Nvidia's RTX 5090 and RTX Pro 6000 cause virtualization bug
The GeForce RTX 5090 and RTX Pro 6000 graphics cards from Nvidia show a bug in virtual machines that requires the host system to be restarted.
(Image: Konstantin Savusia/Shutterstock.com)
The GPUs GeForce RTX 5090 and RTX Pro 6000 from Nvidia's current Blackwell generation apparently lead to a bug in virtualizations that even requires a restart of the host system so that they are correctly recognized by the system again. Specifically, the GPUs hang when the virtual machines are reset.
Instead of returning to an error-free state, the GPU does not respond. The kernel returns the following error message: "not ready 65535ms after FLR; giving up". A Function-Level Reset (FLR) resets devices connected via PCI-Express, such as a graphics card. This is normally a standard procedure when passthrough devices are reassigned. It is noticeable that the card also becomes unreadable for lspci: "unknown header type 7f".
Videos by heise
Who discovered the bug?
The bug was discovered by the GPU cloud provider CloudRift, which was able to reproduce it on several Blackwell systems. CloudRift is now offering a reward of up to 1000 US dollars for information leading to the cause or solution of the problem.
Private users of RTX 5090 graphics cards have also observed the bug. The entries in the Proxmox subreddit and the Level1Techs community are multiplying. Interestingly, there are no known bugs with graphics cards from the previous generation, such as the GeForce RTX 4090. This suggests that this bug is limited to the Blackwell family. Nvidia itself has not yet commented on the error.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(mma)
