Uneven CPU thermals!

I wonder if that attachment is just glued to the plate.

While playing a game, I can see that the power used for the CPU hit 82.58W while the max CPU temp was 83.5, so I can’t complain about this.

Power usage

Could you double-check those numbers using HWiNFO64 instead? Something doesn’t seem right with those readings.

2 Likes

Yep that’s what i wanted to say. Above 80w is impossible even above 70 is far out of Reach. 65w is the Coded Limit inside the Bios unless there was a tamper with smokeless_UMAF. But naah there would be Thermal Throttle even at the best Thermal Transfer above 60w
But well @Alex_Uta you could increase you CPUs efficiency as you got the Ryzen 9 and you can undervolt :slight_smile:

1 Like

Yeah, those numbers are quite off :sweat_smile:

HWINFO64

@PSierra117 I’ll need to look into it (undervolt).

1 Like

I just tried a cold run and you’re 100% right.
It started with a max CPU Package Power of 64.869 W

Cold run


1 Like

I was away for a holiday, so I was catching up on this thread, seeing very interesting comments and results.

I have replaced the LM with some thermal pas I had laying around, and I have 16k+ results again in CB r23, results I had only seen the initial runs just when I got the replacement board.

I made some pictures of when I removed the LM.




I used a bit of brainpower, I ran the machine hot with CB R23 and flipped it and let it cool down so I could easily remove it, and that worked. Could be a useful tip for anyone looking to remove the heat-sink and do some servicing on it.

Most of the LM ran to the sides and a little layer was left touching around 50 % of the die.
You can see the peel I got in the pictures.
I think Framework is not using enough lm to account for the out flow to the sides once it liquefies.
This would explain initial results doing great, but once the LM runs thin and “leaks” to the sides you get worse thermals and or performance.

So I ran new benchmarks after I replaced the LM with the thermal paste. I noticed 2 things, my max temp on the sensor4/core4 didn’t reach 100c, instead it touched 97c. The big delta is still there. The fan ramp was much slower, and ran longer at lower rpm (from what I could hear).
This leads me to believe the plate that is attached to the head-sink is making sub optimal contact at that exact space where the core is.
I read LM could damage copper easily so I understand there needs to be another metal in direct contact but something is not right here.

I have ordered some new thermal paste, liquid metal and pads from thermal grizzly to do more testing. I also need more paste to redo the dGPU so I needed more thermal paste anyway, just added the rest to the order, I can always use it on other machines I have.

I might also decide to lap the heat-sink but I want to do more testing first. The thermal paste I have now is at least working and makes the fan run at lower rpm and or less frequent.

6 Likes

Great Workaround for loosening the heatsink. And good Ideas for the Pool. “Not Enough LM” was also one of my first suggestions after i started digging into that Topic. Great Pictures. I Hope some Guys of the Tech Compartment are watching in this Thread. As it seems most of the RMA Work and Fault Isolation is outsourced with such a little Team Framework has.

2 Likes

Yikes. I just downloaded Cinebench R23. I did two tests. 12,841 and 12,811. Bouncing around 38-40 watts.

Edit. Doing another run now. I realized I had Windows in “Balanced” power mode. I put it to best performance and I’m running another test. Right now the fans are at full tilt, the bottom of the laptop is quite warm (it’s sitting on a stand), but the air coming out of the vents feels like ambient temp. Package at 100C, one of the cores is at 100C. A couple other cores are in the high 90s. A few cores are in the low 80s, high 70s. Room temp is roughly 22.5C (72-73F).

Not sure if the score will be different in best performance mode, but the temps are roughly the same as during the last test, and if anything, the wattage and clock speeds are a tiny bit lower. Pretty darn close though. The utilization is the same as well. If it scores higher, I’m not sure where it’s finding that performance, lol.

Interesting. In best performance mode, it did get a much higher score. 14,157. Despite the temps, the wattage bouncing around 37-39, and core clocks hovering right at 3,700-3,800 Mhz.

Edit: I don’t know what my problem was when feeling the air temp that night. I checked it again the next morning, because it didn’t make sense to me that it wasn’t warm at all. And sure enough, the air coming out of the side vents (from the CPU) was definitely quite warm during Cinebench runs. No idea why I didn’t feel it the night before.

@PSierra117 about this…

I removed the extra bit that was on the heatsink and, it won’t actually work.
After removing that part it decreased the pressure and now I won’t have the same performance as before.

I think that the problem is caused by the glue that’s used, as it won’t be quite even and it won’t allow a perfect heat transfer.

@taezea good job with yours! I would just keep it that way.
After modifying two different parts I can say that it won’t get better.

Heatsink

Part removed

I tested this with two different parts, and both of them are quite the same.
Now I’m waiting for a new heatsink :upside_down_face:


Heatsink Issue

As you can see in this picture, after removing the extra part the heatsink is not quite flat.

Heatsink Final

4 Likes

Okay wow thanks for the Work. Did you already buy a Spare Heatsink? Or why do you habe 2 of those shims.
Hoe did you get it off?
Is it really glued to the Vaporchamber? If yes thats more than dumb. Because why do they use LM on the Die if the next Heattransfer step is blocked by thermal glue.
It would be a great idea to get that part soldered to the Vaporchamber. But soldering on a Vaporchamber is like fiddling with a Pipe bomb. You need solder thats melzing at pretty low temps. If its heated to hot it explodes.

You can overbend the holding mechanism away from the board to increase the Mounting pressure. Or you could use a new Coppershim with ptm on both sides to overcome the gap.

Yeah, I modified a new heatsink and the original one. So, now I’m waiting for the 3rd heatsink, which, I don’t really plan to modify.


It was quite easy, just gotta slide a blade under it for like 1.5 - 2 mm and after that either continue with the blade or, just use the plastic part from the screwdriver to lift it up.


It is not a solid block… so, I’d only assume that’s some sort of “special glue”.


But, at the same time, you don’t want it to get soft at 100C.


A different problem is that after you finish making it nice and flat it will be quite thin and there will be a weak point.
I was thinking about using a shim but, I’m not sure if the heat transfer is as good and, also there’s a risk that it can move.

2 Likes

So that basically means that that plate that’s installed to protect the copper from the corrosive reaction liquid metal would have with copper is also causing sub optimal heat transfer at certain points in the 2 heat-sink/pipe solutions. For me this was at the spot where core sensor 4 is seated, both times.

Normally you would nickel plate the copper to have a protective layer and not put glue and plate to the copper increasing the heat resistance/heat transfer.
I wonder why they have chosen for this solution. Now when you remove the plate you will have issues with the fact that you reduced the thickness and thus the mounting pressure.

Maybe that could be solved, that is something to look into possibly?

I have now ordered 2 Heatsinks an Clear Keyboard and RGB Macropad. I keep my current one as is and take one of the new Heatsinks for more destructive Mods :wink: and some Coppershims and so. Lets see i’m on Fire. I need 60w constant TDP :stuck_out_tongue_winking_eye:

3 Likes

There’s a bit more to it. Unfortunately, it doesn’t look like we can find an easy solution.

I decided to remove each layer from one of the heatsinks and, as far as I can see, there are 4 different layers + the base.

@PSierra117 please have a look at the following photos.
I feel that this could help you find a solution.

Layer 1

Layer 2

Layer 3


Layer 4


4 Likes

Is that a copper mesh?

I’m quite puzzled by this design.

Yup. Layers 3 and 4 are a cooper mesh.

Same. I was expecting to find a block, not multiple layers with some empty space between layers 3 and 4.

1 Like

Well i knew before your Pics how it looks inside there. Because its a Vapor Chamber. There is near Vacuum inside and a tiny Amount of Liquid inside the Coppermesh. The Liquid evaporates on the Hotspot and liquifies on the Heatpipe Area and through the Copper Wick its transfered back to the Hot Area.
This is alot faster than transferring Heat through a solid Cube of Copper. A Vapor Chamber is just another Form of a Heatpipe.
Here a great Video which shows the Principals and the insides of Heatpipes.

What’s Inside the Worlds’ Fastest Heat Conductor? - Youtube “The Action Lab”

*edit
The Vaporchamber works perfectly fine, It’s just the added shim+LM+ Glue? where the Problem is. If the Vapor Chamber wouldn’t work as intended there would be all Cores sitting at a 100C. When a CPU has a high Delta between Cores there is a big variance on Heat Transfer directly above the Die. When all Cores are perfectly sitting under a copper Block they are all nearly at the same Temp until the Block is Heatsoaked and than starts throttling. Thats why the Heat has to be force away as fast as possible. A Vaporchamber+Heatpipe Design is therefore quiete proven. When the Air coming out of the Device is not really Hot but the CPU is than there is one Step between Transfers where its to slow.

3 Likes

could you measure the shim? Than i can order some CopperShims that Size and thickness and maybe some shims a bit thicker, for trying some different Pressure :slight_smile:

18mm x 13.4mm with a thickness of 0.4mm on the outer part
When I measure with the glue part, the thickness is 0.45mm

You could easily go with 21mm x 16.4 mm while trying a thickness between 0.45 and 0.55mm.

2 Likes

would 15x15 fit the die? I can only Source 20x20mm or 15x15mm in various thicknesses.
edit:
Well what a Question, i just get the 20x20 one. It fits save :man_facepalming: