a Framework. Which is a pity, since Framework seems to offer exactly what I want in a laptop.

But at this point, I paid >€2000 for this laptop, and after just 6 months of usage (and QC issues, something evidently wrong with the QC on this unit since rattly and buzzy noises are common during daily use, expansion cards dying on me at the worst possible moment, and countless other issues and support tickets), it physically hurts to see my friend’s €1200 7840U ultra-portable laptop with regular paste and teeny-tiny fans absolutely slaughter my twice as expensive chonky 16" liquid-cooled workstation in any benchmark and when we play together.

We’ll see. I have warranty on this thing until May of 2026, so I would hope this situation is at least addressed by then. If it isn’t, I will resell and move on and that will be that.

I’m getting really, REALLY tired of the lack of response from Framework here. All they needed to say was ‘we see there’s a problem, we’re looking into it’ and I’d accept that. This silence is immature and frankly completely defeats the point of Framework’s business model.

I’ll give it a month. If they stay completely silent and keep just offering an RMA, I’m considering small claims, 'cause I’m way out of my return window and this is frankly unfair to everyone who bought a 16. I really don’t want to have to do something like that with a company whose mission I seriously care about, but this is childish behavior.

5 Likes

RMA board finally arrived, during “normal” use temps seem more uniform across cores and the fans are very quiet in comparison.

Keeping fingers crossed it stays that way after a few thermal cycles.

2 Likes