At the time of its inception, many considered SLI to be the next big step in GPU performance. It never quite lived up to these grand expectations, however.

Now NVLink is stepping up to do the same, but will its fate be any different? Let’s talk about the differences between SLI and NVLink, and figure out which is truly better.

NVLink is a wire-based serial multi-lane near-range communications link. As you might guess from the name, it is produced by Nvidia.

If that sounds like it’s complicated, don’t worry. The core concept is actually quite simple. Effectively, NVLink functions as a connector between two graphics cards.

So, why would anyone want to connect two of their graphics cards? Because simply put, two is better than one. By connecting two of the same graphics card through NVLink, you can achieve greater computational power than would otherwise be possible.

This is ideal when trying to achieve really high-end results with your PC. Say, for instance, you already have the most recent, powerful, and overclocked GPU on the market. There will be some tasks that it still would not be able to complete or that you might want to complete faster.

That’s where NVLink shines. Without some way to bridge GPUs together, you’re stuck with having to wait for a new card to be developed or potentially sorting out a multi-PC setup.

How Does It Differ From SLI?

If you know anything about SLI, then you might notice that the NVLink setup described sounds an awful lot like SLI. So, what’s the difference?

SLI worked under a master and slave system. This is a computing term that means one card controls the other card (or cards) it shares a connection with. Effectively, this means that one card (typically the first card connected) spends a large amount of its computational power directing and collecting the data produced by the other cards.

As a result, you’re not really getting all the power each card has to offer. Instead, you’re getting a proportion of the power each card has to offer.

Moreover, SLI bridges could only transfer so much information at once. GPUs perform a lot of complicated mathematical calculations. When working that much raw information, the two (or more) GPUs share a lot of information, and, as a result, bottlenecks are common.

NVLink works to solve both of these problems by introducing a concept used by some Wi-Fi routers known as mesh networking. In essence, mesh networking replaces SLI’s master and slave configuration with a system allowing each GPU to work completely independently from one another. Each GPU can talk with the other GPUs and the CPU directly without worrying about going through any masters.

Another side effect of this is that the NVLink connection is capable of much faster transfer speeds than SLI ever was, so bottlenecks aren’t an issue.

A more tangible example of this difference is that NVLink’s mesh networking allows your GPUs to pool their VRAM. With SLI, this was completely impossible, but with NVLink, it allows for much more complicated calculations.

As you can see, NVLink has a lot of advantages over SLI, which shouldn’t be too surprising. SLI is effectively being canceled by Nvidia going forward due to its poor performance.

But there are cases where NVLink remains useful and others where NVLink doesn’t quite measure up. What makes sense for you is something you need to find out yourself, though there are always plenty of options.