my test was done on my Framework Laptop 13 AMD (delivered 1st week of January 2024).
To be honest, I was a bit skeptical, but the test results speak for themselves (2 pictures say more than a thousand words
I bought the PTM 7950 pad (30x20*0.25mm) over ebay for €11.52 inclusive postage (a Polish dealer delivered very fast to Germany within 4 days).
The most work was to clean the processor; wasn’t sure whether leftovers of the stock paste hanging at the capacitors could influence the stability, so I used Isopropyl alcohol, Q-tips and the included brush to clean the top of the processor. BTW, its die size (AMD 7840U) is 16mm x 11,5mm. This means, the (smallest) 30x20 size of the pad is enough for 2 applications.
So to make it short: The pad works! And I got the impression, that the fan operates less.
Because everybody knows that the lifespan of electronic components increases with lower operating temperatures, I can’t understand why Framework isn’t using that pads in its production as a standard (like another big laptop manufacturer is it already doing).
Lm in laptops has always produced disproportionately positive results but came with a bit of risk and mess, this stuff seems to have pretty much all the pros with none of the cons.
Irony? No, I was genuinely impressed that it looks like even chassis temp is down (the motherboard and dimms). Which alone makes the change out worth it.
I don’t know why people thought it was irony.
lets say typical room temp is 22°C, at 36°C it becomes uncomfortable to be on your lap.
So you only really can afford the chassis surface temp to increase by 14°C before it becomes uncomfortable.
The graphs there showed that improving the thermal conductivity by the heatsink got rid of enough heat to buy you a significant comfort margin, which is not what I expected.
Enough that I am considering getting some PTM7950 to try on my old MSI notebook.
Yeah, for example the 2°C decrease in the yellow line from 61C to 59C is a decrease of 3.6°F (141.8F to 138.2F) for those that can relate to Fahrenheit. That might not seem like much, but it’s nice to know the increased thermal conductivity also leads to cooler internal chassis temps, which should cause cooler external chassis temps. Good for while using on your lap, and also a good thing for battery longevity. And also might be a big difference if used lightly on e.g. a bed where the vents may be blocked.
I personally did have to take a look at the chart again to see what you were talking about for the 2°C as that had maybe detracted away from the possibly more impressive 911°C drop on the red line CPU diode sensor. Maybe that was why, but I saw no irony and I’m glad you pointed out that detail! I’m sure others may have missed it as I did.
(emphasis mine)
Perhaps it was just a miscommunication? Idk, but hey from the other side of the world, and thanks for posting your results! Good info for AMD 7840U.
And we can all celebrate the wonders of PTM7950
Edit: whoops, 11°C drop on the red line not 9°C, I temporarily forgot how to do basic arithmetic
Thanks so much for re-testing! Great to hear there was very little/no change long term from the pad!
Though interestingly, this was my biggest take away:
I didn’t realise just how much impact on thermals only a year’s worth of dust would have! If a significant thermal improvement can be had just from cleaning out the dust, that’s actually really good to know. Thanks also for providing such a good guide for how to access the fans for thorough dusting.
If y’all can come up with a suite of tests that you find useful, I might do my own test before and after dusting just to see how much dusting alone impacts thermals, before purchasing any new thermal interface materials. If such a large boost can be gotten for free literally just through cleaning, I consider that an absolute win!
My Framework hasn’t been my main workhorse as I have a separate work laptop, so there probably won’t be as much build up as somebody who does use it daily. But if even under those conditions the dust build up has an impact, then maybe having a 6-12 monthly laptop clean out would be a good idea? Am mainly thinking about just how many issues Old Laptop has with thermals which I litterally can’t fix because Old Laptop is stuck together with glue
Maybe a new thread for more generalised thermal testing based on different variable might be a good idea? There’s so many different variables and different interface materials it’d be interested to have a centralised place for folks to share their findings under relatively standardised conditions!
Set the fanspeed to some fixed value, the thermal limit to some fixed value and the power limit as high as it goes. The run a stress test and record the package power after some determined amount of time (and check it is actually at the thermal limit and not limited by something else, if it isn’t reduce the thermal limit and retry).
Imo that’s a pretty solid way to test relative cooling performance.
I did do this kind of testing between stock and ptm earlier in this thread over a range of temp limits and fan speeds (I also did the testing with lm but the results were within margin of error of ptm so I didn’t bother to update the charts).
Edit: Almost forgot, make sure you do it on the same surface, that makes a huge difference. Propping the back up just a little drops temperature by quite a bit at max fan so that’s a variable you’ll want to control.
I think I may have gotten around it using smokeless, at least I managed to do 44W for 15min now which is a lot longer than I was able to do before (I’ll do some propped up testing later, may get even more from that). You could also just stay under 80C, which makes stapm tick up again but where’s the fun in that.
Also gonna have to make sure this didn’t make idle power worse
Oh to clarify, I meant test for temps at a given power, but by “accounting for starting temps” — ensure that the test starts at the material’s respective idle temperatures. E.g. wait 5 minutes to let temps get down to idle so that the thermal material’s temperature floor can be recorded. And then set a fixed power limit so that the temperature ceiling is locked to it.
That way, we have the floor temp (idle) and the ceiling temp (for a fixed power limit) for each material, and then those can be compared. Does that make sense?
Addendum: I think power at a given temp would also be a nice, quick performance measure!
Er I meant testing for the speed at which temperatures rise over time which would affect how long a processor could turbo for (or I think the max power and the duration it can be held).
I guess that should be separated out into it’s own test? So then we measure:
the duration in seconds and the max power from idle to thermal throttling (~100C) at max load
temps at a given power
power at a given temp
Though if in 1. the max power limit can be reached below the thermal throttle threshold (and/or if the processor can turbo indefinitely), then the measure would just be the temperature at that max power limit (equivalent to 2. and 3.)
To automate this simply, perhaps 1. can be skipped entirely. We peg the fan to max, run the initial heat-soak, then step through power by 1W increments and measure the respective temperature. Which is basically:
Though with just that, we can’t gather how quickly temps rise from idle to throttle. Hm, is that even a valuable measure if we already have 2. and 3.? I kinda think no now :s
Just an update. I installed PTM7950 on my old MSI notebook, and the difference is night and day. The fans spin up so much less now…
The notebook keyboard itself isn’t getting as warm as before either. A very noticeable change.
(I don’t have hard numbers to compare it to before unfortunately, but the difference is staggering)
I had to apply it twice as I had to thin out the pads for the gpu ram to allow it to make proper contact on the gpu as the PTM7950 is thinner than what was there before.
And I got the impression, that the fan operates less.