Tagged: cuda, geforce, gpgpu, gpu, graphics, nForce, Nouveau, nvidia, Optimus, quadro, technology, tegra, tesla
This topic was published by DevynCJohnson and viewed 1541 times since "". The last page revision was "".
- Topics - 437
Nvidia is a GPU manufacturing company. The corporation also makes system-on-a-chip (SOC) devices. Nvidia devices and drivers are commonly mentioned among Linux users. Nvidia makes a few different devices that Linux supports. Unfortunately, the drivers are closed-source, but they are still commonly used among Linux systems. Therefore, it may be beneficial to learn about some of the devices.
GeForce (http://www.geforce.com/) is a graphics processing unit (GPU) that was first made August 31, 1999. AMD's Radeon GPU is the only note-worthy competitor to Nvidia's GeForce GPU. GeForce has many features as mentioned below.
- 3D Vision - stereoscopic 3D images
- 4K - four times as many pixels on 1920x1080 screens
- Adaptive V-Sync - eliminates frame tearing and stuttering
- Battery Boost - increase battery life
- CUDA - discussed in detail later in this article
- DSR - (Dynamic Super Resolution) provides more detailed resolution
- FCAT - (Frame Capture Analysis Tool) maximizes performance by finding and fixing problems
- G-SYNC - synchronize the display's refresh rates with the GPU activity
- HairWorks - better support and rendering for hair-like objects
- HBAO+ - improve ambient occlusion rendering
- MFAA - (Multi-Frame Anti-Aliasing) provides better anti-aliasing
- Optimus - discussed in detail later in this article
- PhysX - physics simulations are processed by the GPU
- SLI - the ability for multiple GPUs to work together
- Surround - multiple displays acting as one
- TXAA - a specialized anti-aliasing algorithm
- VXGI - (Voxel Global Illumination) realistic reflections, shading, and lighting
nForce is a motherboard that is compatible with AMD and Intel microprocessors. These are high-performance motherboards that can support large amounts of RAM (up to 8GB in the nForce 700 series). Most of these motherboards can be overclocked.
Tegra is a system-on-a-chip (SOC). The chip is used for mobile devices and uses an ARM CPU. Like other Nvidia products, these chips are high-performance and optimized for gaming.
The Quadro is a GPU similar to GeForce. However, GeForce is designed for gaming while Quadro is intended for general computers. Nvidia's Quadro is equivalent to AMD's FirePro.
Tesla GPUs (named after Nikola Tesla) are designed for intense computing (like supercomputers). Many of the sales for this GPU are military-related.
FUN FACT: The Tesla K80 has 24GB of RAM.
An open-source alternative to Nvidia's proprietary drivers exists. The alternative is called "Nouveau" (http://nouveau.freedesktop.org/) and is licensed under the MIT license. Nouveau supports Nvidia's video cards and the Tegra SOCs for x86, x86-64, and ARM architectures on both Linux and BSD systems. Nouveau is a Gallium3D-style driver. Gallium3D is a library and API used to simplify the process of programming graphics drivers. Nvidia's proprietary drivers perform better than Nouveau.
The Direct Rendering Manager (DRM) is the part of the Linux kernel that communicates with GPUs. Userland programs send data to the GPU by using the DRM's API. The Kernel Mode Setting (KMS) driver is a kernel driver that manages the display controller. Graphic card drivers provide a DRM and KMS (such as Nouveau).
NOTE: Each GPU is represented in /dev/ as "/dev/dri/card#" where "#" is a number.
Compute Unified Device Architecture (CUDA) is a programming model for parallel computing. Nvidia uses this platform in their graphics processing units (GPUs). CUDA allows GPUs to also act like CPUs. This means the graphics card can process non-graphics-related data. A GPU that can do this is known as a General-Purpose computing on Graphics Processing Units (GPGPU). Linux, OS X, and Windows supports CUDA. Many of Nvidia's products uses CUDA.
Nvidia Optimus (or just "Optimus") is a power and performance optimization ability used to help reduce power-consumption in laptops and mobile devices. A GPU with support for Optimus will reduce or shut-off power to the GPU when not in use. Optimus provides a way to make the GPU perform well while still conserving power. Bumblebee (http://www.bumblebee-project.org/) is a term used to refer to an open-source implementation of Optimus for Linux.