Uneven CPU thermals!

I ran this through WINE on Linux, but it looks like I didn’t win the lottery with my Batch 1 unit.

image

One thing I noticed though is that the CPU temp was always at 100c and the power draw was between 45-50W, averaging in the middle. This would seem to indicate to me that the cooler is doing it’s job of dissipating 45+W and that I just didn’t win the silicon lottery… unless of course I’m not supposed to be at 100c under those conditions.

It looked like my CPU frequency was bouncing between 500mhz and 5Ghz, averaging around 3.8Ghz.

Is this relatively standard performance?

That looks a bit low.
You should have at least 15k points in Cinebench.

I just had a new test after I started the laptop and browsed for 10 minutes.

Single run, 17020 pts.

Cinebench

I did some further Testing, after my New Heatsink performed that well i wanted to examine why the old one was that bad.

Well the Heatspreader is more like a Heat"narrower".
I put a Razorblade under the Edge and tried Lifting. The Edges popped up really easy.
The Heatspreader is soldered onto the Vaporchamber, but with less Area than the literal Die Size.
The Last Pic shows a Paper in Die Size next to the Sodered Area.




Furthermore i dissassmble the Laptop once again and installed the old Heatsink with a Coppershim in 20x20mm with 0,8mm Thickness. The Shim was lapped and cleaned and put onto the Die with a PTM7950 Sandwich.

Well my new Scores are even more impressive. I am pulling 55-57W Constantly with the new Shim on the old Heatsink.
10 min RUN


Single Run

4 Likes

Well that’s a bummer. Fortunately I don’t really need the performance, the quadcore i5 Thinkpad T470 my Framework replaced had more than enough power for the simple text editing I do. I guess I’ll have to see if this knowledge gnaws at me enough to try some of the stuff other’s are doing in this thread. Realistically, I can probably deal with it until an upgrade path is released.
Thanks for the info.

To those trying the shim method, would a thin heatpipe fit in that gap as well? I realize they’re meant to move heat laterally, but I wonder if they’d be more performance across that small of a height jump than pure copper.

2 Likes

Are there any visible differences between the new unmodified heatsink and the old unmodded heatsink?

Hard to spot but yes there are. The new Heatsink has visiblr Solder on the Right edge of the Vaporchamber and the pressmarks on the upper middle edge are more pronounced.


2 Likes

okay now 2 Days later, my modified and PTM7950 sandwiched Shimmed HEatsink is better than ever.
Pulling 75W at the Beginning of the Cinebench Run Going down to stay at 56W Constant Power Draw. Freqency from 4,5 ghz Allcore down to 4,3 ghz.
I have 3 Cores Hitting a 100C and the lowest is Sitting at 93C.
That’s how it should have been from the Beginning.
Only Thing i could improve is Doing this all with Liquid Metal, but thats unnecessary.

9 Likes

What was your score on that run?

Reliantly above 16k


This is from this Morning. Just 73W Peak, but still better than stock :wink:

4 Likes

Keep this up and you can make some extra dough making modified heat sinks :rofl:

3 Likes

I am having no improvements with my newly ordered heatsink and different thickness of shims, I had tried a 0,3mm and 0,8mm. (I believe when I tested them both stock, the new one performed a little worse than the one that came with the board)
However I never sanded the heatsink, when I went with my fingers over the area that had the plate soldered, I feel unevenness, like little dimples. It could be an artifact from manufacturing when the plate got soldered on.
I ordered some sandpaper of different grades to see if I can even it out better, as these dimples might be the cause of seeing no improvements.
If that doesn’t help, this heatsink is a dud. :frowning:

Hi! i am thinking about switching to PTM7950 for mine. I am having very bad thermal issue. a delta of 30 degrees from hottest to coolest core and thermal throttle at 35w. Is there any things to beware when disassemble the cooler? Do I have to wait for the thing to cool down before doing that?

Heat it up with (optional the Midplate removed, loosen the Screws a little bit).
flip it over, let it cool.
Even than you need a little bit of force to remove the Heatsink.
Its easiest to remove it with the Expansion Bay and the Upper Cover with the Fingerprint Reader removed, but its possible to do without.

It’s pretty impressive that you were able to achieve that with a PTM sandwiched shim. Usually the sandwich would’ve lowered heat transfer by a probably verifiable amount. Wonder if that’s more an indication of how effective PTM actually is, or a knock on how poor the original spreader was.
Might have to find myself some shims to play around with for a rainy day…

1 Like

The Original Shim was soldered yes. Which means it had noticable lower Heat Resistance than the PTM. But the Area where it was soldered was significantly less than the bare CPU Die. My Shim now is 20x20mm so its Area connected to the Vaporchamber is much higher. Before it was a choke point now its a Heatspreader.
If i would somehow solder this shim to the Vaporchamber it would get somewaht better again yes.

2 Likes

I have my new FW16 in hand (Ryzen 9 7940HS with 7700s module installed) and I did some initial testing on the liquid metal but I wasn’t really satisfied with the numbers I was getting directly out of the box and after a few runs. I currently have a layer of ptm7950 on it and am doing some further testing with that, but it appears that performance so far is very close to the same as the original liquid metal.

I didn’t realize I was out of conductonaut liquid metal or I would have ordered some sooner but I should have that in a couple of days as well.

Preliminary numbers were about 15,400 CB R23 score for 10 min run, 16,400 for single cold run, and continuous power draw balanced out just below 50W package power.

PTM7950 single layer I saw bad numbers at first but after heat cycling I ended up seeing very very similar numbers to the original liquid metal. It’s impossible to get a direct comparison due to temperatures fluctuating in my house a lot lately.

PTM7950 double layer I immediately see a drop in performance and even after 15 heat cycles and a few hours of burn in it’s not as good as just a single layer.

I’ll have the new higher quality liquid metal in soon and I’ll have to make a game plan when I get it for the best way to protect the board for application (this type stays liquid at room temp). I’m hoping I can finally get scores more in line with what this CPU should be capable of.

I have not tried lapping the shim yet but I may try that if the liquid metal is not satisfactory in the end. I really do feel like the cooler here should be able to pull 60+w of heat away from the CPU. I have had laptops that seemed to have a much weaker cooling design push more than I would expect but this one won’t give the CPU any headroom at all, meaning any CPU bound load is going to ramp the fans all the way up. It’s strange because the GPU cooler can dissipate 112W of power with significantly lower fan speed while remaining easily below 80C.

@PSierra117 I apologize if you already said and I missed it. But would you mind sharing where you got your copper shims? I might buy a heatsink and do some experimenting myself.

You can find them on amazon, I found mine on ebay. I think I searched for these keywords, shim, copper, thermal and various options came up.

1 Like

Yeah same. Amazon 20x20mm 0,8mm and 0,5mm thickness. I just used the 0,8mm one and lapped it on either Side and deburred the edges. Just for sagety you should put some Capton Tape on the diodes and Capacitors next to the Die. FrMework did coat them. But better safe than sorry.

2 Likes

20-day update because I received my order with “Thermal Grizzly PhaseSheet” and decided to replace the current PTM pad.

Single Run, 16686 pts

10 Minutes, 16720 pts

1 Like