XGCDB - Glossary
What is SLI?

SLI or Scalable Link Interface is a multi-GPU solution developed by Nvidia used for linking two or more video cards together to produce a single output.

SLI was first introduced by 3dfx in 1998 and it was used in the Voodoo2 series graphics cards. Nvidia bought 3dfx, an acquired the technology which was called Scan-Line Interleave, at that time. In the year 2004 Nvidia introduced "new" SLI intended for use on PCI Express. Basicly SLI is an parallel processing application, that allows two or more graphics processing units to share the work load when rendering a 3D scene. Motherboard must have two PCI-Express x16 slots, for two identical graphics cards.

Two graphics cards are usually connected by SLI bridge which is used to reduce bandwidth constraints and allow two cards to communicate directly. Some low-end to mid-range graphics cards do not need bridge connector to run in SLI mode.

SLI offers one anti-aliasing method and two rendering for splitting the work between the video cards:

 

  • SLI Antialiasing Splitting. The antialiasing workload between the two graphics cards this rendering mode offers up to double antialiasing performance. This mode is intended for games which are not GPU-bound, superior image quality in place of better performance.
  • Split Frame Rendering or SFR. This rendering method analyzes and split the frame horizontally in order to divide the workload 50/50 between the two GPU-s. Depending on geometry this method will balance workload between the two GPU-s.
  • Alternate Frame Rendering or AFR. The frames are rendering in sequence, one GPU process even frames, and the other one processes odd frames, one after the other. Master card recives results of slave card via the SLI bridge, and outputs the completed frames. This would double the performance of the graphics cards, as the rendering time cuts in half.

sli
Two GPUs tehnology
In January 2006, Nvidia released the 7900 GX2, a dual-GPU card. This grapics card has a pair of slightly lower clocked 7900GTX GPU-s "bridged" together. This card was only available to OEM companies, but later version GeForce 7950 GX2 was available to consumer market.
Next Nvidia's graphics card with two GPU-s was GeForce 9800 GX2, released in March 2008.
On January 2009, Nvidia introduced GeForce GTX 295, it combines two GeForce GTX 260 GPUs.

Quad SLI
When Nvidia introduced the 9800GX2 it was with two such cards in an SLI mode. It is called Quad SLI because two GX2 where connected by SLI connectors. Quad SLI has shown improvements by enabling 32x anti-aliasing in SLI-AA mode, and support for much higher framerates at the 2560x1600 resolutions in modern games, but there where no massive improvements with common resolutions of 1280x1024 and 1600x1200.

3-Way SLI
Nvidia revealed a triple SLI setup for the nForce 700 series motherboards. This mode is only available on Vista operating system. It can be implemented on the high-end video cards such as GeForce 8800 GTX, 8800 Ultra, 9800 GTX, 9800GTX+, GTX 260 and GTX 280. In practical terms, this system delivers up to a 2.8x performance increase over a single GPU system.

Quadro Plex
The Nvidia Quadro Plex system supports up to four GPU-s per unit and it is designed for large-scale 3D visualizations.

Hybrid SLI
It is meant for use on laptop, setup consists of an IGP as well as a GPU on MXM module. When the laptop is plugged to a power socket the IGP would assist the GPU to boost performance.  When the laptop was unplugged from power socket, MXM module would be shut down in order to lower overall graphics power consumption.

Post to MySpace!