Thanks Linus.
This clip is one of the most famous moments in Linux history. It comes from the Aalto Talk with Linus Torvalds, hosted by the Aalto Center for Entrepreneurship in Otaniemi, Finland, on June 14, 2012.
What happened
During the audience Q&A, a woman described trying to get Linux working on a laptop with Nvidia Optimus switchable graphics. She had hoped Nvidia would help, but she said the company told her flat out that it would not support it. Torvalds answered that Nvidia had been one of the worst trouble spots the kernel developers had dealt with among hardware makers. Then he turned to the camera and said “So, Nvidia, f*** you” while raising his middle finger, to applause from the room.
He then added more context, as TechSpot reported: “I think it’s really sad when you sell hardware and use Linux and you are being really difficult about it. Don’t get me wrong I’m not saying that other companies are perfect either.”
Why it struck a nerve
In 2012, Nvidia shipped only a closed-source binary driver for Linux. Optimus laptops switch between Intel integrated graphics and an Nvidia GPU, and they were a constant headache. Users relied on community workarounds such as Bumblebee. The open source Nouveau driver had to reverse-engineer the hardware. Every kernel update could break the proprietary module.
What changed since
A lot, if not everything:
- In May 2022, Nvidia released open source GPU kernel modules under a dual GPL/MIT licence, starting with the R515 driver. They support Turing and newer GPUs. Older cards still need the proprietary modules.
- The source lives at NVIDIA/open-gpu-kernel-modules, and Nvidia publishes a snapshot with every driver release.
- Distributions such as Ubuntu and SUSE now package the open modules, and hybrid graphics on laptops is far less painful than it was.
The user-space parts, like CUDA, OpenGL and Vulkan, are still proprietary. Still, the kernel side is in a very different place than it was in 2012.
For sysadmins today
On a RHEL-family system such as AlmaLinux, run modinfo nvidia | grep -i license to see which kernel module you have. “Dual MIT/GPL” means the open modules. On GPU servers, pin the driver version alongside the kernel, because the kernel module and user-space libraries must come from the same release.
Related guides: Data Points on GNU/Linux Growth: What Tux Machines Traffic Says About Linux Interest | Where to get Linux?