mirror of
				https://github.com/NixOS/nixos-hardware.git
				synced 2025-11-04 17:27:14 +08:00 
			
		
		
		
	
		
			
				
	
	
		
			101 lines
		
	
	
		
			3.8 KiB
		
	
	
	
		
			Nix
		
	
	
	
	
	
			
		
		
	
	
			101 lines
		
	
	
		
			3.8 KiB
		
	
	
	
		
			Nix
		
	
	
	
	
	
{ lib, ... }:
 | 
						|
{
 | 
						|
  imports = [
 | 
						|
    ../../../common/gpu/nvidia.nix
 | 
						|
    ../../../common/cpu/intel
 | 
						|
    ../../../common/pc/laptop/acpi_call.nix
 | 
						|
    ../.
 | 
						|
  ];
 | 
						|
 | 
						|
  hardware = {
 | 
						|
    nvidia = {
 | 
						|
      prime = {
 | 
						|
        intelBusId = "PCI:0:2:0";
 | 
						|
        nvidiaBusId = "PCI:1:0:0";
 | 
						|
      };
 | 
						|
    };
 | 
						|
 | 
						|
    # is this too much?  It's convenient for Steam.
 | 
						|
    opengl = {
 | 
						|
      driSupport = lib.mkDefault true;
 | 
						|
      driSupport32Bit = lib.mkDefault true;
 | 
						|
    };
 | 
						|
  };
 | 
						|
 | 
						|
  # Sleep
 | 
						|
  # -----
 | 
						|
  #
 | 
						|
  # The system will not resume from sleep properly while on battery power in
 | 
						|
  # either offload mode or sync mode.  When it tries to resume, it gets to a
 | 
						|
  # state with a cursor in the top left hand side of the panel, the power LED
 | 
						|
  # goes from flashing to solid, and thereafter cannot be interacted with (even
 | 
						|
  # over SSH) and must be power cycled forcefully.  Sometimes it doesn't even
 | 
						|
  # finish going to sleep before this behavior kicks in.
 | 
						|
  #
 | 
						|
  # When on AC, the machine either wakes up from sleep before being asked to
 | 
						|
  # (or maybe never gets to sleep state), or it goes into a sleep state and it
 | 
						|
  # appears consistently resume properly when it does.
 | 
						|
  #
 | 
						|
  # But the machine actually sleeps and resumes reliably when tlp is disabled
 | 
						|
  # fully or partially.  Disabling RUNTIME_PM and AHCI_RUNTIME_PM appears to be
 | 
						|
  # enough to allow it to work when tlp is active.  I couldn't figure out a
 | 
						|
  # more granular way to get it working, despite trying to do a per-device
 | 
						|
  # binary search via powertop.
 | 
						|
  #
 | 
						|
  # My personal configuration to make this work looks like this:
 | 
						|
  #
 | 
						|
  # {config, lib, ...}:
 | 
						|
  #
 | 
						|
  # {
 | 
						|
  #   services.tlp = {
 | 
						|
  #     settings = {
 | 
						|
  #       # DISK_DEVICES must be specified for AHCI_RUNTIME_PM settings to work right.
 | 
						|
  #       DISK_DEVICES = "nvme0n1 nvme1n1 sda sdb";
 | 
						|
  #
 | 
						|
  #       # with AHCI_RUNTIME_PM_ON_AC/BAT set to defaults in battery mode, P51
 | 
						|
  #       # can't resume from sleep and P50 can't go to sleep.
 | 
						|
  #       AHCI_RUNTIME_PM_ON_AC = "on";
 | 
						|
  #       AHCI_RUNTIME_PM_ON_BAT = "on";
 | 
						|
  #
 | 
						|
  #       # with RUNTIME_PM_ON_BAT/AC set to defaults, P50/P51 can't go to sleep
 | 
						|
  #       RUNTIME_PM_ON_AC = "on";
 | 
						|
  #       RUNTIME_PM_ON_BAT = "on";
 | 
						|
  #     };
 | 
						|
  #   };
 | 
						|
  # }
 | 
						|
  #
 | 
						|
  # I'm thinking this is too aggressive to put into shared config, and folks may
 | 
						|
  # be concerned with the hit on battery life.
 | 
						|
  #
 | 
						|
  # throttled vs. thermald
 | 
						|
  # -----------------------
 | 
						|
  #
 | 
						|
  # NB: the p53 profile currently uses throttled to prevent too-eager CPU
 | 
						|
  # throttling.  I understand throttled to have been a workaround solution at
 | 
						|
  # the time the p53 profile was created (throttled's original name was
 | 
						|
  # "lenovo_fix").  thermald would have been preferred if it worked at the
 | 
						|
  # time.
 | 
						|
  #
 | 
						|
  # I read
 | 
						|
  # https://wiki.archlinux.org/title/Lenovo_ThinkPad_X1_Carbon_(Gen_6)#Power_management.2FThrottling_issues
 | 
						|
  # as saying that thermald is fixed under the circumstance that led to the
 | 
						|
  # development of throttled given version 5.12+ of the kernel combined
 | 
						|
  # with version 2.4.3+ of thermald.  At the time of this writing, the
 | 
						|
  # stable NixOS kernel is 5.15 and 2.4.9 of thermald.
 | 
						|
  #
 | 
						|
  # In the meantime, I also ran the "s-tui" program which can stress test the
 | 
						|
  # system, while eyeing up the core temps and CPU frequency under three
 | 
						|
  # scenarios: under thermald, under throttled, and with neither.  None of the
 | 
						|
  # scenarios seem to have massively improved fan behavior, core temps, or
 | 
						|
  # average CPU frequency than another.  The highest core temp always seems to
 | 
						|
  # hover around 90 degrees C, the lowest CPU Ghz around 3.4 on a 3.8Ghz machine.
 | 
						|
  #
 | 
						|
  # I ended up choosing throttled because subjectively, the fans seem quieter
 | 
						|
  # when it's stressed and it allows the average temps to get a degree or two
 | 
						|
  # higher when running throttled than when running in the other two scenarios,
 | 
						|
  # but still substantially under critical temp.
 | 
						|
 | 
						|
  services.throttled.enable = lib.mkDefault true;
 | 
						|
 | 
						|
}
 |