We're approaching another inflection point, where it's time to cast off legacy ports for the good of all things gadgets. In the firing line: pretty much everything.
If you've had anything to do with transferring data, ever, it's likely that you've used the humble USB key. Thumbdrive. Flash drive. Just so long as you don't call it "my USB".
Still predominantly USB 2.0 devices, these things range from agonisingly slow for cheap, no-name brands, to somewhat fast, depending on how much you pay.
We say somewhat fast, because, theoretically, USB 2.0 is limited to 480MBps. Divide by eight, and that gets you 60 megabytes per second of potential throughput. Except that it never pans out as such. Thanks to overheads, non-native controllers, slow rotational speeds and otherwise, the fastest transfer speed that we've ever seen out of USB 2.0 has been 32MBps — almost half the theoretical limit.
The original USB spec was never intended for file transfer. Specced at 12Mbps for the full-speed port (with some scraping in at 1.5Mbps), it was designed for keyboards, printers, scanners and mice. Zip drives were still considered as being the ultimate in data transfer at the time. For those who don't know what they are, thank your lucky stars. The only storage hardware more flaky than Zip drives were 3.5-inch floppies, with the dreaded "click of death" a common affair.
Manufacturers saw more potential than the initial vision, though, and USB storage started to appear.
While USB 2.0 was intended to address the performance issue, its performance was underwhelming due to its CPU-bound nature, especially compared to the competing FireWire standard (with the less-catchy official name of IEEE 1394). FireWire often gave better, more consistent performance, despite only being rated for 400Mbps, leading video editors to flock to it as the interface of choice. Thanks to the power of Intel, though, USB achieved one thing that FireWire never could: ubiquity.
Neither interface could run all of the devices that we wanted them to, or as well as we wanted them to.
USB 2.0 needs to die, but it's not the only one
And so comes the need for one high-speed interface to rule them all. While the need for fast file transfers has typically been the domain of video editors and hardware enthusiasts, with ever increasing file sizes, 1080p video becoming a commodity, fascinating applications being explored and the advent of SSDs, USB is now the biggest bottleneck in the data chain.
For USB 3.0 to have the best chance of success, though, plenty of interfaces need to die. It's had its time, but FireWire, in both 400 and 800 variations, needs to go. It's of primary import to Apple-based video editors, who are no doubt invested in the aged interface. But if there's one thing we've learned about video folks, it's that they can never have enough data throughput, and they'll go with whatever interface provides the best result. Forget the promised 1600 and 3200Mbps versions, too; they don't have a chance on the consumer scale.
Likewise, as much as it's served us well, eSATA doesn't quite cut the mustard. There are a few things in its favour: it offers up to 6Gbps, and will actually get near that theoretical speed. It also has a direct link to the SATA controller, allowing it to hit higher, more consistent speeds than USB.
What it isn't is powered, making it useless for truly portable devices.
There are two competitors aiming to take on the mantle of the true universal interface: USB 3.0 and Thunderbolt.
USB 3.0, compared to its predecessor, gets a speed bump to 5Gbps. That's a potential 625MBps, for those playing at home. To date, we've only seen the interface hit 203MBps reads, but we suspect that this has to do more with non-native, older SATA 3Gbps to USB 3.0 adapters than any limitation with the device. It also carries power, charges devices quicker, has dedicated lanes for transmit and receive, rather than USB 2.0's shared connection, and is backwards compatible with USB 2.0 devices. You'd figure this would make it a no-brainer for implementation as the universal port, but there's a few things standing in its way.
The first is that not all USB 3.0 implementations are created equal. Cheap add-in cards for desktops don't help speed matters; while many have two ports on them, most only plug in to a PCI-E x1 2.0 slot, limiting the total transfer for both ports simultaneously to 5Gbps — the same speed that a single USB 3.0 port is meant to be. As long as you only use one device, you won't notice, but throw in the second, and things might get hairy. HighPoint's RocketU 1144A is an exception to the rule; by allocating a single controller per port, and basing the interface on PCI-E x4 2.0, it ensures that each port runs at its intended speed simultaneously — assuming that you can actually find the equipment to saturate the bus in the first place. This is temporary — in time, as these interim measures go away and native support comes in, the problem will disappear. Take in mind we're talking theoretical speeds here — we're still waiting for equipment to exist that hits the true speed ceiling for USB 3.0. Real speeds versus theoretical have been starkly different in the past.
Another stumbling block that serious data enthusiasts care about: just like USB 2.0, USB 3.0 is still host based — that is, every USB device that you attach will take a hit on the CPU, potentially damaging overall performance. The great swathe of mainstream consumers are unlikely to care; if you do, Thunderbolt is likely what you're hoping will save the day.
Everyone's waiting for Intel
The biggest hurdle for both interfaces is Intel. Native support for both is not planned until its Z series chipsets, coinciding with the Ivy Bridge processor launch in Q1 2012. This means that manufacturers have to use third-party chipsets from the likes of NEC to run their USB 3.0 ports, increasing the manufacturing costs of their boards, and potentially power consumption, too. It's also the reason that not all ports are USB 3.0 on modern laptops, with the third-party controllers often only supporting up to two ports.
A direct line into the PCI-E bus, Thunderbolt gives up to 10Gbps transfer speeds (potentially higher once things go optical), and supplies 10W of power. It's essentially a meshing of DisplayPort and PCI-E, a seriously impressive spec that really opens the door for the unified port: peripherals, audio, video, storage, the whole kit and kaboodle. Thunderbolt has the potential to be more universal than Universal Serial Bus. Unlike USB, though, it doesn't have thousands upon thousands of devices already on the market and completely compatible. At last count, we found four, although there are no doubt more out there, somewhere.
There's even more, we guess, if you account for the monitors that it powers as it doubles as a DisplayPort adapter. DisplayPort, though, is yet another example of a superior technology that's failed in the face of more popular alternatives; namely, HDMI and DVI. See Beta and VHS, LaserDisc and VHS and every form of high-definition audio ever for further examples.
Thunderbolt faces the same spectre, which isn't helped by its Apple-exclusive launch. There's only one Windows machine at the time of writing that sports a Thunderbolt interface, and the implementation does nothing to advance Thunderbolt's cause.
The laptop in question,, calls the interface by its codename, "Light peak", due to the fact that rather than using a mini-DisplayPort connector, it has opted to bundle the interface into a USB 3.0 port. Positives: it works with USB devices. Negatives: all existing Thunderbolt devices are now rendered incompatible.
Sony barely mentions its expanded capability on its web page, the only hint being that you can hook the laptop up to a dock that contains a graphics card and a Blu-ray drive, over "a single high-speed optical link". Hello, fragmentation.
There's one final stumbling block for Thunderbolt: the cable price. This isn't a scam like expensive HDMI cables; Thunderbolt cables are active devices — that is, they have chips at either end to ensure that the high speed remains high speed. This alone makes Thunderbolt as a universal interface quite unappealing, although professionals and video editors are rightly tempted.
The fruit company
Will Apple's involvement with Thunderbolt contribute to its success? While there's no doubt that moving its legacy FireWire constituency onto Thunderbolt would be a very good thing, we wonder how much impact the Cupertino-based company will have across the entire computing ecosystem, given its recent hostile moves against its professional video-editing user base. The substandard Final Cut Pro X update, along with Apple's slow retreat from enterprise and production, makes us wonder if professional users will be with the platform for long enough for Apple to champion the cause.
A mixed future
As much as we'd love to see one port that does everything, the cost of Thunderbolt alone makes it an unlikely contender, unless non-active, lower-bandwidth cables come into play. USB 3.0 certainly stands a better chance, but it won't appeal to the professional crowd, due to its host-based nature.
Thus, we believe Intel when it says that it sees USB 3.0 and Thunderbolt as complementary — due to its cost, Thunderbolt is for the professional, and will take over from FireWire and eSATA. USB 3.0, meanwhile, will do what it does best — everything else. It might be a two-horse race, but the horses are well and truly strapped together.