3D accelerator

from Wikipedia, the free encyclopedia

A 3D accelerator is an extension of the graphics card in a personal computer that specializes in the calculation and display of three-dimensional objects. This extension can calculate triangle-based algorithms (such as Z-buffers , texture mapping ) and anti-aliasing by means of hardware acceleration much faster. Every PC sold today has at least a rudimentary 3D accelerator that is integrated into the graphics card. Replacing it with a faster graphics card is quite easy, the 3D performance can be increased more than tenfold. 3D accelerators are also used in game consoles such as the Xbox 360 , Wii and PlayStation 3 . Cell phones are increasingly being equipped with corresponding accelerators in order to offer more appealing operation and to be playable.

history

Diamond FireGL 1

The first 3D accelerators for the private market appeared in the mid-1990s. Originally it was plug-in cards that were additionally installed (e.g. 3dfx Voodoo Graphics ). The signal from the old graphics card was looped through this expansion card and was only interrupted when 3D content was displayed. The old graphics card was still responsible for displaying 2D content (DOS console, Windows interface).

However, special 3D cards have been produced for the professional market since the 1980s (e.g. IBM PGC , TIGA or Diamond FireGL ). These cards were mainly used in mechanical engineering, construction ( CAAD ) and circuit design in electrical engineering in the form of CAD / CAM programs application. The initially proprietary software interfaces were standardized with the OpenGL initiative. In their area of ​​application, these special cards were far superior to the 3D accelerators from the consumer market in terms of speed, while at the same time they were unsuitable for computer games.

At the end of the 1990s, the 2D and 3D functions were combined on one graphics card (e.g. 3dfx Voodoo Banshee ). In the further development, on the one hand, the pure drawing performance of triangles was further increased through massive parallelization ( filling rate in megatexels / s). At the same time, the accelerators were expanded to include additional capabilities. This included the integration of the T&L unit into the graphics card, (programmable) pixel shaders and physics engines . Because of their specialization in graphics calculations today are GPUs to CPUs superior in processing power. The number of transistors in the current (June 2012) graphics processor from AMD (Radeon HD 7970, approx. 4.3 billion) and the current top model from Intel ( Ivy Bridge-HE-4 , approx. 1.4 billion, of which approx 405 million for the internal graphics part). The development of the integration density of graphics processors has even exceeded Moore's law with an annual factor of 2.4 .

The connection of the 3D accelerator to the overall system made special demands on its construction, since very large amounts of data have to be transferred. The bandwidth of the ISA bus was nowhere near sufficient. This necessity led to expansions such as the EISA bus or VESA local bus . The successor standard PCI soon no longer met the increased requirements. A new standard for graphics cards was introduced with the AGP slot. Only with PCI Express is a general slot for 3D accelerators common again.

Another function that graphics cards can be expanded to is the decoding of video data (e.g. H.264 ). The large computing capacity of the 3D processor can also be used for this.

Accelerated calculations

Individual evidence

  1. ↑ Page no longer available , search in web archives: Qualcomm presents 3D chip for mobile phones - report on pcwelt.de from 23 August 2007@1@ 2Template: Toter Link / www.pcwelt.de
  2. ↑ Manfred Bertuch: bread and games. Focus on 3D graphics hardware. c't, May 1996, accessed March 27, 2015 .
  3. ↑ C't magazine: The TIGA is awakened, The graphic interface TIGA and its competitors , issue 11/1990, page 220.
  4. ↑ Comparison of Diamond FireGL with Riva TNT-2 in C't-Magazin 07/1999