[SOLVED] The AMD GPU idle power draw has doubled

Strauchdieb · 2025-10-30 16:06:13

Could anyone else confirm whether they have experienced a similar change?

I connected an AMD RX 6800 to a monitor with a resolution of 3440 x 1440 and a refresh rate of 144 Hz via DP.
A month or two ago, my idle power consumption was still around 7–8 W, but now it's more like 15–17 W.

At the moment, I am using a recent system and GNOME environment.
Disabling undervolting and adjusting the resolution and refresh rate had little to no effect on power consumption.

I realize that there are a multitude of possibilities and pitfalls. Perhaps someone has a suggestion.

Last edited by Strauchdieb (2025-11-06 12:10:08)

Luciddream · 2025-10-30 20:17:30

Do you have any other monitors connected? Usually this happens when the GPU memory clocks stay stuck in the highest possible setting, even if there is 0% GPU usage. Multiple monitors have been causing this in the past (e.g RDNA), and it's still not fixed for that GPU generation. RDNA 4 works fine though with multiple monitors.

Strauchdieb · 2025-10-31 19:00:34

Do you have any other monitors connected?

No, it's only this one monitor. I even deactivated VRR in the monitor settings (via the OSD, not the software settings), and then the memory stuck (?!) at the maximum clock speed without variance.

Usually this happens when the GPU memory clocks stay stuck in the highest possible setting, even if there is 0% GPU usage. [...]

I'm aware of this memory clock problem and have experienced it occasionally (for example after gaming). Switching the monitor off and on again helped. Moreover, I saw other mitigations: https://forum.manjaro.org/t/vrr-flicker … tor/174951 or https://gitlab.freedesktop.org/drm/amd/-/issues/4320 .

For clarification: this is idling now (only foot terminal opened) on my system.

GFX Clocks and Power:
	456 MHz (MCLK)
	9 MHz (SCLK)
	1825 MHz (PSTATE_SCLK)
	1000 MHz (PSTATE_MCLK)
	793 mV (VDDGFX)
	27.00 W (average SoC)

GPU Temperature: 49 C
GPU Load: 0 %
MEM Load: 2 %

SMC Feature Mask: 0x00003763a36f7dff
VCN: Powered down

Clock Gating Flags Mask: 0xb8198a0d
	Graphics Fine Grain Clock Gating: On
	Graphics Medium Grain Clock Gating: On
	Graphics Medium Grain memory Light Sleep: Off
	Graphics Coarse Grain Clock Gating: On
	Graphics Coarse Grain memory Light Sleep: On
	Graphics Coarse Grain Tree Shader Clock Gating: Off
	Graphics Coarse Grain Tree Shader Light Sleep: Off
	Graphics Command Processor Light Sleep: Off
	Graphics Run List Controller Light Sleep: Off
	Graphics 3D Coarse Grain Clock Gating: On
	Graphics 3D Coarse Grain memory Light Sleep: Off
	Memory Controller Light Sleep: Off
	Memory Controller Medium Grain Clock Gating: On
	System Direct Memory Access Light Sleep: Off
	System Direct Memory Access Medium Grain Clock Gating: On
	Bus Interface Medium Grain Clock Gating: On
	Bus Interface Light Sleep: Off
	Unified Video Decoder Medium Grain Clock Gating: Off
	Video Compression Engine Medium Grain Clock Gating: Off
	Host Data Path Light Sleep: On
	Host Data Path Medium Grain Clock Gating: On
	Digital Right Management Medium Grain Clock Gating: Off
	Digital Right Management Light Sleep: Off
	Rom Medium Grain Clock Gating: Off
	Data Fabric Medium Grain Clock Gating: Off
	VCN Medium Grain Clock Gating: Off
	Host Data Path Deep Sleep: Off
	Host Data Path Shutdown: Off
	Interrupt Handler Clock Gating: On
	JPEG Medium Grain Clock Gating: Off
	Repeater Fine Grain Clock Gating: Off
	Perfmon Clock Gating: Off
	Address Translation Hub Medium Grain Clock Gating: On
	Address Translation Hub Light Sleep: On

The memory clock drops if I open an additional application, e.g. the Nautilus file manager. This seems to differ depending on the application and window size. Even then, though, the GPU needs 12–14 W at 96 MHz.

GFX Clocks and Power:
	96 MHz (MCLK)
	12 MHz (SCLK)
	1825 MHz (PSTATE_SCLK)
	1000 MHz (PSTATE_MCLK)
	793 mV (VDDGFX)
	14.00 W (average SoC)

GPU Temperature: 49 C
GPU Load: 0 %
MEM Load: 8 %

SMC Feature Mask: 0x00003763a36f7dff
VCN: Powered down

Clock Gating Flags Mask: 0xb8198a0d
	Graphics Fine Grain Clock Gating: On
	Graphics Medium Grain Clock Gating: On
	Graphics Medium Grain memory Light Sleep: Off
	Graphics Coarse Grain Clock Gating: On
	Graphics Coarse Grain memory Light Sleep: On
	Graphics Coarse Grain Tree Shader Clock Gating: Off
	Graphics Coarse Grain Tree Shader Light Sleep: Off
	Graphics Command Processor Light Sleep: Off
	Graphics Run List Controller Light Sleep: Off
	Graphics 3D Coarse Grain Clock Gating: On
	Graphics 3D Coarse Grain memory Light Sleep: Off
	Memory Controller Light Sleep: Off
	Memory Controller Medium Grain Clock Gating: On
	System Direct Memory Access Light Sleep: Off
	System Direct Memory Access Medium Grain Clock Gating: On
	Bus Interface Medium Grain Clock Gating: On
	Bus Interface Light Sleep: Off
	Unified Video Decoder Medium Grain Clock Gating: Off
	Video Compression Engine Medium Grain Clock Gating: Off
	Host Data Path Light Sleep: On
	Host Data Path Medium Grain Clock Gating: On
	Digital Right Management Medium Grain Clock Gating: Off
	Digital Right Management Light Sleep: Off
	Rom Medium Grain Clock Gating: Off
	Data Fabric Medium Grain Clock Gating: Off
	VCN Medium Grain Clock Gating: Off
	Host Data Path Deep Sleep: Off
	Host Data Path Shutdown: Off
	Interrupt Handler Clock Gating: On
	JPEG Medium Grain Clock Gating: Off
	Repeater Fine Grain Clock Gating: Off
	Perfmon Clock Gating: Off
	Address Translation Hub Medium Grain Clock Gating: On
	Address Translation Hub Light Sleep: On

I will try out some live systems (CachyOS, Fedora 43 and so forth) and update this post with the results.

edit: short interlude
Similar results were obtained when testing Fedora-KDE-Desktop-Live-43-1.6.x86_64, Fedora-Workstation-Live-43-1.6.x86_64, cachyos-desktop-linux-250828 and EndeavourOS_Mercury-Neo-2025.03.19: 7–9 W.

Last edited by Strauchdieb (2025-11-02 10:39:56)

Luciddream · 2025-10-31 19:45:20

In my case it was the vertical blanking that caused the stuck memory clocks, similar to the first post you linked. I could use CRU in windows and fix it, by lowering the vertical blanking just enough to be inside the desired mhz range. I've been using wayland forever though, and I never found a way to update the EDID in wayland to have a similar fix for Linux. Eventually I just upgraded my GPU.

Last edited by Luciddream (2025-10-31 19:46:01)

seth · 2025-10-31 22:14:51

A month or two ago, my idle power consumption was still around 7–8 W, but now it's more like 15–17 W.
At the moment, I am using a recent system and GNOME environment.

Did you check whether this is kernel or userspace driven?
Ie. eg the LTS kernel or not-gnome (eg. weston, sway or X11)?

Strauchdieb · 2025-11-06 12:09:38

The problem appears to be self-inflicted (the LTS kernel and/or Sway will also be affected).

A brief explanation:

It is possible to undervolt the RX 6800 quite nicely, so I did this with Lact and a reduction of -100 mV. Therefore, I added amdgpu.ppfeaturemask=0xffffffff to the kernel parameters.
In one game (Total War: Warhammer III via Proton), there is a 1 in 10 chance of the game freezing at the same point in the loading screen (with a ring timeout).
In an attempt to find a solution, I changed the ppfeaturemask (for example: 0xfffd7fff, deactived PP_GFXOFF_MASK and PP_STUTTER_MODE). This resulted in the increased power consumption, but I didn't realise this was the cause.
In this context, I stopped undervolting. However, even without it, the GPU still crashes in this game. I will create an issue in the Mesa project ~~if possible~~.

Last edited by Strauchdieb (Yesterday 12:58:54)

jl2 · Today 10:37:17

However, even without it, the GPU still crashes in this game. I will create an issue in the Mesa project if possible.

Do you have other OCs, like XMP? That caused many game crashes for me.

Strauchdieb · Today 12:39:10

Thank you for pointing this out to me. It rings a bell, but I didn't venture into this corner.
I have an AMD Ryzen 7 5700X and an ASUS ROG STRIX B550-A GAMING, and I do use some of their features:

Memory: Ai Tweaker and D.O.C.P. to achieve 3,200 MHz
CPU: AMD Overclocking for ECO Mode (45W)
CPU: Curve Optimizer with a negative magnitude of 26
CPU: a positive CPU Boost Clock override of 200 MHz

Nevertheless, compared to other games, the game is caching the entire DRAM and some swap space.
So it could be a few BIOS quirks? There is a new beta update for my motherboard that I could try.

What did you do? Did you deactivate it for the memory and set the timings and voltage manually?

Arch Linux

#1 2025-10-30 16:06:13

[SOLVED] The AMD GPU idle power draw has doubled

#2 2025-10-30 20:17:30

Re: [SOLVED] The AMD GPU idle power draw has doubled

#3 2025-10-31 19:00:34

Re: [SOLVED] The AMD GPU idle power draw has doubled

#4 2025-10-31 19:45:20

Re: [SOLVED] The AMD GPU idle power draw has doubled

#5 2025-10-31 22:14:51

Re: [SOLVED] The AMD GPU idle power draw has doubled

#6 2025-11-06 12:09:38

Re: [SOLVED] The AMD GPU idle power draw has doubled

#7 Today 10:37:17

Re: [SOLVED] The AMD GPU idle power draw has doubled

#8 Today 12:39:10

Re: [SOLVED] The AMD GPU idle power draw has doubled

Board footer