Even though I knew that video modes were a nightmare mess that was made barely tolerable by standards, I had no idea the hell that awaited once one passed the 1200p edge.
A short history of video modes before we begin (this helps the pain later). Video is a three-dimensional concept that must yield to the laws of computer science and become a two-dimensional bitstream of arrays in order to go down the wire to the screen (and also in order to be stored in memory, but let's not complicate things more). You may be wondering if I added an extra dimension in the previous sentence but I did not -- that additional dimension is time.
In computer science a list of numbers is an array (stop it, Melvin, you know where I'm going with this). A two-dimensional array is a list of lists (think of an outline). A picture, or a frame of video, is a 2D array of pixels (and each pixel is an array of component values -- usually RGB). You can quickly see why pictures take up a lot of space:
(0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0) (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0) (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0) (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0) (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0) (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0), (0,0,0)
That's just a 6x6 image filled with black. But it's formatted for humans. A computer would see that closer to:
That raises a very important issue: how big is that, and what's what? It's just a mess of zeroes without some kind of structure around it to describe what's what. That's metadata. Specifically, it's a glob of metadata that conforms to a pre-set standard and describes what kind of pixel description is being used, how long a row is, and how many rows make up a single image. Using this information, a computer can break up that bitstream into chunks at the row boundaries, decode the pixel colors, and then draw row after row on-screen and create an image from it.
As one might imagine, it's very important to get that description correct. If you cut at the wrong boundary or your pixels are a different format (or size!) then it all goes to pot fairly reliably. There are ways to detect when this goes haywire, but it's better if it doesn't in the first place.
That brings me to video. So take the above and then add in metadata about time. That is, how many frames will be delivered in one second, what a frame's data will start with (just in case a partial frame is delivered -- don't worry about it), and what the current stream time is for any given frame (that is, when to display it in relation to other frames should it need to hold a frame on-screen longer than another one). There's a lot of other stuff, but it starts to get rather complicated after this.
To simplify all of this, there are video modes. These are pre-agreed upon configurations of all of the above that two devices can use to simplify this whole process. You've probably heard them by the name "resolutions" but it's the same thing (Melvin, hush).
The standard-bearer resolution at one sad point in time was 640x480 for the Video Graphics Array, or VGA. Realizing this sucked rather massively, the industry moved to 800x600 and slapped "SUPER!" in front of it and we got SVGA. Both of these supported the mind-boggling limit of 16 colors due to an infinitesimal pixel storage size. Because the set of computer scientists doing video did not intersect the set of artists in the world, black and white are considered colors for the sake of this discussion. You got 14 colors, and you liked it.
Realizing this wasn't really so "super" the 1024x768 resolution was put forth and slapped with the fancy XGA name. Among the improvements was a widening to 8 and 16 bpp which allowed for 256 and 64K colors in those modes. Things got crazy after that and Wikipedia can tell you all about it should you decide to inflict that madness upon yourself.
The Point, Please Get to It.
As the frames got larger and the pixels wider and the refresh rates faster, there was an increasing demand put on the connection from the computer to the display. With the old VGA connector type it was not-trivial-but-not-impossible to keep adding these new modes to it because it was essentially broadcasting the video signal over the wire as a waveform. Enhance the transmitter and receiver and voila! -- more pixels.
Digital was quite different. After VGA (the connector) came DVI -- Digital Video Interface. It's a mess of a standard because it was bridging the analog world into a brave new digital era, but it did impressively well for what it was. However, the digital connection was limited by available bandwidth to 1920x1200 (if you've been keeping up with the names, prepare for a winner: WUXGA -- Widescreen Ultra Extended Graphics Array; when in doubt, add a word). That means that if you want to power a higher-resolution screen then you'll need two of those links to get the bandwidth needed to break past that limit. More on that later.
When DVI was made it came in several flavors, but the most common were DVI-A (analog), DVI-D (digital), and DVI-I. The I in DVI-I stands for Integrated, which is the nice way of saying "we considered all you laggards and included some analog pins in here so we can deprecate that god-awful VGA adapter and you can just hook up an adapter to extract the analog signal and power your radiation box that way". It was a really efficient naming scheme, I think.
While they were busy shoving pins into this connector, someone had the bright idea that displays might keep getting larger and more dense and it would be nice to consider the future just a little. So a collection of pins in the middle were designated a second digital "link" and could be used to shove more bits down the pipe. Using both links, you could drive about 4K pixels sixty times a second (2560x1600@60Hz). Thus we have Single-Link DVI and Dual-Link DVI.
Still Not Seeing a Point
Because displays could live on the sustenance provided by SL-DVI for many years, lots of things just didn't support DL-DVI (or DVI-A/I for that matter). That means things like displays that didn't need it, video cards that couldn't push it, cables for those folks, and KVM switches for the same reasons (oh, and cost). If you were the lucky owner of a 1920x1200 display then you were perfectly fine living in the Single-Link world. That includes all common HDTV formats, by the way, so a very large number of people are in this bucket (HDMI is compatible with SL-DVI for the most part).
But displays did, in fact, grow larger. As they grew larger, the previous version of larger trickled down to a more common position of the distribution curve and now we're seeing a video mode grow in popularity that hits this pain point right in the Y-joint: Quad XGA (aka Quad HD), and Widescreen Quad XGA. If you take 720p video and double the height and width, you'll get Quad HD (2560x1440). If you take that 16:9 signal and make it 16:10 you'll get WQXGA (2560x1600). Remember that resolution? Yeah, that's the limit of DL-DVI. We found the edge of the world (again)!
(By the way, you may have also heard it by another -- trendier -- name: 4K. That refers to the number of pixels per frame and is a nice shorthand. Another name for 1080p is 2K video, along the same line.)
The Point (Really!)
I bought one of these bastards. Namely, I bought a 32" QHD display to replace my 27" Apple Thunderbolt display because I wanted to use it with more than just my Mac (seriously, Apple?). With a DisplayPort to Mini-DisplayPort cable the Mac runs it just fine. With a DL-DVI, DisplayPort, or HDMI (1.3+) cable the other computers can use it just fine. With a SL-DVI cable -- or a DL-DVI cable running through a SL-DVI KVM -- I get 1900x1200.
It turns out that this really sucks. It sucks more that the DL-DVI KVMs for four computers (really) run about $400+, the DisplayPort KVMs appear to be limited to exactly two hosts (WTFSRSLY?), and HDMI KVMs only advertise Single-Link resolutions.
However, if you read the descriptions of the HDMI KVMs you'll notice something funny. They advertise being HDMI 1.3 switches. Yes, in spite of saying that they only handle 2K video, they are using the 4K standard and will happily support switching your lovely 4K display (or near-4K). Also, since HDMI includes full 7.1 audio, you can switch that as well.
So, very long story short: if you need to switch your 4K display for a price under 4K, use HDMI to connect it instead of DisplayPort or DL-DVI. In fact, if you handle the KM part yourself, you can get a 4K HDMI switch for about $30, and it'll do PIP and have a remote as well.
In the end, it's all about the bandwidth and the associated video mode standards. If the pipe is big enough, the pixels will flow.