Introduction To Eyefinity
Triple screens – there’s a lot to like. The immersion these type of setups can offer is nothing short of revolutionary, but there are issues, big ones. This isn’t a fluff piece to simply promote triple screens or Eyefinity, this is an in-depth analysis of why it can be brilliant, yet rarely is.
While setups using three or more screens have been available for many years now via TripleHead2Go and even SoftTH, interest increased dramatically when in 2009, AMD announced Eyefinity – a multi-monitor solution that allows the grouping of multiple monitors into a single large surface (SLS) at the GPU driver level, thus allowing the OS to see multiple screen real estate as a single, high resolution surface. NVIDIA, caught unaware, scrambled to bring out a similar feature-set and finally got there in mid 2010 with NVIDIA Surround. While we hope to cover both NVIDIA surround and SoftTH in the future, this article is mainly Eyefinity based.
In this article, we’re going deep and attempt to cover literally every facet of triple screen gaming, so if I’ve missed anything or made an error, let me know in the comments. We’ll cover:
- How to get Eyefinity up and running and it’s many configuration options.
- Why DisplayPort is both ideal and mandatory for Eyefinity.
- Some of the common mistakes made with most configurations.
- How a perfect setup should be done.
- The software bugs and poor triple screen support of almost all games.
- My own experience with triples so far.
- Some mandatory benchmarks to show how demanding it is to run in 2013.
- What it’s like to experience and whether it’s even worth it in the first place.
The hope here is that by the end of the article, the reader will have an exhaustive understanding of everything triple screens related – both the good and bad. A secondary hope is that pressure can start to be placed on game developers to work a little harder on their triple screen implementations before I become permanently seasick (more on this later).
Let’s jump in with a little triple screen montage to show what all the fuss is about and why it’s pretty darn cool.
By the end of this article, you’ll be able to see how many of the setups in this clip were incorrectly setup, which ones were not, and reasons for why.
What You Need To Run Eyefinity in 2013
I’m going out on a limb here, but I’m going to assume you already know you need three monitors, right? Good. However, there are some important facts you must understand before you run out and buy three random monitors and expect it all to play nicely. I’ll try to cover the important and somewhat headache inducing bits below. First, we need to talk about DisplayPort. Please note, everything in this article relates to DP 1.2 only.
Any reference AMD GPU powerful enough for Eyefinity will typically have:
- 2x DVI, 1x HDMI and 2x DisplayPort v1.2.
The obvious first problem you’re likely to encounter (alternatives explained below) is having to mix and match connections to get Eyefinity running as there aren’t three identical outputs on reference cards. Next, this typical array of outputs can give the impression that Eyefinity should run fine from a 2x DVI + HDMI configuration. This is not the case. HDMI support is linked or shared with one of the DVI outputs so you are forced to choose one or the other. So, ignoring DisplayPort for a moment, it’s important to understand you have only two true independent video outputs – DVI+DVI or DVI+HDMI, and this is the reason you need something more to get triples working – DisplayPort is that ‘something’ and it’s absolutely mandatory for at least one monitor to run from a GPU DisplayPort output. So why has AMD chosen to pull DisplayPort in to complicate things?
Why not just a third DVI out or HDMI? There are very good reasons for this that I’ll try to quickly cover. Simply adding a third DVI out and calling it a day would have worked at launch, but would have had a limited time frame and eventually AMD and NVIDIA would have had to have found a better solution, and fast. DVI is on its last leg as it is and has real bandwidth limitations going forward. HDMI is not much better with bandwidth and has royalty issues ($10,000 to each high volume manufacturer plus US$0.04 per device). It would also have been expensive to implement either as each output requires independent hardware such as a Pixel Clock which synchronises the signal speed between GPU and Display (you can denote from this that DVI and HDMI can only stream one signal for one monitor, thus requiring a clock for every monitor it wishes to drive).
If there is one thing you should take away from these pages, it is that DisplayPort is a major driving force for why we have solutions such as Eyefinity today. Without it, I doubt AMD and NVIDIA would have found it feasible to bring such technologies to market. I’m confident both camps had been sitting on concepts like these for years in research labs, waiting for the right display interface to surface to make it a reality.
Let’s look deeper at the bandwidth problem:
Each separate DisplayPort 1.2 output supports around 17.26 Gbit/s of bandwidth. To put this in perspective, a standard DVI output supports only 4.95 Gbit/s, dual-link DVI doubles this, and HDMI 1.4 tops out at around 10 Gbit/s. All these numbers are meaningless unless you have something to compare them to, so here are a few useful examples to show why DP makes sense:
- 1920 x 1080 at 60hz – 4.43 Gbit/s
- 1920 x 1080 at 120hz – 8.86 Gbit/s
- 2560 x 1600 at 60hz – 8.75 Gbit.
- 2560 x 1600 at 120hz – 17.5 Gbit/s
A common desire even today would be a 120hz triple screen configuration using three 1920 x 1080 displays. This takes a whole lot of bandwidth and while three Dual-DVI or HDMI outs would just about manage this, they would be at their respective limits. In comparison, DP 1.2 can drive double the pixels before it tops out, and that’s future proofing right there, folks. Also, DisplayPort 1.2 doesn’t really have a Pixel Clock in the traditional sense, it works more like USB or PCI-e in that it uses a packetized data interface that can handle multiple display streams at the same time. The streams are sent and then ‘clocked’ at the displays DP receiver. This means DisplayPort can support multiple displays that have different video resolutions (and refresh rates), and hence different pixels clocks, over the same output. The limit to what can be sent is again down to DisplayPorts current bandwidth capacity of 17.28 Gbit/s – looking at the above example i listed, you can see that running three 1920 x 1080 @ 60hz is possible over a single connection.More Connection Options