NVIDIA

From Gentoo Wiki
Jump to:navigation Jump to:search
Other languages:
Resources

NVIDIA is a popular graphical chipset manufacturer. NVIDIA GPUs can use either open source (nouveau, x11-drivers/xf86-video-nouveau) or proprietary (closed source) drivers (nvidia-drivers, x11-drivers/nvidia-drivers).

Which driver to choose?

Because it is more up-to-date and efficient, the proprietary driver is often seen as a must-have for more computationally intensive 3D applications and games. At the same time, NVIDIA is gradually ending support for older GPUs. A proprietary driver is available on a limited number of architectures, currently only for amd64 and arm64. These limitations may lead to the fact that the only alternative can only be the open source nouveau driver.

Feature support

Chipset Code name Series OpenGL OpenCL VAAPI / VDPAU Vulkan CUDA[1] Latest nvidia-drivers support[2] nouveau support[3]
NV04 Fahrenheit Riva TNT, Riva TNT2 1.2 No No No No 71.86.xx Yes
NV10 Celsius GeForce 256, GeForce 2, GeForce 4 MX 1.2 No No No No 71.86.xx Yes
NV20 Kelvin GeForce 3, GeForce 4 Ti 1.3 No No No No 96.43.xx Yes
NV30 Rankine GeForce 5 / GeForce FX 1.5 No No No No 173.14.xx Yes
NV40 Curie GeForce 6, GeForce 7 2.1 No No No No 304.xx Yes
NV50 Tesla GeForce 8, GeForce 9, GeForce 100, GeForce 200, GeForce 300 3.3 No Yes No 1.1 340.xx Yes
NVC0 Fermi GeForce 400, GeForce 500 4.6 1.1 / 1.2 Yes No 2.1 390.xx Yes
NVE0 Kepler GeForce 600, GeForce 700, GeForce GTX Titan 4.6 1.1 / 1.2 Yes 1.2 / 1.3 3.0 470.xx Yes
NV110 Maxwell GeForce 700, GeForce 900 4.6 3.0 Yes 1.3 5.2 latest Partial
NV130 Pascal GeForce 10 series 4.6 3.0 Yes 1.3 6.1 latest Partial
NV140 Volta Titan V 4.6 3.0 Yes 1.3 7.0 latest No
NV160 Turing GeForce 16 series, GeForce 20 series 4.6 3.0 Yes 1.3 7.5 latest git-sources
NV170 Ampere GeForce 30 series 4.6 3.0 Yes 1.3 8.7 latest git-sources
AD10x Ada Lovelace GeForce 40 series 4.6 3.0 Yes 1.3 8.9 latest git-sources

A full list of NVIDIA GPU capabilities is here.

Hybrid GPUs systems

Some systems allows seamlessly switches between two GPUs. It is typically used on systems that have an integrated Intel GPU and a discrete NVIDIA GPU. The main benefit of using that technology is to extend battery life by providing maximum GPU performance only when needed.

Proprietary implementation from NVIDIA called Optimus and it could be used only with nvidia-drivers. For an open source implementation of Optimus that can be used on systems with nouveau driver, see Bumblebee.

Available software and articles

References