First of all, congrats to Intel on shipping a technology I’d derided as vaporware for over a year.* That said, here’s why Thunderbolt will lose to USB 3.0. From Tested (emphasis mine):
Thunderbolt is designed to replace many of those cables and cuts out the middleman. Instead of connecting a USB controller to the PCI-Express bus, and then connecting your devices to that, Thunderbolt lets you plug a device directly into the PCI-Express bus … On the PC end, it’s connected to the PCI-Express bus, where it packages the native signals, sends them across the cable, and then another controller unpackages them on the far end.
So basically Thunderbolt takes you from having 1 central controller for all your devices in the USB case to having a controller for each one. This increases Thunderbolt’s OEM peripheral production costs vs. USB, which has been capable of handling up to 127 devices per controller since its 1.0 release. Controller cost matters much more to a peripheral OEM who’s selling a $200 device than to a PC OEM who’s selling a $1K+ machine.The second strike against Thunderbolt is the a feature that, ironically, should be a strength: daisychaining. For a Thunderbolt device to properly support this feature, it has to ship with 2 ports. Not only does that increase costs as above, it also causes design (aesthetics and functionality) issues and complicates manufacturing. Mind you: daisychaining can be a good thing … if you’re a user: it saves you the cost of buying a hub. But it only transfers that cost to the device OEM. And when there’s another more established connectivity standard out there that’s “fast enough” and cheaper to implement, the OEM’s going to choose it instead. Another ironic downside of daisychaining is that because it exists, the incentive for developing hubs is sharply reduced. Ever seen a Firewire hub? Neither have I. This means that peripheral OEMs either absorb the cost of an extra port or ship a 1 port device that severely limits the utility of the end user’s PC Thunderbolt port. This latter issue also arises for devices that use solely other protocols Thunderbolt is supposedly compatible with. Unless those are Firewire devices, they’re likely to have only 1 port too, so if you decide to connect your USB 2.0 external HDD to your new MacBook Pro’s single Thunderbolt port, you either won’t be able to use any additional devices on the port OR you’ll have to manually move the HDD all the way to the end of the daisychain. Another issue Thunderbolt will have to overcome is poor/buggy OEM implementation. I’ve yet to come across an external HDD whose USB functionality didn’t work correctly right out of the box. OTOH, I’ve had two such devices whose Firewire implementation was faulty as shipped. One drive shipped with only 1 Firewire port (see issue above), and proceeded to make itself invisible to Windows at random times. Windows eventually declared all the data on it corrupted, and I wound up having to switch to USB and reformat it, almost mistakenly wiping out another drive in the process. I had a similar issue with an Adaptec Firewire drive enclosure. This situation is likely to be much worse in Thunderbolt’s case because of its claimed support of other interfaces. In fact, the above support is yet another issue: not only is there the risk of incompatibility due to poor implementations, but throwing Thunderbolt into a device connection that natively supports a different interface adds yet another point of failure to the connection. It’s just something else to troubleshoot if things don’t work. It adds complexity. Did I mention how hard it is to troubleshoot a daisychain? With USB, disconnecting a device means severing only the connection between it and the host PC or it and the hub. With a daisychain, unless the device in question is at the end of the chain, you have to disconnect 2 cables and then connect the other devices that were ahead of and behind the troublesome device to each other. 3 times the effort. That may sound easy, but try doing in a tight, dark space under your desk. If you don’t wind up at least muttering expletives for a minute straight, you’re a saint. I know this because I have a Firewire HDD daisychain at home. Nice to have, hell to troubleshoot anything that’s not at the end of the chain. What does this all mean? The port and controller costs will prevent Thunderbolt from appearing on non-high end devices for a long, long time. Perhaps even forever, since USB will always be cheaper and easier to implement, while offering more compatibility and comparable performance (assuming the USB 3.0 interface isn’t a bottleneck). Of course, Intel is likely to do its best to limit USB 3.0’s compatibility by dragging its heels on native chipset support. It won’t be able to do so for long, though: USB 3.0 devices are pouring onto the market, and AMD has the tech on its chipset roadmap. Thunderbolt devices? Only 2 have been concretely announced, and neither of them are on store shelves. Don’t get me wrong, Thunderbolt is a technical marvel and a good development for people who hate cables, but simple economics, bugs and poor implementation are likely to be its demise. *Technically, since Thunderbolt showed up as a watered down copper version of the original optical concept, maybe I’m still right 😉