On the 11th gen, the CPU performance, while plugged-in and charging, is dependent on the charge level of the battery. e.g. Runs slower when battery is at 20%.
Has anyone modified the EC such that CPU performance / workload needs trumps battery charging when it comes to power allocation priority?
Ideally, I wish there’s a BIOS option, as well as an exposed EC call to toggle / modify power allocation preference.
How old is your 11th gen? If the battery is super old, the CPU power is restricted to prevent the internal resistance from dragging the voltage to below minimum level when discharging. I think it’s a bug because that shouldn’t be an issue while charging.
Kinda but on the 7x40u, and in my case more in the other direction.
You may be misunderstanding what is happening though, the drop in power limit isn’t to prioritize charging but because it can’t rely on the battery to cover spikes anymore so reduces the power limit so that the charger alone can cover everything.
That is certainly something that could be improved all over the ec but this is one of the more reasonable ones even if maybe overly conservative.
Edit: probably still simpler just getting a 100w pd charger than modding the ec for that.
They lowered psys power as well? This may slow down battery charging IIRC since psys is the total power draw from the computer
Yeah…really trying to see if there’s an upstream / non-single-point (non consumer-side) solution. And I really don’t want to throw money at solving what I consider a Framework-induced behaviour at the moment.
‘power’ management is odd with Framework in general (from my consumer-level understanding / perspective). Anything related to power is / has been quirky. Started with RTC cell battyer, USB-A voltage, USB-PD charging, USB flash drive connection stability, USB-C with iPhone, sleep power consumption with / without USB-A / DP / HDMI cards, battery charging / discharging, CPU power limit while charging (this thread)…
You could probably slightly tweak the values but at some point you have to turn down the peak power or risk the battery noping out and potentially also crashing. You could probably lower the threshold to like 10% and increase the allowed power after that a little by removing the weird 90% of negotiated power limit logic (and disabling input current prochot with battery connected but in this situation it actually would kind of make sense).
You may be able to raise some power limits xtu/intel-undervolt or something, only some of the stuff the ec sets are hard limits (in the case of my 7840u it’s pretty much just the peak power thing, the rest I can override from the os)
That is true, though it also is at least partially a consequence of doing cool stuff like having 4 pd capable ports. In most current laptops you get 1 or 2 which can be handled by a single pd controller which then directly talks to the charge controller or at least has a lot less special logic in the ec. That setup has a lot fewer things to go wrong and they can just copy proven firmware from the last generation.
Agree, It would be really nice if they put a few comments in their ec code why they are doing certain things (would probably not only help us but also internally) there is a lot of weird stuff in there. It looks like the pd-controller firmware is not developed by framework themselves and unfortunately also not open source. The whole having to sync 2 pd controllers through an ec thing is certainly not making things easy.
Well they at least learned from that XD.
Don’t think I have heard of that before.
Well at least the intels do get 5V charging XD.
I still think that is the overly agressive ocp but that is in or behind the pd controllers firmware.
Also almost certainly pd controller firmware
That was on the redrivers iirc which also have firmware that framework has pretty little control over.
This one is mostly them being pretty conservative (or a bit too spicy while having a battery available like with the amd fw 13, suuure let that thing do 175W peak power, why not XD)
It was at one point! When did that change??
I meant pd-controller at that point, the ec is fortunately open source.
At the end of the day, as a consumer, it doesn’t matter where the issue is coming from. Framework as a component integrator ought to iron them out. We’re buying a product. It’s judged as a product. Anyway, I’m getting carried away.
