A Technical Introduction to SLI

A Technical Introduction to SLI
By , Last updated on: 12/3/2014

SLI was originally only an exceptional upgrade for gaming/high-end laptops, and even then, was primarily only used in desktops. But as technology progresses at its rapid rate, SLI is being seen more and more, now even in some mid-level laptops. So, let’s talk a little bit about SLI, and hopefully help you on your quest of deciding whether or not you need an SLI configuration.

First off of all, thanks to SLI, Nvidia has not only become the commanding industry leader for graphics card, but has also broken several benchmark records. Its closest competitor, ATI, was left in the dust, so to speak, and thus had to come up with its own multi-GPU (Graphics Processing Unit). So, starting a couple years ago, ATI researched and developed its own technology: CrossFire. But, we’ll get into that in another article.

And now the moment you’re waiting for, what exactly is SLI and how does it work? SLI stands for Scalable Link Interface, and there you have it. Not enough information? Alright well let’s continue. There are two basic ways in which SLI can work, called Split Frame Rendering (SFR) and Alternate Frame Rendering (AFR). Just to go over some basics: frames are still images, and several still images are put together to make a moving, or animated picture. Thus, every TV program, movie, home video, etc, just consists of several frames put together, which appear so quickly that as the image is sent to your brain, it is processed as a moving image.

So keeping this in mind, SFR and AFR do just as their names suggest. SFR takes each still image and divides in half, or split it, horizontally. This doesn’t necessarily mean equal halves, the processing unit determines where exactly the dividing line goes, essentially equalizing and optimizing the processing of both cards. So, for example, if the bottom half contains much more detail and needs more processing, the horizontal line may be moved down to compensate. AFR, on the other hand, gives alternative frames to each card to process. AFR requires a lot of cross-communication between cards, slowing down the process. Also, since this isn’t quite as sophisticated or efficient as SFR, it is usually not used.

Going by the classic saying, bigger really is better. Well rather, more is better. Obviously, two cards working to process, regardless of the SFR or AFR method, will be significantly faster than a single card (assuming all cards are of equal capacity). In benchmark tests, normally, 2 GPUs with 256MB are comparable to a single card with 512MB. This should hold true for 2 512MB vs 1 GB, as well. So at this point there really is no advantage of two over one. But, here’s the advantage: usually, the price of 1 GPU with double the RAM can be more expensive than the corresponding 2 GPUs configured in SLI. If that didn’t make sense, 2 512MB cards are much cheaper than 1 1GB card, for example. So in the end, you’re looking at significant cost advantages. Sometimes this doesn’t hold true, but, in general, you get the picture. Now, if you’re not sure that you need to upgrade to SLI, check out our other article on whether or not to go with an SLI configuration.


Add Comment