Frequently Asked Questions

ddcutil does not identify a DDC/CI capable monitor.

Check the following:

  • Is DDC/CI enabled in the monitor's On Screen Display?
  • Is driver i2c-dev loaded?
  • Does the current user have permission to access the I2C bus associated with the monitor?

Command ddcutil environment will analyze your system environment and make suggestions. See enviroment Command

ddcutil doesn't work on my laptop.

Laptop displays do not support DDC/CI.

But recent laptops use Embedded Display Port (eDP) panels which have an I2C interface. Why doesn't ddcutil work?

The I2C connection on eDP panels implements slave address x50, which allows programs to read the EDID in the expected way. However, it does not implement slave address x37, i.e. it does not support DDC/CI.

If eDP panels did implement DDC/CI, ddcutil would support it. For each I2C bus that could possibly be associated with a monitor, as part of display detection ddcutil attempts to read the EDID at slave address x50. If successful, it then attempts to communicate on slave address x37. If that fails, and display detection is occurring in the context of ddcutil detect --verbose, ddcutil then looks for a possible reason, and checks to see if the display is a laptop. If so, it then reports that laptops to not support DDC/CI.

ddcutil is slow.

Typically, ddcutil spends 90% of its elapsed time in waits mandated by the DDC/CI protocol.

Two options can have a big impact on performance:

  • Option --bus. If the /dev/i2c bus number of the monitor is given, ddcutil skips the initial phase of searching the /dev/i2c devices for monitors. Note that if this option is specified, no other monitor selection option such as --display should be specified.

  • Option --sleep-multiplier. This option adjusts the length of time ddcutil spends in DDC/CI mandated waits. For example, if the DDC/CI protocol specifies a 40 ms wait between the time a command is sent to the monitor and the time a reply is read, and --sleep-multiplier .2 is given, ddcutil will only wait (.2 x 40 ms) = 8 ms. Some monitors have been found to communicate successfully with --sleep-multiplier values as low as .1. On the other hand, some monitors with poor DDC/CI implementations perform better if the sleep time is increased by using a value greater than 1.

ddcutil can't communicate with a monitor when it's plugged into a docking station.

ddcutil does work when the monitor is plugged directly into the laptop. The problem affects DisplayPort, DVI, and HDMI connectors on the dock.

This is a problem with newer docking stations that implement DisplayPort Multi Stream Transport. Until kernel release 5.11, the drm driver layer did not implement I2C writes.

For the gory details, see this thread on the Intel-gfx developers list.

I'm using Nvidia's proprietary driver. ddcutil doesn't seem to be be working.

Symptoms include "DDC communication failed" on ddcutil detect, and lots of "Maximum retries exceeded" errors on command ddcutil getvcp known.

You may need the Nvidia secret handshake. See Special Nvidia Driver Settings.

The capabilities command reports the MCCS (aka VCP) version, but elsewhere ddcutil says the version is unknown.

DDC/CI has two ways to get the the MCCS version:

  • As part of the capabilities string returned in response to a DDC Capabilities Request.
  • As the response to a VCP Feature Request for feature xDF (VCP Version).

Sometimes they disagree.

Command ddcutil capabilities reports the response to the DDC Capabilities Request. However, ddcutil regards this output as purely informational. (See capabilities command)

ddcutil therefore relies on VCP feature code xDF to determine the VCP version. (For USB connected monitors, it queries HID usage x00800004.) It is possible that feature code xDF is unsupported, even though the capabilities response specifies a version.

The output of the capabilities command is incorrect.

For example, a Dell P2210 monitor has VGA, DVI, and DisplayPort inputs, but capabilities only reports:

 Feature: 60 (Input Source)
    Values (unparsed): 01 03
    Values (  parsed):
       01: VGA-1
       03: DVI-1 

The capabilities string reported by a monitor is often incorrect. While the ddcutil capabilities command parses and reports the string, this is solely informational.

The only way to know for sure if a monitor supports a VCP Feature Code is by testing using the getvcp and setvcp commands.

The following command will attempt to read all VCP codes that ddcutil understands, other than those that are write-only:

ddcutil getvcp known

To attempt to read all possible VCP codes, whether understood by ddcutil or not, except for those that are known to be write-only:

ddcutil getvcp scan

To see all the values defined for a non-continuous (NC) feature code, i.e. one with discrete values, use the vcpinfo command with the --verbose option. For example, to see all the values defined in the Monitor Control Command Specification for VCP feature x60 (Input Source):

# ddcutil vcpinfo 60 --verbose

VCP code 60: Input Source
   Selects active video source
   MCCS versions: 2.0, 2.1, 3.0, 2.2
   MCCS specification groups: Miscellaneous
   ddcutil feature subsets: 
   Attributes (v2.0): Read Write, Non-Continuous (simple)
   Attributes (v2.1): Read Write, Non-Continuous (simple)
   Attributes (v3.0): Read Write, Table (normal)
   Attributes (v2.2): Read Write, Non-Continuous (simple)
   Simple NC values:
      0x01: VGA-1
      0x02: VGA-2
      0x03: DVI-1
      0x04: DVI-2
      0x05: Composite video 1
      0x06: Composite video 2
      0x07: S-Video-1
      0x08: S-Video-2
      0x09: Tuner-1
      0x0a: Tuner-2
      0x0b: Tuner-3
      0x0c: Component video (YPrPb/YCrCb) 1
      0x0d: Component video (YPrPb/YCrCb) 2
      0x0e: Component video (YPrPb/YCrCb) 3
      0x0f: DisplayPort-1
      0x10: DisplayPort-2
      0x11: HDMI-1
      0x12: HDMI-2

In practice, any given monitor will implement only a few of the NC values, and some monitors will implement undocumented values. The only way to know for sure is by testing.

ddcutil detect --verbose reports that both I2C address x50 (EDID) and x37 (DDC) are responsive, but DDC communication fails.

There can be any number of reasons for this situation.

The same DisplayPort connected monitor appears twice in the output of ddcutil detect.

Sometimes the same DisplayPort connected monitor is detected at 2 different I2C bus numbers. ddcutil detect reads the EDID at on both buses, but typically DDC communication succeeds on only 1 bus. ddcutil attempts to filter out the invalid bus. The invalid bus appears to be related to the DisplayPort multistream facility. You can ignore the invalid entry.

The detect command reports different EDIDs depending on how the monitor is connected to the video card.

Strange as it may seem at first, a monitor can have more than one EDID. The EDID for a VGA connection will have a different video input definition (byte x14) than that for a digital connection. The VGA version of the EDID will have timing information.

The detect command reports that I2C bus address x30 (EDID block number) is inactive.

EDID version 1.4 allows for addtional 128 byte EDID blocks. I2C bus address x30 specifies which block to read. These blocks generally contain extended timing information. ddcutil is interested only in the contents of the first EDID block. The check of bus address x30 is purely informational.

setvcp does not change the feature value

Several Iilyama monitor models require that command scs (Save Current Settings) be issued immediately after a feature value is set. setvcp must be invoked with the --noverify option, followed immediately by scs.

The monitor supports Dynamic Contrast Range (DCR). setvcp for feature x10 (brightness) does not change the value

At least some monitors with Dynamic Contrast Range (DCR) disable feature x10 (brightness) and possibly other features when DCR is enabled. This has been reported with both Lenovo and Benq monitors.

The new value for feature x10 is different from the value specified on setvcp

It is common for the new value of a Continuous feature to be off by 1 from the specified value. This appears to be the result of integer to floating point to integer conversion in the monitor.

At least one monitor (Dell 2407wfp) does not allow setting brightness less than 30. setvcp values in the range 0..50 are mapped to to 30..50.

ddcutil hangs on Radeon RX 5xxx and 6xxx cards (Navi and Navi2 GPU)

There have been multiple reports of ddcutil command failures on Radeon cards with the Navi and Navi2 GPUs (RX 5xxx and RX 6xxx cards).

In some cases ddcutil hangs. (See for example issues Stop poking AMDGPU SMU I2C #194, Screen freeze when changing brightness #223).. The SMU related hang should have been fixed as of kernel 5.14.4. ddcutil has had a workaround since release 1.1.0.

In other cases, the monitor's EDID is simply not detected.

If your kernel version is 5.14.4 or later, I suggest you file a bug report at the amd driver issue list.

Include: - drm output - the kernel log file (/var/log/kernel on Ubuntu) - output of the following commands

$ sudo ps -aux | grep ddcutil
$ sudo awk ' {print FILENAME ": " $0 } '  /sys/bus/i2c/devices/*/name

KDE Plasma 6

ddcutil commands fail after upgrade to version 2.0 and KDE Plasma 6

ddcutil 2.0 added a cross-instance locking facility. The problem addressed was random failures when multiple ddcutil or libddcutil instances clobber each other's DDC transactions. Internally, ddcutil uses what are called display handles to represent /dev/i2c devices that have been opened at the OS level. Proper use of the API is, for each DDC transaction, to open a display, creating a display handle, perform the transaction, and then close it. This sounds expensive, but in fact the time spent opening and closing is negligible compared to the time spent performing the I2C operations of a DDC transaction.

When the cross-instance locking facility is enabled, and a display is opened creating display handle, ddcutil calls flock() on the underlying /dev/i2c device and holds the flock until the handle is closed. Unfortunately, at startup PowerDevil creates a display handle for each /dev/i2c device and never closes it. So the flock's are never released, blocking other ddcutil instances.

Cross-instance locking can be disabled using option --disable-cross-instance-locks. This option can be specified on the ddcutil command line, or in configuration file $HOME/.config/ddcutil/ddcutilrc. The latter is the only way to control libddcutil, which is used by PowerDevil. For a description of configuration files, see https://www.ddcutil.com/config_file.

Powerdevil crashes repeatedly with ddcutil 2.1.0

ddcutil 2.1.0 API introduced an initialization function, ddca2_init(), into the libddcutil API to give the client, e.g. PowerDevil, finer control of libddcutil's initialization. If the initialization function is not called, then default settings are used. There was a bug in default initialization that caused PowerDevil to crash repeatedly. This was fixed in ddcutil 2.1.2. Also, PowerDevil was changed to explicitly call the initialization function, thereby avoiding the bug. With two different fixes for same problem, it may not be obvious what change made PowerDevil work again.

Does ddcutil work in containers?

User Nikolay Dandanov has reported success running ddcutil in a Docker container:

I defined a new Arch Linux image for Docker using the following simple Dockerfile:

FROM archlinux:latest

RUN pacman --noconfirm -Syu && \
    pacman --noconfirm -S ddcutil && \
    rm -rf ~/.cache/*

I saved this to a file named Dockerfile and built a Docker image using ~~~docker build -t ddcutil:v0.0.1~~~.

I then ran the container and also mapped all i2c devices (otherwise, these were not available in the container):

docker run $(for dev in `ls /dev/i2c*`; do echo "--device $dev "; done) -it --name ddcutil ddcutil:v0.0.1

Questions about building ddcutil

configure complains that a required package does not exist, but it is installed on my system.

When building ddcutil, error messages from pkg-config (which is called by configure) can be misleading. If configure complains that a package is not found but it seems to be installed, it's likely that what's missing is the associated development package (with a suffix like "-dev").