I decided to try this stuff out on my AMD Framework 13…
It seems to me that it’s working so well that the fan curve is now way off.
It takes longer before the fan ramps up than it did previously. Which means that the aluminum casing is (subjectively) warmer than it was previously… and that when I run something like Cinebench, it’s actually spiking up to like 95C before the fan kicks into a higher speed and eventually brings it down to 79C. (Screenshot below is from an earlier run).
I’ll probably pull off the heatsink and try again in a few days with a new piece to be sure I didn’t screw up the install, but it really feels like the fan is just confused by the low temperatures on the other sensors since it stays absolutely silent for so long after the initial load starts. (Not sure what that misread package temperature minimum of 5C is about either).
I’ve repasted mine with liquid metal and it’s showing this behaviour as well. I think it’s just for some reason, there’s a huge delay between the temps rising and the fans kicking in. It could be intentionally so that fans turn on less, hence a quieter experience.
I just had mine arrive, took a gamble and luckily I had plenty more than I needed. I’ll include some pictures at the end of how big the die was and how much of my 30mm*30mm was left afterwards.
Frankly I’m perplexed by these results. Honestly at first I thought I must’ve messed up at plugging the fan back in cause I couldn’t hear it at all while running cinebench, but I checked and it was at max cpu usage and around 60-65 degrees and not increasing. Mind you I did all these tests at Best Power Efficiency because I forgot I had it set before and couldn’t redo the Before tests. Anyways, I’m perplexed because while my temperatures and noise were downright magical in their improvement, it also benched far worse and didn’t reach the same power draw (multicore 567pts before, 488 after. 25.97W before, 21.85W after. Both plugged into the wall at 80% charge (my max)). Anyone have any idea why this might be? I know ptm needs to be broken in and heat cycled a few times to reach its peak, but it already had the amazing cooling, but still benched lower. Ran a second time after letting it cool some and got similar results. Anyone know what might be up with this? Ryzen 5 7640u, 16gb*2 4800mhz cl40
Total stats, multicore:
Before: 567pts, max cpu temp 76.9C, max VID 0.925V, max power (package) 25.97W (cores) 24.52W
After: 488pts, max cpu temp 65.9C, max VID 0.906V, max power (package) 21.82W (cores) 20.65W
As silly as it might sound, have you tried rebooting?
Windows can be weird sometimes.
Otherwise, have you checked peak temperatures, and if utilization is 100%?
Might be worth it to sanity test with another stress test as well.
Heya @Michael_Wu ! Did you happen to fully retest the thermal pad on your original mainboard after the 1 year mark? Would love to see if it actually held up over time!
Have been investigating thermal paste options for laptops, mainly for my (non-Framework) work laptop (though having leftovers to apply to my personal Framework would be a bonus too!) and most comparison posts don’t do longterm testing or even just laptop testing at all. These pads seem super promising, so would love it if you had the data to confirm either way!
Heya! Last I checked, the thermals were still great, but I haven’t fully retested to see how it compares to my original post. I swapped in the AMD mainboard (there was a bit of dust on the old mainboard’s heatsink) before leaving for travel and have/will be busy for a bit. I’ll have a chance when I get back in about 2 weeks to swap back, retest, and post here (will tag you). Also will probably experiment with PTM7950 on my AMD board.
@jhoff80@TheTRUEAsian@a4955 any updates/are y’all still experiencing the weirdness y’all posted about? I’m wondering if the stock BIOS has fan curves that are tuned specifically for the stock paste which don’t jive well with PTM7950/liquid metal? I’m not sure how that would be though…
This might give some clues/insights: what temp/frequency was it throttling at stock vs. aftermarket? Is aftermarket somehow letting the CPU reach higher frequencies because the overall temps are lower, and perhaps it’s programmed to throttle more from a higher frequency?
For example,
stock paste allows a max of 5.0GHz, which thermal throttles down to 3.3GHz sustained
aftermarket allows a max of 5.1GHz, which thermal throttles more, down to 3.0GHz sustained
Which could cause the aftermarket paste to perform lower, and explain this:
but I’m totally spitballing here; I’ve never seen that behavior and don’t think it makes much sense, but it’s a possibility?
Aftermarket may also allow the CPU to stay at higher frequencies for longer, resulting in warmer overall thermals (and thus a warmer aluminum chassis). With stock, it may only be able to stay at 5.1GHz shortly and may reach over 100C or whatever the thermal limit is in 1 second. While aftermarket may be able to stay at 5.1GHz for 3 seconds.
Guessing here as well, but I think that may be what’s happening with this:
and higher internal temps may cause other things to underperform? IMO probably wouldn’t, but this could easily be tested by running the mainboard outside of the laptop.
Definitely still experiencing it. I remember reading in another thread that this delayed fan behaviour is intended (to prevent the fan from constantly ramping up) and apparently it’s based on a temperature sensor that’s located farther away from the cpu. I would say my temps are a lot better using liquid metal compared to stock. While gaming, it’ll draw 35 watts with an average temp of about 70 C while on stock paste, I never saw it go above 30 watts.
@TheTRUEAsian hmm interesting and that would make sense. So with the higher draw, I’m guessing you’re not also experiencing the better thermals but lower benchmark numbers/performance oddity that @a4955 was? And that you’re actually getting better thermals and performance with liquid metal?
Yup. What I think may be involved as well is hotspot temperature versus core temps. While initially booting, I’ll see the hotspot temp get to around 100 C while the core temp is around 55 C to 60 C before the fans begin to really ramp. Afterwards, while core temp will gradually rise to 70 C, the hotspot temp will drop a lot more quickly. Another thing is giving the TIM time to “break in”. I know that PTM7950 requires a few heat cycles for it to work its best but I didn’t expect the same for liquid metal. After a week of using liquid metal, I decided to repaste it (was going to do another mod but then realized it wasn’t necessary) and after using fresh liquid metal, my core temps increased by 10 C which was really weird. After giving it a few days, it went back down to 70 C average.
Yeah, at this point I’m fairly sure this is the cause more than anything else at this point, that it’s basing the fan on a sensor that no longer heats up as quickly as it does when stock.
For an example, if I run a quick Cinebench 23 run, the fan stays inaudible the entire run, even as the temperatures spike up like this. But a second run in a row the fan eventually kicks into gear. The first run is actually a little slower (like 200 points) than the second when the fan is running, which is kind of crazy since it’s the opposite of how computers usually behave.
I’d guess some micro-movements/spreading/settling in-between the two surfaces.
This wasn’t happening on stock right? I’ll test before I install PTM7950.
Interesting, thanks y’all for the info/reassurance that it won’t completely make things haywire. Might be able to also manually warm/cool up said temperature sensor to confirm/test behavior. If there’s anything I should test since I’m still on stock thermals, lmk!
I’ve now upgraded from Intel 11th Gen to AMD 7840U.
Ran the 7840U for a while in stock paste and switched to the Honeywell after two days. Not particularly scientific but looking at HWiNFO I would estimate a 3-4c drop at idle/browser workloads.
Running games with my eGPU attached however is a weirdly binary change - the fans just don’t spin up loudly anymore when the CPU is running at 25% (about 15W). Does’t seem to cross whatever threshold that is in the fan curve anymore.
Really nice change in terms of subjective experience. I can now also go and work in the local library all day without a charger, but that’s just the AMD difference I suppose!
I just tried ptm on my amd framework 13 and am a mix of impressed and disappointed, it is pretty nice but calling it lm without the drawbacks may have oversold it a bit. At high fan speeds it makes a ton of difference, at low to no fan speeds there is very little.
Here I mapped the the achived power running stress (–cpu 8) with the power limits set as high as they go with different fan-speeds(off, about 2100 which is as low as it’ll go and not audible to me and just short of 7000 which is as high as it’ll go and pretty audible) and temperature limits. I recorded the power after 10min of heat-soaking from the last step. Ambient temps were between 23C and 23C
In the case of 7000 rpm I seem to have run into some other wall, not sure if it is skin temperature or vrm throttling or something else entirely but over time the power limit would just reduce down to 30W at higher temp limits. in the case of the 7000 ptm result I was not able to run the 97 and 100C tests because of that limit even when pre-cooling before. The stock paste barely managed but it didn’t go much higher than the 30w hard limit anyway. Still though, massive improvement.
The min and no fan results are a bit less amazing but at least not worse.
I am kinda tempted to test lm but this is the first new laptop I have had in ages so I may want to baby it a bit more first XD.
If framework shipped this thing with ptm stock they could have probably avoided the “relatively loud fan” cmplaint in the reviews. I personally don’t find it all that loud and am pretty happy that it is able to cool this much. My t480s with liqiud metal and undervolting barely manages 30w with maxed fan and that one blows the hot air straight at your mouse hand.
Strongly advise against it. I Liquid Metaled an 11th gen and fried the motherboard. It wasn’t poor application, the application was perfect but I dropped the laptop and the metal seeped out and got past the conformal coating. Ended up killing the board like I said. The risk makes it not worth imo
Not really, I don’t have that concern if it’s factory applied for multiple reasons. The first and most obvious is that if a board gets fried, then it’s likely Framework who would pay to replace it in a warranty scenario. Not so with me.
Less obvious is that Framework will have taken steps to design the cooling system to hold the Liquid Metal in place and prevent it from coming out from underneath the fan, which they did not do for the FW13 because it wasn’t designed for that.
I put a honeywell pad on my AMD Framework 13 and it seems pretty good I didn’t really check the thermals before, but after they stay around 75ish under load while playing Baldur’s Gate 3, and the fans seem quieter/more consistent, so that’s been rather nice! Totally anecdotal report, but there you go.
I didn’t really check the thermals before, but after they stay around 75ish under load while playing Baldur’s Gate 3, and the fans seem quieter/more consistent, so that’s been rather nice! Totally anecdotal report, but there you go.