Hopefully it stays that way! I think the safety of PTM7950 is worth the minuscule performance difference.
4 Likes
I agree with you on this one. I would much prefer the 13 sticking to either regular thermal paste we can easily replace with PTM, or switching to PTM instead of thermal paste.
I have yet to find a single real reason to use liquid metal over PTM.
6 Likes
Has anyone else found liquid metal being applied to the stock Intel core CPUs on their framework 13 when opening it up ?
i want to give people an update regarding the initial ebuy7 link in the OP. I ordered a sheet from them in September and want to tell people that in my package there are no longer any stickers refering to honeywell and came in two sealed bags. the tools in one and the material in the other. it doesn’t initially ring potential fake material yet until i apply later today i will record my results and then say if i believe the source is now potentially fake.
2 Likes
Back to answer my own question: about 27.75x11mm. Seems safe to go a little bigger, and that helps with placement too.
My semi-scientific measurements agree with the consensus that PTM7950 helps on the 12th Gen: idle temps are 3-4 degrees cooler (~45C to ~41C in ~23C ambient). Under a s-tui stress test the core temps measured up to 5C cooler and the fan took two minutes to hit the observed max 6800rpm, whereas it got there in 30 seconds before.
Not earth-shattering improvements for my real-world use: I’d kind of hoped some of the short periods of fan noise would go away or quiet down but that part of the experience is roughly the same. Still better, though.
2 Likes
I managed to bypass the STAPM, but how did you get 45W power? I can’t unlock the power limit
I used Smokeless UMAF. After using the custom EFI image to boot, you don’t need the image further, you can boot into UMAF whenever you want on the boot menu
Bootmenu
After editing, press F10 to save, then hit ESC to return to the boot menu, go to “UEFI Misc Device 3” again, Enter.
Before editing the power goes up to 51W in a few seconds then goes down to 33W, the sustained power limit, unfortunatly it goes down to 28W due to the notorious STAPM
Decrease the power/thermal limit are without problem
Noticed that I changed the slow PPT time and it shows as well, stressing the CPU confirms that it’s working
Increase the power limit does nothing, the power is still capped to 51W → 33W as shown
I can change the time from 5s to 20s, to some extent… I didn’t increase the TjMax as the CPU can change from 40C to 90C in a second, even if I just want to see whether the displayed TjMax increased on HWINFO it’s still too risky if booting up the OS runs the CPU full tilt above 5 seconds.
The most headache is the STAPM
I set the STAPM to manual, make it not to measure skin temperature otherwise the power will go down after 1 min of gaming. I also set the system configuration to 35W and this supposedly to make the power not to decrease by a timer even if the temperature itself is lower.
28W
It didn’t work, it goes to 28W after 10 mins of gaming. The 35W in HWINFO was a lie.
Change the sustained power limit to 35000mW, doesn’t work.
Set the Tskin Time Constant to 99999s, didn’t work, can’t get rid of the countdown. Disabling STAPM boost(set Override to 1 and STAPM Boost to 0) doesn’t work, get straight to 28W, even the 33W “sustained” power was gone.
I eventually found that you need to think counterintuitively. Set the STAPM to track the skin temperature, but only take action after 99999 seconds, not tracking the temperature disables to 99999s as well(falls back to 10 min).
Gaming for 2 hours, sustained 33W, not falling back to 28W.
What if you set the fast PPT limit to desired level and increase the time? Didn’t work, the time is capped to 30s, set to 240s or 99999s did nothing.
Since you already found a way to increase the sustained power to 45W, with my finding of way to bypass the 28W STAPM, you might be able to get higher sustained power for hours and beyond.
Without tweaking these, potentials are wasted, gaming with 28W only gets me 69 to 72C without repasting or other cooling modification, and I can reduce the noise by set the fanduty to 50% and the temperature not even reaches 80C.
I think these findings can be helpful to FL16 owners as well, the 7840/7940HS has a “sustained” power limit 54W, but STAPM will cut the power to 45W after half an hour, by bypassing the STAPM, FL16 users can get the sustained power true to its name
1 Like
You do need to set that at runtime using ryzenadj or something, the soft limits get reset by the ec all the time (at least whenever you unplug/plugin power)
Can you bypass the STAPM timer on ryzenadj as well, so no need the atrocious UMAF tuning?
No you need both, you need umaf to bypass stapm and raise the hard limits and ryzenadj to set the soft limits. You don’t technically need to raise the hard limits for 45w, before you get stapm cockblocked you can set the soft limit to 45 even without umaf but it won’t hold it for long.
I don’t think smokeless umaf is atrocius, it’s actually pretty nice compared to what you have to do to get access to bios settings that manufacturers hide from you on other platforms, the atrocious bit is the bios hiding those settings from you, or the bios in general in this case I guess. The bios in the framework really doesn’t fit the ethos being so extremely locked down and featureless.
1 Like
Sorry, I was trying to say is that my steps in UMAF for bypassing STAPM is atrocious… Did I do anything wrong? I guess my “workaround” of 99999 isn’t quite neat. What did you do in UMAF in order to bypass the STAPM?
About the hard/soft limit thing. If I want to increase power, I’ll need to increase the PPT in UMAF, then increase the PPT on ryzenadj, without UMAF, ryzenadj is not able to increase the PPT on its own.
Should be able to a little bit but stapm will make it not very useful and if you are allready in smokeless might as well turn up the max power XD.
With the ec setting the soft limits back with every re-plug you need to set the soft limits when you need them. I don’t actually mind that even if I would love to have some setting to set those the presets are relatively sane and I don’t usually need the 45w+ beast mode.
-
System Configuration set to 54W and Tskin Time Constant to 99999 without disabling System Temperature Tracking, bypassing the STAPM. I don’t know whether 99999 is the max value before overflow but if you want to game for more than 27.7h continuously better to get a desktop. not required on Windows(see 3.)
-
Set the Fast PPT Limit, Slow PPT Limit and Slow PPT Time Constant to higher values,
this unlocks the hard limits,not really as you can still adjust via ryzenadj even if it’s Auto on UMAFed BIOS as long as [unlocked] is displayed on HWINFO. -
Increase the soft limits in userspace via ryzenadj to desired values. On Windows you can put in on Task Scheduler and have your settings refreshed back to your desired value 10 seconds after re-plug, and to reset stapm-timer again and again. On Linux you need to do (1.) to bypass STAPM and re-apply the power parameters after re-plugging.
./ryzenadj --stapm-limit=54000 --stapm-time=99999 --fast-limit=45000 --slow-limit=41000 --slow-time=10
excerpt from readjustService.ps1
#some Zen3 devices have a locked STAPM limit, this workarround resets the stapm timer to have unlimited stapm. Use max stapm_limit and stapm_time (usually 500) to triger as less resets as possible
$resetSTAPMUsage = $true
function doAdjust_ACmode {
$Script:repeatWaitTimeSeconds = 5 #only use values below 5s if you are using $monitorField
adjust “fast_limit” 46000
adjust “slow_limit” 41000
adjust “slow_time” 10
adjust “stapm_time” 99999
#adjust “tctl_temp” 93
adjust “apu_skin_temp_limit” 90
#adjust “vrmmax_current” 100000
#replace any_other_field with additional adjustments. Name is equal to RyzenAdj options but it uses _ instead of -
#adjust “any_other_field” 1234
#custom code, for example set fan controll back to auto
#values (WriteRegister: 47, FanSpeedResetValue:128) extracted from similar devices at https://github.com/hirschmann/nbfc/blob/master/Configs/
#Start-Process -NoNewWindow -Wait -filePath "C:\Program Files (x86)\NoteBook FanControl\ec-probe.exe" -ArgumentList("write", "47", "128")
if($Script:acSlider -eq $Script:betterBattery){
#put adjustments for energie saving slider position here:
enable "power_saving" #add 10s boost delay for usage on cable to reduce idle power consumtion
}
}
function doAdjust_BatteryMode {
$Script:repeatWaitTimeSeconds = 10 #do less reapplies and less HWiNFO updates to save power
adjust “fast_limit” 46000
adjust “slow_limit” 41000
adjust “stapm_time” 99999
adjust “apu_skin_temp_limit” 90
#adjust “any_other_field” 1234
if($Script:dcSlider -eq $Script:betterBattery){
#put adjustments for energie saving slider position here: for example disable fan to save power
#Start-Process -NoNewWindow -Wait -filePath "C:\Program Files (x86)\NoteBook FanControl\ec-probe.exe" -ArgumentList("write", "47", "0")
}
if($Script:dcSlider -eq $Script:bestPerformance){
#put adjustments for highest performance slider position here:
enable "max_performance" #removes 10s boost delay on battery
doAdjust_ACmode #set limits from cable mode on battery
}
}
I played a simulation game that’s hard on both CPU and GPU, I managed to increase the graphics by a notch(from low to medium, still worse than using dGPUed gaming laptop), when doing that with only 28W the CPU will underclock to ~1.6GHz and the game would be vary laggy as you cannot put high load on both CPU and GPU at the same time without one being underclocked to cap the total power. After increasing the sustained power to 41W I got 90~96C GPU and 96~100C CPU, with CPU frequency at 3.3~3.7GHz.
This is on stock, unmodified cooling, with PTM7950 or liquid metal, the power could be pushed to even higher. Only achievable on 100W PD(67W actual total power draw on 41W sustained APU power and full battery), the battery can only discharge at 1C(61W on FL13, 85W on FL16) when using battery the power jumps all over the place between 33W to 41W, and on Framework 60W power supply the battery is going to discharge.
Not a lot as it turns out XD
--stapm-limit=<u32> does nothing on Linux, on Windows you can set “System Configuration” or “SPL Control” to 54W/54000 on BIOS.
--stapm-time=<u32> I set this to 99999 on Windows seems to work, not tested on Linux.
--skin-temp-limit=<u32> This is power in Watt not temperature set this to a higher value(such as 50000) does work, doing this alone increase the power consumption to 33W sustained.
--apu-skin-temp=<u32> Skin temperature in degree Celsius, set this high enough effectively bypasses STAPM, I set this to 90.
Was messing with those at one point before I caved and tried umaf, at least back then they didn’t seem to do all that much.
I am still just using the power and temp limits in ryzenad and just leave stapm turned off in bios/umaf, I am using my own skin temp management (if hurt don’t put on skin XD).
I do get why they need skin temp management but they could at least have used an actual skin temp sensor instead of just guessing the skin temp.
Apparently they do have sensors but it’s not exposed. I played video games without any BIOS or ryzenadj settings. After power goes down to 28W I lifted the laptop and used two desk fans to blow into the bottom surface of the laptop, after a minute the power slowly but surely went back to 33W. If repasting/LMing lower the Tctl but power still reduced to 28W, not repasting but cools the outer surface of the laptop increase power to 33W, apparently there’s a sensor.
Are you sure that wasn’t just because it dropped below 80c for long enough?
Edit: I just checked my spreadsheet, without bypassing stapm mine can do 33W while staying under 80C (both with tpm and lm). With a temp limit set to 80 you’ll never get stapm throttled but you also won’t get a lot more than 33W with the stock heatsink.
No, and paradoxically, using a desk fan increases the CPU temp!
Notice and compare CPU (Tctl/Tdie), CPU Skin Temperature, CPU Package Power, CPU PPT SLOW Limit % and APU STAPM Limit %.
As shown, when running GPU load, the temperatures were well below 80C but STAPM capped power to a solid 28W and the cooling fan ran at max RPM, using a desk fan to cool the bottom surface increased the power between 28W and 33W. When running CPU load, the STAPM didn’t cut power all the way and the fan ran slower despite the CPU temp (Tctl) is much higher. Using a desk fan to cool the skin even makes the Tctl temp increased to 100C, triggering TjMax thermal throttling.
My explanation is the temperature between junction and skin depends on thermal conductivity. The better the thermal conductivity the lower the CPU temp the higher the skin temp. in the Ryzen 7, the iGPU has larger surface area than the CPU, thus at a given cooling system, both thermal conductivity(K) and thickness(d) are constant, for iGPU load, the cross-sectional area(A) is much larger than CPU load. For single-core CPU load, the A is even smaller. That’s probably why High CPU temperatures at low utilization. The fan of the FL13 does not consider Tctl, that’s why the fan slows down on CPU load.
In conclusion, improve the thermal conductivity by repasting/PTM7950/LMing will reduce the CPU temperature, but it might increase the skin temperature, making bypassing the STAPM limit even more necessary. This applies only to FL13, on FL16 the opposite is true as the power is often capped by Tctl due to uneven thermals, in this case, repasting increases CPU power.
I wonder if we are using different sensors, I have never had stapm throttling under 80C but I am not 100% sure what sensor I looked at for that.
Looking at the stapm throttling it really was just decrementing the power by some factor over time as slong as it was over 80 and incrementing when under 80.
While I do agree that ptm/lm increases skin temperature I think it is because you have a lower fan speed for the same power output.
It’s also why you should not run the fw passive even if it can totally run like 10-15W without any fan pretty much indefinitely, the skin temp just gets extremely/illegaly hot.
Given it is pretty much not necessary at all with the stock paste that is a pretty reasonable statement XD.
In my case I have a dbrand skin on it which makes skin temperature a lot less of an issue for me but an even bigger one for the computer XD
