I started using Bluefin and Fedora Silverblue at the end of 2023. The concept of primarily using cloud technology and containers matched perfectly with my workflow.
I gave a few talks about it:
Everything I talked about holds true and works as expected. I even built my own Tuxedo drivers for my images:
The Tuxedo pipeline stopped working after I switched to the Fedora 43 version (regardless of whether I used Silverblue or Bluefin). It failed because the build user ran into errors while compiling the kernel module.
Having a stable image that only updates on a successful build was perfect and rock solid, until my notebook started experiencing sudden reboots, freezes, and kernel dumps. At first I thought an old issue had resurfaced, where my NVMe drive needed a firmware update.
Tool & Software Download | Samsung Semiconductor Global:
- NVMe SSD-990 PRO Series Firmware *(8B2QJXD7) To improve read-operation stability. (Release: December 2025) *(7B2QJXD7) To address the intermittent non-recognition and blue screen issue. (Release: September 2025)
But I found threads across multiple Linux distributions describing the same problem and offering different workarounds:
- https://www.reddit.com/r/linux_gaming/comments/1q1bg71/8_threads_in_2_weeks_amd_gpus_crashing_on/?solution=cfc29e7b2777aea4cfc29e7b2777aea4&js_challenge=1&token=bbbe4bf1c9a2b5160829c4be34da5861b59ff771b8920b4e2840b1bd3abf3abe&jsc_orig_r=
- https://bugzilla.redhat.com/show_bug.cgi?id=2420039
I tried several approaches:
- https://github.com/tuna-os/tunaOS
- https://docs.fedoraproject.org/en-US/quick-docs/kernel-installing-from-koji/
I rebased to Bluefin LTS (which the documentation now notes as unsupported) and could no longer boot. I tried TunaOS (LUKS wasn’t working with the ISO I used), various kernel boot parameters, and downgrading amd-gpu-firmware to versions from early, mid, and late 2025. Eventually I went back to Kernel 6.17.8, after one post mentioned the issue didn’t exist there and had appeared in later versions.
Kernel 6.17.8 was significantly more stable, but when the latest firmware package 22060401 arrived, the crashes started again.
I’ve been dealing with this issue for five months. Some days I hit five reboots or freezes; on others just one. Having a stable system for more than a day was rare.
I’m not sure whether building kernel modules or akmods would help, but my knowledge in that area is limited. Too often the image-building pipeline worked fine for weeks and then suddenly failed. I never managed to stick with the stock images from ublue-os; I always had to build my own to include the settings and applications I need in the core system. Updating once or twice a week is fine, but each update download is at least one gigabyte.
So yes, immutable desktops are stable, but the download sizes are large, and adding special drivers or kernel modules is complicated, if not impossible. A good example is this Bluefin discussion, where a docking station driver was ruled out due to the inherent limitations of the immutable model. If you depend on hardware that needs out-of-tree drivers, you’re largely on your own, and that’s exactly the situation I found myself in. That, combined with the AMD GPU instability, is what pushed me back to a conventional distribution. I had no chance to use a pinned image, as I needed some time to recognize that the issue is not hardware related.
To be clear: I’m not blaming Bluefin or Fedora. Both are excellent distributions, and the immutable desktop concept is genuinely compelling. The combination of my specific hardware issues and driver requirements just pushed me beyond what the model handles well today.
Immutable desktops have been great to use, and I’ll keep Bluefin on my other notebook (which has no AMD GPU). But my primary work machine is now running CachyOS.
