Dr Max Grossmann

Using the Framework Laptop 13 with Debian 13 (trixie)

In 2025, I ordered and obtained a brand-new Framework Laptop 13 with AMD Ryzen™ AI 300 Series - Ryzen™ AI 7 350.

Changes to this document

2025-10-01
Post created.
2025-10-04
Added microphone configuration.
2025-10-06
Added touchpad notes.
2026-01-24
Added NVMe crash investigation and fix.
2026-02-04
Added SSD disappearance issue. Switched to 60W charger.
2026-03-16
Updated SSD disappearance issue: the problem has recurred repeatedly.
2026-10-28 (current version)
Rearranged sections so that later problems appear in chronological order. Added Wi-Fi problems with Linux 7.1 from Backports and touchpad hardening. Updated SSD disappearance issue and overall assessment after one year.

Table of contents

Assembly

Needless to say, I ordered the DIY edition. The official guide seems basically correct, though I was unable to identify a “white line” (Step 8).

Linux kernel

Linux 6.12.48 works very well. However, I took the liberty to add Backports and install Linux 6.16.3 (apt install linux-image-amd64/trixie-backports). The graphics and audio issues described below applied to both versions. I have since returned to the standard trixie kernel and recommend it; see below.

So, despite what others write on the Internet, it literally does not matter one bit. Linux 6.12.48, which comes with Debian trixie, is recent enough. Wi-Fi works perfectly well, and so does Bluetooth and everything else I can think of. I am writing these lines on kernel 6.12.48.

Always just use Debian. You'll never need to worry about anything. I recommend completely ignoring other Linux distros, especially Ubuntu and similar slop.

I also recommend using btrfs, but that too is not specific to a Framework laptop.

Screen not redrawing after video playback

I am a user of dwm and I encountered the following issue: after I had played a video, the screen essentially froze in place and refused to redraw those parts on the screen that had changed. Whenever I switched to another dwm “tag” (a workspace, if you will), the screen was redrawn, however.

In any case, that was fixed by putting amdgpu.dcdebugmask=0x12 into GRUB_CMDLINE_LINUX_DEFAULT in /etc/default/grub and running sudo update-grub. This disables PSR (Panel Self-Refresh).

By the way, screen brightness can be controlled by means of teeing to /sys/devices/pci0000*/*/*0/drm/card0/card0-eDP-1/amdgpu_bl0/brightness.

Bluetooth audio and microphone

Regrettably Debian has decided to go full pipewire. Never go full pipewire. I like pulseaudio especially because it has always worked flawlessly (as has systemd, by the way—another irrelevant controversy to the detriment of users). One of the issues with pipewire is that I simply could not get it to work with Bluetooth devices. Blueman just showed No audio endpoints registered and refused to connect with my headphones. Audio did play, but pavucontrol (another excellent piece of software) was no help in getting it all to work.

That one was resolved with systemctl --user disable --now pipewire pipewire-pulse wireplumber pipewire.socket pipewire-pulse.socket && sudo systemctl --global mask pipewire pipewire-pulse wireplumber pipewire.socket pipewire-pulse.socket and then systemctl --user enable --now pulseaudio.service pulseaudio.socket Remember to install pulseaudio-module-bluetooth, which automatically pulls in pulseaudio.

To get the builtin microphone (the “Internal Stereo Microphone”) to work, you need to use the “Play HiFi quality Music (Mic1, Mic2, Speaker)” profile of your Family 17h/19h/1ah HD Audio Controller. See Configuration in pavucontrol.

First impressions of the hardware

The build quality in general is very good. The keyboard is good, too (especially because it is very quiet). Right-clicking works through the bottom right corner of the touchpad. Compared to the keyboard, clicking in general is kind of noisy, which is a downside.

One issue that could limit my enjoyment is that it has only four slots for expansion cards, with one of the four taken up by the power supply. So far this has remained only a theoretical concern. Also, I would like to get a Mini-DisplayPort expansion card.

Crashes under heavy I/O (January 2026)

On January 20 and January 24, I encountered random crashes in the middle of the night. The screen started flickering and the device was completely inoperable. Further investigations revealed that these crashes occurred under heavy I/O, namely during my nightly btrfs scrub start / cronjob. The issue could be reproduced by running the same plus glmark2 --run-forever plus having some spiky CPU benchmarks. These crashes were presaged by audit messages in dmesg and concluded with nvme nvme0: controller is down; will reset. The fix recommended in these messages and on the internet (adding kernel parameters) was not quite sufficient, but read on.

After a tremendous amount of research, I came to a theory that is related to power supply triggering buggy NVMe power state management. At the time, I used a 30W USB-C charger at home, not the 60W charger I purchased with the laptop. Under heavy load, 30W are not sufficient (even 60W can be borderline), and the power controller will draw some current from the battery to make up for the difference. That in turn causes the battery to enter the discharging state (as may be verified with upower). This kicks in a feature called PCIe Dynamic Link Power Management. As anyone knows, such features are inherently problematic on Linux. BAZINGA!

I was able to resolve this issue by disabling that feature in the BIOS. I am for now also keeping the suggested kernel parameters (nvme_core.default_ps_max_latency_us=0 pcie_aspm=off pcie_port_pm=off plus amdgpu.dcdebugmask=0x12, see above). I have not yet tried removing these kernel parameters. Disabling PCIe Dynamic Link Power Management appears to be the key causal factor (but there may be interactions, and for now I am happy to throw these parameters at the kernel). Despite substantial effort, I have been unable to reproduce the crashes under this configuration. I will update this post should that change. Note that disabling these power management features increases power consumption.

The issue has not recurred since.

SSD disappearing during btrfs scrub (February 2026)

Separately, I once awoke to a screen saying that no operating system could be found. Once again, this happened during my nightly btrfs scrub start / run. After a reboot, everything once again looked normal.

While WD BLACK SN850 is known to suffer from “sudden death,” the SN850X that I have is not immune either. Reports of BSODs, drive disconnections, and ASPM-related crashes are not hard to find. However, all SMART data looked entirely normal. Despite tremendous effort, I was unable to figure out the core issue, or to reproduce the crash. I thus followed the ubiquitous recommendation to “reseat” the SSD (though I had already put it in very tightly and there was no discernible slack), and switched to the official 60W charger.

Unfortunately, by March 2026, the SSD disappearance had recurred again and again—roughly once every two weeks, always during the nightly btrfs scrub, and even under the appropriate 60W charger. This was frustrating.

As of October 2026, the SSD still disappears during the nightly btrfs scrub, but much less frequently: it now happens only very rarely. The issue remains unresolved.

Wi-Fi problems with Linux 7.1 from Backports (August 2026)

In June 2026, I once again installed the kernel from Backports, which soon moved from Linux 7.0 to Linux 7.1. Under Linux 7.1, the Wi-Fi card (a MediaTek MT7925, driver mt7925e) became unreliable: traffic stalled while the connection still looked fine, downloads became very slow after switching between access points, and sometimes the connection dropped or could not be established at all. At the end of August 2026, I went back to the standard trixie kernel (currently Linux 6.12.113).

Others report the same symptoms. At least two separate driver bugs are involved:

The kernel currently in Backports (7.2.11) contains the first fix but not the second. I therefore currently recommend the standard Debian trixie kernel, not the one from Backports. For completeness: I still take the firmware packages (firmware-linux, firmware-mediatek) from Backports. With this combination, I have had no Wi-Fi problems.

Touchpad hardening and tap-to-click (October 2026)

In recent weeks, the touchpad appears to have “hardened.” This is difficult to describe. I therefore recommend enabling tapping by putting the following into /etc/X11/xorg.conf.d/30-touchpad.conf:

Section "InputClass"
    Identifier "Framework Touchpad"
    MatchIsTouchpad "on"
    Driver "libinput"
    Option "Tapping" "on"
    Option "TappingButtonMap" "lrm"
EndSection

With TappingButtonMap set to lrm, a one-finger tap is a left click, a two-finger tap a right click, and a three-finger tap a middle click.

Overall assessment

I am impressed by the Framework Laptop 13. It's a high-performing laptop, aesthetically pleasing, and pretty compatible with Debian GNU/Linux.

After one year of usage, I am still happy with my Framework 13.

Benchmarks

Here is some further information about my previous laptop (a ThinkPad T440s) and my new Framework, including hardinfo2 benchmarks.

Specification ThinkPad T440s Framework Laptop 13
Processor
CPU model Intel Core i7-4600U AMD Ryzen AI 7 350
Architecture Haswell (4th Gen Intel) Zen 5
Cores / threads 2 cores / 4 threads 8 cores / 16 threads
Base / boost clock 2.1 GHz / 3.3 GHz 3.5 GHz / 5.1 GHz
Memory & storage
RAM 11.6 GB 31.4 GB
Storage type Micron SATA SSD NVMe PCIe 4.0
Storage capacity 2 TB 4 TB
Storage model Micron CT2000MX500SSD1 SanDisk WD_BLACK SN850X
Graphics
Graphics processor Intel HD Graphics 4400 AMD Radeon Graphics (RDNA 3)
Graphics architecture Gen7.5 (Haswell GT2) RDNA 3 (GFX1152)
Benchmark results
CPU Blowfish (Single-thread)
Higher is better
10.45 53.95
CPU Blowfish (Multi-thread)
Higher is better
29.37 412.57
CPU Blowfish (Multi-core)
Higher is better
19.50 325.06
FPU Raytracing (Single-thread)
Higher is better
1984.40 4584.70
GPU OpenGL Drawing
Frames per second - Higher is better
376.75 fps 3321.50 fps
Storage R/W speed
MB/s - Higher is better
1305.48 MB/s 4423.41 MB/s
General specs
Form factor 14" 13.5"
Year released 2014 2021+