Nvidia FX5200 GPU With 256 MiB RAM
|Founded||April 5, 1993|
Santa Clara, California,
|Revenue||US$10.92 billion (2020)|
|US$2.85 billion (2020)|
|US$2.8 billion (2020)|
|Asset||US$17.32 billion (2020)|
|Total equity||US$12.2 billion (2020)|
Number of employees
|18,100 (October 2020)|
|Subsidiaries||Nvidia Advanced Rendering Center|
After proposed acquisition: Arm Ltd.
Nvidia is the by far biggest marker of graphics card chips for x86-64 computers. Their chips are found in a wide range of desktop and laptop computers. Using Nvidia graphics hardware on GNU/Linux machines requires a proprietary binary blob graphics driver because Nvidia is, as a company, very hostile towards free software and free software users. Linux kernel community describes them as one of the most difficult companies they have to work with.
There is a free driver for Nvidia cards available called nouveau which can be used - but it's horribly slow on modern cards. The reason is not due to lack of effort from the free software community, it is actually due to various choices made by Nvidia. Their newer GPUs use signed firmware which is closed source. The practical result is that free drivers can not clock or power-manage Nvidia GPUs which results in horrible performance. You can watch the nearly 50 minutes long video Nouveau Status update from Fosdem 2020 for additional details.
Linus Torvalds described Nvidia as "The single worst company we've dealt with" at a Q&A at Aalto university on June 14th, 2012.
Non-Free driver is required for acceptable performance
Nvidia offers a proprietary binary blob driver and prefers that you use this over the free driver. Some distributions, like Manjaro Linux, include this driver. Others, like Debian and Fedora, do not make installation of this non-free driver easy.
The driver does have some problems, it tends to be specific to kernel and X server version. This means that Nvidia's driver will to some degree dictate what core components your system can use.
Unlocking Firmware VGPU Restrictions
There is a project called
vgpu_unlock at https://github.com/DualCoder/vgpu_unlock that:
"enables the use of Geforce and Quadro GPUs with the NVIDIA vGPU software. NVIDIA vGPU normally only supports a few Tesla GPUs but since some Geforce and Quadro GPUs share the same physical chip as the Tesla this is only a software limitation for those GPUs. This tool aims to remove this limitation."
We do not have any Nvidia hardware we can use to test it, so all we can say about this project is that it looks interesting.
- nouveau driver feature-matrix: https://nouveau.freedesktop.org/wiki/FeatureMatrix/
- nouveau manual: https://man.linuxreviews.org/man4/nouveau.4.html
- NVIDIA Annual Reports 2020. Nvidia (December 2020).