[Honeywell PTM7950 Phase Change Thermal Pads/Sheets] Application, Tips, and Results

That makes more sense! :slight_smile:

Your temperatures still seem too high, I would investigate.

I dug into this a bit and below are my findings.

All of this was run from Windows 10 on High Performance profile (so idle temps were a bit higher). Ambient temps around 77F.

First run:
Starting at 43-44C ThrottleStop chip sensor (I forgot to check but I’m pretty sure all cores were below 45C). Running Prime95’s Small torture test, the first vertical line on the graph shows 86C @ 12.3W:

The next vertical line shows 89C at 58.8W:

Afterwards, it steadily rises to a max of 100C @ 59.5W.
(Note at lower idle temps, I can hit 60W+).

Second run:
I let the laptop cool a bit to 47-48C (ThrottleStop’s chip sensor and I forgot to check if all CPU core temps were above 45C. I think they were, but not sure.). Then I reran Prime95’s Small torture test. First vertical line shows 95C at 52.4W:

Afterwards it does still steadily rises to 100C. But the duration is much shorter (probably since it started at a higher 95C).

Third run:
I let it cool down a bit from the second run and then stressed the CPU in short bursts with ThrottleStop’s CPU bench (starting/letting it run for 1-2 seconds), then stopping and making sure all cores are above 45C. You can see where I did this in the graph:

Then running Prime95’s Small torture test, the first vertical line shows 95C @ 56.0W:

Afterwards, a short duration steady rise to 100C.

Fourth run:
I let the laptop cool to 42-44C (ThrottleStop’s chip sensor) and this time ensured all cores were under 45C (ranged from 35-42C).

Prime95’s Small torture test, first vertical line 82C at 10.1W.

Next vertical line 90C at 58.5W.

Afterwards, a significantly longer duration rise to 100C.

In conclusion, it seems that even with all cores above 45C, the steadily-rising-to-100C behavior remains. I also reread this post which has data on CPU change on stock vs. Kryonaut vs. Conductonaut liquid metal. It seems that all of them have the steadily-rising-behavior as well, though the duration to 100C is longer with “better” paste. So I don’t think the phase change behavior of PTM7950 changes things too much – I could be wrong/missing something though.

2 Likes

The fact that you’re seeing “steadily-rising-to-100C behavior” is really good…because you’re getting a longer 60w duration. On mine, it’s almost instantly at 100c…so my unit going stays at 60w for like 0.5 second…then start dropping TDP.

@Second_Coming oh interesting – I did another run quickly, here’s some data (my current ambient temp is ~77C):

The farthest left image is the entire run, and the subsequent images are at each next second. It seems like this run peaks at 59.5W, but stays in the ~59W range for ~3.5 seconds. Then it starts dropping steadily down to ~40.7W at the end of that 20 second duration. After that, it drops to ~28W (26.6W in the graph, but you can see how further along in the last image that it does indeed sustain 28W).

Is this a significantly longer/“better” than your unit?

1 Like

OMG, this PTM7950 is amazing! (Arrived after about 2 weeks…well worth the wait)

I’m able to boost to 63.275W, from idle of 34c (warm day). AND it stayed above 60W for a good 2.5-3 4.0 to 4.5 seconds. (Polling ever 500ms, and able to get 9 rows of results higher than 60W)

image

Think this is the end game TIM for the Framework Laptop…without going into liquid metal territory. I’m truly amazed.

2 Likes

@Second_Coming wow that’s impressive performance and I’m glad someone else with a Framework is able to confirm it working well!

I’m also curious to see how this performs with the 12th gen chips :nerd_face:

Same here. Hope someone with a 12th gen board will be able to test this out and share their experience.

I might go get the PTM7950 (I got the i7-1260p model) but Linux temperature sensor readouts leaves much to be desired. I really wish if someone can port Hardwareinfo and even things like Thaiphoon burner to Linux.

Ok I will definitely get it, the laptop was overheating resulting in random Wifi disconnects during light gaming (Overcooked 2). I had to disable all but one of the efficiency cores to reduce temps (the BIOS doesn’t allow us to fully all efficiency cores).

@Jieren_Zheng I’m looking forward to your results – a before and after would be wonderful!

I will try, Linux have horrible temperature sensor readouts as I have said above.

I think the simplier performance would be probably something like how badly it throttles with all e-cores enabled and how much cooler than stock when only one e-core enabled.

I was wondering, unless you did it already, how much wattage can you do while keeping the laptop at around 95C consistently? Like a new max sustained TDP

Yeah, I wish there was something more in-depth as well. Currently, on Linux, I’m just using s-tui (GitHub - amanusk/s-tui: Terminal-based CPU stress and monitoring utility) and its built-in stress tool and running it as root to show power usage in the bottom left hand corner. From that, I can compare temperature, CPU frequency and power usage.

I’m still wondering this as well! The i7-1165G7 is fine for my current needs and I’m not sure when I’ll have time to dive into that/find out how to do that. I think it’s some combination of Intel’s thermald (on Linux) and the embedded controller.


Here are 2 of the charts I did last night, I noticed the temperatures are both high due to the iGPU running hard, something that is present when I was using a laptop with an i7-3667U and an i5-5257U.

This was in Overcooked 2 single player, title screen. The overheating occurrence I had was when I had all cores enabled playing multiplayer and hosting the game. I notice lower temperatures when I disabled 7 of the e cores and but still observed game lock ups. I decided to not host the game with 7 e cores disabled did the game become playable. The ambient room temperature was definitely higher then too.

Notice that my CPU usage wasn’t high, so I guess the game hosting probably increased CPU usage too.

From what I see, I don’t see any indication that your system is actually overheating/thermal throttling (at least the CPU, not GPU). I’m going to try to not get too much into troubleshooting territory as that’s not the topic of this thread. That being said:

It doesn’t look like any of your CPU cores (even with all cores enabled) go above 85C – those temps are well under the 100C max. What’s your CPU frequency at 85C?

To determine if your system is actually thermal throttling, you can stress the CPU/GPU to see if it maintains the proper frequency at the given temperature. s-tui’s stress test is a simple way to check the CPU.

With the information provided, I think that your system might not actually be overheating and the wifi disconnecting is a separate issue.


But zooming out of troubleshooting – thanks for the graphs, and now there’s a “before” to compare with PTM7950!

Yea, I am trying to reproduce it frankly, it was overheating previously (I didn’t manage to measure the ambient at the place previously, it is definitely warmer than the room when I created those graphs), the sensor readouts were hitting 100-105C. And when that happened, my Wifi disconnected and I got dropped out of the game.

@Jieren_Zheng Huh, very interesting. If the CPU goes over 100C, the system should shutdown (I’m pretty but not entirely sure). I wonder what those sensors were for. In any case, I hope you come to a solution!

That’s why I was saying I am not sure that the sensor readouts were real. It could be a momentarily spike and faded out before the system took action though. I notice the system alarms had delays in reporting so I guess the temperature sampling rate isn’t as quick so users don’t get crashes for temporal spikes.

I just got mine from Aliexpress now.

Might get to work on it later tonight and see how much better it runs :slight_smile:

@Jieren_Zheng ooo, yay! I hope everything goes smoothly and am excited to see your results!