Hey folks, I've noticed that power-profiles-daemon recently added support for AMD's panel power saving technology with this MR: https://gitlab.freedesktop.org/upower/power-profiles-daemon/-/merge_requests/137 It appears to be controlled via a panel_power_saving SYSFS property. This reminded that Intel has a similar technology. RPL's datasheets confirm this: https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/display-power-savings-technologies/ > The Intel® DPST technique achieves back-light power savings while maintaining > a good visual experience. This is accomplished by adaptively enhancing the > displayed image while decreasing the back-light brightness simultaneously. > The goal of this technique is to provide equivalent end-user-perceived > image quality at a decreased back-light power level. > Intel® OPST solution uses same HW infrastructure as Intel® DPST. Frames are > processed using frame change threshold based interrupt mechanism similar > to Intel® DPST. Intel® OPST SW algorithm determines which pixels in the > frame should be dimmed to save power keeping visual quality (such as contrast, color) > impact to acceptable level. Since there is no backlight for OLED panels, > the power savings come solely from pixel dimming. However, it doesn't seem like i915 has any support for this. Searching online was ineffective too: - I found mentions of /sys/class/drm/card0/power/i915_dpst, but it doesn't seem to exist, at least not anymore. - A i915.dpst parameter was also brought up, but it doesn't seem to exist either. Interestingly, all mentions of dpst on Linux were referring to Android, so maybe this was a downstream thing with Android? In any case, I found this email in the archive which confirms that DPST was not supported (at least in 2012): https://lore.kernel.org/intel-gfx/50ADFFB3.8030907@xxxxxxxxxxxxxxx/ Since AMD's equivalent to DPST is now supported by their driver and being used by userspace, it seems like a sensible idea to implement this for i915. I've confirmed that this feature has existed since (at least) Haswell (released over a decade ago!) so implementing it would likely net a few energy consumption improvements for a large chunk of Intel laptops. The best approach here would probably be to expose a similar attribute to amdgpu's "panel_power_savings", with a scale that controls the feature's aggressiveness, then update userspace tools, including power-profiles-daemon, to set the value based on the intended energy scheme. Thanks for reading, José Relvas
