home | about | newsletters | contact | advertising | shop | radio | courses | widget | site map

Helping Normal People Get Their Geek On And Live The Digital Lifestyle

Is It Possible To Make $141,657.15 While Playing With Your PC?
» Learn More About PCMech Premium Program
Big Things Are About To Happen Here

Login: Password: Remember me

SLI Technology

Posted Dec 11, 2005 by Force Flow  

Introduction
In the summer of 2004, nVidia released a new technology into the gaming market.  The nVidia Corporation is known primarily for developing and supplying GPUs (Graphics Processing Units), motherboard chipsets, and contributing to other forms of multimedia hardware.  The latest buzz about nVidia centers around developments made with GPU and chipset technology.  These technologies are used in joint operation to produce a gigantic leap forward in gaming hardware horsepower.  This technology is called the “Scalable Link Interface”.


Surprisingly, SLI technology is not a new concept.  The first line of SLI products were produced by a video card company named 3dfx.  SLI was first used in arcade game consoles and other professional markets.  3dfx then released a consumer version of SLI in 1998 with their Voodoo2 video card line, which was called “Scan Line Interleave” at the time.  nVidia then reintroduced the basic concept on the PCI Express bus, while ATi still continues to tweak their version of a linked video card system named “CrossFire”.  Many gamers who have always used ATi products are debating whether to bite the bullet and go for nVidia’s SLI, or play the waiting game and sit tight for the final release of ATi CrossFire.


nVidia’s new SLI (Scalable Link Interface) technology has brought a new incentive for gamers to invest in nVidia video card technology.  Broken down into its simplest concept, SLI allows systems to utilize the power of two identical video cards for a single output display.  Utilizing any more than two is merely theoretical (or hypothetical) at this point.  The purpose of SLI is to offer better performance per cost of a system as a whole.  Typically, a single, significantly more powerful video card rivals the cost of two less powerful cards.  Plus, these two cards can be purchased separately (one initially, and the second as a later upgrade), defraying the cost of an expensive gaming system.  This certainly helps builders on a lower budget and helps keep an older machine usable for a longer period of time.


With a traditional gaming computer, you would find a motherboard, a CPU, RAM, and a video card making up a system’s core components.  SLI technology makes it possible to install two video cards, and have them operate simultaneously for a boost in image rendering performance.


First, we will introduce you to what the original SLI technology was, how it was used, and a brief history leading up to the current form of SLI technology, including current and past market influences leading to its demise and rebirth.  This is followed by architecture comparisons between technologies, the performance benefits of current SLI technology, and the hardware requirements for building a computer that would support SLI. 


After taking a look at the hardware side of SLI, we will dive into the software side of SLI, which is where the real magic happens.  Discussed are SLI’s rendering and Antialiasing methods in detail, plus some comparisons between their respective settings, which are critical pieces of the SLI setup as a whole.


Wrapping up the contextual portion of the document, we end with a brief look at the innovation of gaming technology and with our analysis of the best situation where SLI can be utilized effectively, taking cost vs. performance into account.  Performance in this case would equate to how efficient the technology would maintain an acceptable output response.  In other words, it must maintain an acceptable frame rate and quality.


3dfx SLI
The first line of SLI products from 3dfx were used in arcade game consoles and other professional markets. The Voodoo 2 series was the first SLI line available for consumers. In order to make it work, two PCI Voodoo 2 cards were required be linked together on the PCI bus, and connected by a ribbon cable. Vaguely similar to the way a TVs produce frames, 3dfx’s SLI technology allowed each card to process alternating horizontal lines in a single frame in order to produce a single complete frame.





This process allowed users to increase texture filtering to make use of trilinear filtering, increase the screen resolution to a new maximum (at the time, it was an impressive 1024×768), and increase the number of polygons that could be drawn, thus doubling the rendering throughput.


A major draw toward SLI technology was that it ensured that a gaming rig would not undergo premature obsoleteness. Once a single Voodoo 2 card was unable to keep up with the gaming demands that a gamer was putting on it, all they had to do was throw in a second video card and they would be ready to fly once more. The ability to add a second video card was a huge advantage over other competitors at that time.


Unfortunately, 3dfx crashed and burned due to its lack of innovation in later product lines, and merely tried to milk old product designs for all they were worth. The company was later gobbled up by nVidia, and the SLI project was forgotten and shelved for two reasons. Advancements in leaps and bounds with more powerful video cards made SLI unnecessary, and video cards had progressed beyond the limited PCI bus and had started to take advantage of AGP. It was not possible to link AGP and PCI buses to utilize SLI, or to have more than one AGP slot on a motherboard, for that matter. Therefore, nVidia did not feel it was worth investing in SLI technology at the time. Clock speeds and driver optimizations became the main focus and driving force behind the development of video cards. On the software side of things, optimizing video card drivers became an important key for card development, and often breathed new life into older cards.



nVidia SLI

After ATi gained a secure lead in the video card market with their superior image quality and performance, while nVidia fell behind a bit (more recently with their FX series cards) and needed to take innovation in a new direction.  The 6xxx series emerged, on both AGP and PCI Express buses, which was shortly followed by the reintroduction of SLI technology, thanks to the development of the PCI Express standard by Intel, finalized in 2002.  A few months later, nVidia moved to PCI Express only, which can be clearly seen with their 7xxx series video cards.


nVidia’s SLI technology employs rendering methods that go beyond 3dfx’s interleaving method on the PCI serial bus.  SLI takes advantage of the increased bandwidth of two 16x PCI Express slots, thanks to the nature of the parallel bus architecture.  Parallel bus architecture allows more work to be done in within same amount of time than what the PCI bus could do with its older serial architecture.  Parallel bus architecture allows more data to be transferred within same amount of time because of multiple dedicated lines (a.k.a “lanes”) running back and forth between the slots and motherboard chipset.



 


However, there is one caveat.  The two PCI Express 16x slots are 16x in name only when used in a dual-card setup.  The secondary 16x PCI Express slot is a branch off of the primary 16x PCI Express slot, thus effectively sharing the bandwidth between the two slots.  When only the primary slot is in use, it receives the full bandwidth, while the secondary slot remains in a deactivated state.  When both slots are in use, each slot is throttled down and receives 8x, thanks to lane width negotiation.


SLI also takes advantage various features of nVidia hardware, such as the capabilities of the GPU (Graphics Processing Unit; the video card’s CPU), but as well as the motherboard chipset, which manages data transfers for the SLI-enabled PCI Express slots, among other tasks. Depending on how a SLI setup is configured, it’s possible to get nearly double the performance than that of a system with only a single video card.  Various Benchmarks have estimated this performance gain to be about 1.5 to 1.9 times (or, 77% to 90%) greater in a dual card SLI setup.  This is a generalized estimation from an array of benchmark reviews on SLI.  The actual performance gain is dependent on a number of factors that will vary from computer to computer: the hardware components used, the ForceWare driver version used, and the benchmark program used.

Posted In: How It Works

1 Comment(s)

  1. martin curnow said:
    4/11/2008 5:46 pm

    I am in the process of buying a new graphics gard ( eVGA e-GeForce 9600 GT KO, 512 mn, ddr3, pcie 2.0). The card spec. sheet states that to just run the graphics card the card I need a 400 watt power supply. But to run this card in a SLI mode it needs a 450 watt power supply.My question is: My computer comes with a 6150 le navida chipset. Can I run these two in a SLI configuration? And if so; do I need the 450 watt power supply?

    [Reply]

Post a Comment

Got The Newsletter?

Exclusive PCMech Content. Sign up and receive our free report: 20 Tips For Becoming a Technology Power User.

NAME:
EMAIL:

PCMech Highly Recommends...

The Hacker's Nightmare is a full 500+ pages of valuable content. It has plenty of diagrams and illustrations and is broken down into small sections with easy step-by-step procedures. This is what I like about this book. It is powerful information that everybody needs, but it doesn't read like a boring computer manual. LEARN MORE

Best of PCMech