Open chrisgleissner opened 3 months ago
Rem0o
commented
3 months ago This would be a very hard feature to justify or explain to the user, since it only serves your very edge case where the control doesn't accept 0% "and" that same command results in a higher RPM than what's actually possible through a higher % command, all of which kinda makes no sense.
chrisgleissner
commented
3 months ago It seems that based on another issue in your Github repo, multiple users are affected by this.
My suggestion would automate a workaround that I (and presumably anyone else with an Nvidia 3060, possibly other models of the 3xxx range) has to apply at the moment in order to get the 0rpm feature to work properly.
The 'Force Override' feature unfortunately does not work all the time since the Gpu does not seem to respect it when certain temp or power use thresholds are breached.
My request was just a suggestion. There may be a more elegant solution.
On Wed, 7 Aug 2024, 17:57 Rem0o, @.***> wrote:
This would be a very hard feature to justify or explain to the user, since it only serves your very edge case where the control doesn't accept 0% "and" that same command results in a higher RPM than what's actually possible through a higher % command, all of which kinda makes no sense.
— Reply to this email directly, view it on GitHub https://github.com/Rem0o/FanControl.Releases/issues/2654#issuecomment-2273911991, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA6JA64YQ32MYO5YNDBAABLZQJGYHAVCNFSM6AAAAABMEBA53WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZTHEYTCOJZGE . You are receiving this because you authored the thread.Message ID: @.***>
Feature suggestion: Fallback Fan Curve
Background:
Many GPUs support a 0 Rpm mode in which its fans turn off. However, this mode is disabled on a hardware (or driver) level if certain conditions are met, such as breach of temperature of power use thresholds. If this happens, even the 'Force Override' feature of Fancontrol no longer works. For example, in the case of the Asus RTX 3060, this results in the fan RPM suddenly rising from 0 (configured via Fancontrol) to 1000 (forced by the GPU), resulting in audible noise. Other than reducing temperature and/or power use, e.g. by increasing system airflow via other fans or not using GPU features such as multiple screen support, no workaround is known.
Suggested Solution:
Associate an 'original' fan curve with another 'fallback' fan curve. The fallback is activated if Fancontrol notices that its attempt to set a certain Rpm (as configured by the fan curve) has been unsuccessful a configurable amount of time, default 10s.
Periodically, Fancontrol checks whether the original fan curve Rpm application started working again. If so, it reverts back to the original fan curve. Otherwise, it continues to use the fallback curve.
In the UI, this could be offered as a 'Fallback Curve' in the context-sensitive menu of each fan curve. When selected, a fallback curve can be chosen from a list of all defined curves.
Potential Drawbacks:
The test as to whether the original fan curve can be applied again may result in an rpm spike and audible noise. For example, in the case of the 3060 Gpu mentioned above, attempting to set 0 Rpm and failing to do so results in the gpu setting an rpm of 1000. This could be mitigated by making the 'fallback reversal' check interval configurable so the user can set it to a very large value.