Graphics processing unit

From Citizendium, the Citizens' Compendium
Jump to: navigation, search
This article is developing and not approved.
Main Article
Talk
Related Articles  [?]
Bibliography  [?]
External Links  [?]
Citable Version  [?]
 
This editable Main Article is under development and not meant to be cited; by editing it you can help to improve it towards a future approved, citable version. These unapproved articles are subject to a disclaimer.

A graphics processing unit (GPU) in a computer is an independent specialized graphics coprocessor. It is responsible for handling the graphics output of the computer, generally to the monitor. GPUs usually handle one or two outputs, often in forms like VGA or DVI.[1] In some computers, GPUs can also handle other functions, such as video decoding. A graphics processing unit is a part of all personal computers, most workstations, and some servers. A graphics unit can be dedicated or integrated. Generally, lower-end PCs or smaller PCs tend to have integrated graphics solutions, while midrange or high-end PCs have dedicated graphics solutions on video cards.

Computers use a graphics API and driver to interact with the GPU. Drivers are generally supplied by the motherboard or GPU manufacturer, and often only work for a specific model or set of models of GPU. A computer program (such as a game) will generally use either OpenGL or DirectX to communicate with a graphics card.[2]

Dedicated Graphics Card

A dedicated graphics unit is normally situated on a video card, and connected to the motherboard by an expansion slot, such as PCIe or AGP, and includes dedicated RAM for the graphics coprocessor to use. Dedicated RAM offers massive amounts memory bandwidth to the GPU. Additionally, because there is less graphics RAM than system RAM, the graphics RAM can be expensive low-latency, high speed RAM.

Gaming PCs sometimes include more than one graphics card, using either SLI or Crossfire, which allow pairs of nVidia or ATi cards to work in parallel on graphics tasks.

List of Dedicated GPUs

All GPUs listed below use a PCI 2.0 16-pin motherboard connector unless otherwise specified. This is not an exhaustive list.[3]

The GeForce 8400 GS is a relatively low-end GPU with a core clock of 567 MHz, a shader clock of 1.4 GHz, and a memory clock of 400 MHz. It must have at least 512 MB of random access memory (GRAM). It supports a 64-bit memory interface and can transfer up to 6.4 GB/s of video data. It can fill textures at 3.6 billion per second. It supports DirectX 10 and below.

The GeForce 210 is another relatively low-end GPU with a core clock of 589 MHz, a shader clock of 1.4 GHz, and a memory clock of 500 MHz. It must have at least 512 MB of random access memory (GRAM). It supports a 64-bit memory interface and can transfer up to 8 GB/s of video data. It supports DirectX 10 and below.

The GeForce 9500 GT is a decent-spec gaming GPU with a core clock of 550 MHz, a shader clock of 1.4 GHz, and a memory clock of 800 MHz for DDR3 and 500 MHz for DDR2. It must have at least 256 MB of random access memory (GRAM). It supports a 128-bit memory interface and can transfer up to 25.6 GB/s of video data with DDR3 memory, or 16 GB/s with DDR2. It can fill textures at 8.8 billion per second. It supports DirectX 10 and below.

The GeForce 9800 GT is a relatively good gaming GPU (and direct descendant of the 9500 GT) with a core clock of 600 MHz, a shader clock of 1.5 GHz, and a memory clock of 900 MHz. It must have at least 512 MB of random access memory (GRAM). It supports a 512-bit memory interface and can transfer up to 57.6 GB/s of video data. It can fill textures at 33.6 billion per second. It supports DirectX 10 and below.

The GeForce 220 GT is a decent gamer's GPU with a core clock of 625 MHz, a shader clock of 1.35 GHz, and a memory clock of 970 MHz. It must have at least 512 MB of random access memory (GRAM). It supports a 128-bit memory interface and can transfer up to 25.3 GB/s of video data. It supports DirectX 10.1 and below.

The GeForce 240 GT is a decent gamer's GPU with a core clock of 550 MHz, a shader clock of 1.34 GHz, and a memory clock of 900 MHz for DDR3, 1GHz GDDR3, or 1.7GHz DDR5. It must have at least 512 MB of random access memory (GRAM). It supports a 128-bit memory interface and can transfer up to 54.4 GB/s of video data. It supports DirectX 10.1 and below.

The GeForce 250 GT is a mid- to high-level gamer's GPU with a core clock of 738 MHz, a shader clock of 1.83 GHz, and a memory clock of 1.1 GHz. It must have at least 512 MB of random access memory (GRAM). It supports a 256-bit memory interface and can transfer up to 70.4 GB/s of video data. It supports DirectX 10 and below.

Integrated Solution

An integrated graphics solution sits directly on the motherboard. It borrows RAM from the rest of the system to do its work. Integrated solutions are generally slower, but use less energy and are less expensive. Integrated solutions are slower both because they borrow lower-bandwidth system memory and because they tend to have slower processors. Some newer integrated GPUs are hybrids, using some dedicated memory and borrowing some system memory. This allows for better performance. Integrated solutions by definition are generally not expandable or replaceable, but some motherboards which use integrated graphics also have PCIe slots that allow for a graphics card to be added.

References