I just switched from the stock thermal paste to thermal grizzly kryonaut (propably overkill for such a low power chip) and I have core to core deltas of 6°C under full load. Would you consider this too much?
Under Cinebench I reach about 70°C after the turbo duration and about 85 while turboing to 3.8GHz. This is in a 24°C room.
Should I redo the thermal paste application?
What do your temps look like?
Thanks for your help
Have you done a cinebench run with to old thermal paste you can compare against?
Generally speaking i would say everything is fine. Assuming the 70°C / 85°C are max temps during the run, you have a good amount of headroom before throttling.
Cant really comment on the core to core deltas though.
Given that two CPUs from the same wafer can have significant different thermal / power / performance properties, i would assume that the same goes for two cores on one cpu die.
I know that Intel (at least on their Desktop platform, dont know about mobile) prioritizes their cores by their max boost frequency.
E.g core 1 can boot up to 5.2ghz, core 3 can only go to 5.0ghz, so in single threaded workloads, core 1 would be prioritized to do the job.
So if cores on a die can have different performance properties, why wouldnt they have different thermal properties as well
Unfortuntatly I didnt look into the core to core deltas before changing the thermal paste.
It’s likely cooler mounting pressure. It needs to be EXACT.
I would bend the CPU retention arms (numbered 1-3) a little to get a higher mounting pressure. I did this on standard paste and got 500+ higher cinebench scores with cores within 5C of each other. Also make sure to start with screw 1 and finish with 3 when tightening down.
I am not thermally trotteling, so I dont think more mounting pressure will give me better thermals, but maybe I will try that
Generally speaking more pressure means greater thermal transfer. See what happens I guess
Oh, that was a mistake on my part, I meant to say “I dont think it will give me better performance”. Better thermals is quite possible though
Not even when it’s boosting? (to 60w TDP)
It throttled with the stock paste, but not with the new one. But my chip also doesnt take 60W. It is only supposed to boost to 3.8GHz under all core load and it takes about 40W for that.
Oh…I see. Then yeah, the heatsink & fan in the Framework laptop, during my week-long test with various thermal paste, has shown that it can sustain about a load up to around 34W. Anything beyond that for extended period (say after 2 minutes)…will eventually reach thermal throttling.
OOTB, your boost period is around 28 seconds, laptop has sufficient thermal headroom for 40w of heat to sink for 28 seconds.
Is there a way to elongate that time, since you said ootb=
Typically via ThrottleStop, but I’m not able to do that on the Framework Laptop. Not sure why… I might be doing it wrong.
I tried that, but it didnt really work for me. Have you managed to increase the duration?
No, not on the Framework Laptop, unfortunately.
@Kieran_Levin , any idea why we can’t seem to adjust tau / turbo time limit?
For example, I would like to adjust mine to only boost to 34W…but for longer than 28 seconds, say, 56 or even 80 seconds.
Have there been attempts on turning undervolting back on?
Maybe this is possible through some bios mods
You can’t, you just can’t… (It’s intel who disabled it)
Didnt intel disable it in BIOS?
Surely there must be some good bios modders who could undo it?
Or is there a different “place” where it is disabled?
If I’m not mistaken, the BIOS is signed. You can’t ‘mod’ anything…
Unless you have one of those three mainboards that were sent to coreboot developers.
If you have access to a quantum computer (PQC related)…then sure.
ah, i didnt know, that the bios has to be signed.