"active cables are commonly used at data rates above 5Gbps. These cables contain tiny chips at either end that are calibrated to the attenuation and dispersion properties of the wire between them. Compensating for these properties "greatly improves the signal-to-noise ratio" for high-bandwidth data transmission."
Instead of the cable actually managing this itself, though, couldn't the cable just contain a single, simpler chip containing the metadata on the cable's particular make and properties, which would allow the host port (and client port) to perform variable compensation based on the information received? I suppose this would create a more complex low-level protocol, or possibly require more pins, but for a "universal connector" type of cable I imagine they're going to see a lot of wear, so I would have imagined moving hardware out of them would have been a goal...
It is probably cost. Motherboards are high-volume products, and therefore the engineers who design them are fanatically cost sensitive. Adding $5 to the bill of materials for a feature that hardly anybody needs would be a career limiting move. Conversely, the people who actually need a 10 Gbps links are exactly the people who can be up-sold to a $50 cable.
In a few years it will be everywhere natively: competition will have driven the parts cost to pennies per board, and average customers will be demanding it.
1. It is likely that the passive components are chosen to match the specifics of the cable. This would be more expensive to create with adjustable balancing.
2. This also gives them the ability to have longer haul optical cables without changing the motherboards, the optical transceivers will be inside the cable.
3. None of the ESATA ports on any computer I have ever owned has never seen a plug. All of the firewire ports are empty. Probably 2/3 of the USB ports are empty. If Thunderbolt goes the way of these, why not move the circuitry out of the empty ports?
According to http://news.ycombinator.com/item?id=2642196, industrial design trumps longevity at Apple. Maybe they like the aesthetics of having the chips in the cable (no extra pins?).
Also note that there is still a controller chip inside each TB-enabled device. The chips in the connectors mostly handle attenuation and SNR tuning (as indicated by Steko).
LightPeak/Thunderbolt is supposed to support copper or fiber cables with the same port and controller. Within the next few years, there should be much longer fiber versions of this cable, with the optical transceivers replacing these chips.
