The new Light Peak spec, now branded by Intel and Apple as Thunderbolt, is getting lots of press today. Everyone is talking about how fast it is and comparing it to 1394 and USB 3.0. Sure its faster, but everyone is missing the point. It is an attempt to unify all data cables into one, specifically by including your monitor in the daisy chain. This is great, but with some big caveats.

There are a couple of nice graphics to try to get this point across at a high level. It puts together a USB-like BUS with a DisplayPort signal, and breaks them out again at the other end. Nobody has done it before, and it is very cool. Sure there are KVM and propriety cable bundles that carry audio/USB/PS2 and DVI or Analog all at once, but these are pale precursors. Merging and splitting encrypted discrete data types on shared pins at this speed is something completely different.

Before getting to the problems it is important to note why nobody has done it before. Video needs huge amounts of bandwidth, it requires fancy hardware-based processing (at least for encrypted traffic), and all at a high clock speed.  Just getting to the high speeds while implementing the complex modern protocols has been a challenge, and tossing in other other complex protocols makes it exponentially more complex. As for the high clock speeds, video cards are expensive items that can afford the high-speed circuitry whereas the various serial ports were meant to be universally cheap. Finally, unlike the Ethernet and TCP/IP protocols which do handle different data types simultaneously, monitors and many serial devices are not tolerant to delays due to collisions or bandwidth constraints. Congrats to Intel, it is an impressive feat.

USB 2.0, the current reigning champ of serial data transfer, was released in April 2000. Almost 11 years later Thunderbolt is released, and we’ll start seeing the wave of products in 2012. If it is crowned the winner over USB 3.0 and DisplayPort, then we can logically assume it will be around for a while, say 10 years. This is great for manufacturers and standard peripherals, and bi-directional 10Gbps sure feels like an infinite  amount of bandwidth.

1080p HD only takes 3.7Gbps (1920×1080 x 60fps x 30bbp), but don’t forget that we apparently all want 2 channels for stereoscopic 3D video, so be sure to double it to 7.4Gbps. Since Thunderbolt has a total of 10Gbps upstream it seems like a perfect fit for a world of HD content. After all, 2.6Gbps upstream  and 10Gpbs downstream (which is basically untouched) is plenty for video cameras, scanners, and even external drives.

It feels like a huge amount of headroom, but when your’re talking about monitors, it really isn’t. And today in is 2011, what will I want on my desk in 2022?

Thunderbolt will hold back screen resolution.

Apple started the journey to higher density pixel count with its retinal display. For years I’ve been looking forward to a vector and scalable world where all monitors weren’t stuck at a max of about 100ppi, save a few very expensive medical imaging screens. Trending towards 250dpi for all screens would allow for a new revolution in quality devices, and the iOS/Android/WebOS wars would make it happen in amazing ways. Despite the trend of making fixed 1920×1080 resolution screens as big as you can, I had high hopes – until today.

Although the released information is still a bit vague, Thunderbolt can apparently handle 2 single-link DVI connections, where each single link DVI is a bit more than 1080p, so you can drive two 1920×1200 monitors for example. On the DisplayPort side, thunderbolt will only support up to 2 DisplayPort lanes, each of which represents 4.3Gbps, which handles two full HD streams nicely.

A true DisplayPort has a maximum effective throughput of 17.28Gbps. A 3840×2160 monitor running at 60hz and 30bpp uses roughly 15Gbps of that, which is more than the total 10Gbps supplied by the Thunderbolt spec. My 2560×1600 monitor uses 7.3Gbps which appears to fit into the Thunderbolt 10Gbps limit, but Thunderbird won’t support binding two DisplayPort lanes together. And don’t even think about 2K or larger video.

Just for comparison, a 10″ tablet at 250dpi is 8″ x 250 pixels per inch x 6″ x 250 pixels per inch x 30 bbp x 60 fps = 5.4Gbps, which is already over the Thunderbolt single screen limit. The 3840×2160 screen that uses 17.28Gbps from above is just a 17.62″ 4:3 screen at 250ppi.

Not only will Thunderbolt hold back the screen ppi revolution, but it won’t even be compatible with native resolutions for today’s high end monitors. Even though the prices have come down considerably, large format screens may remain a costly specialty item.