Super speed: a brief history of USB 3.0, 2007-2018

USB 3.0 is coming, and the hour approaches when the computer and electronics industries can sink their collective teeth into a new, faster USB interface for the first time in ten years. USB 2.0, with 480Mbps High speed, launched in April 2000, and USB 3.0, with 4.8Gbps Super Speed, will launch in the first consumer devices in early 2010. As this happy day draws closer, USB 3.0-related news has come fast and brisk, and it has been hard to follow. Let’s review the milestones of the past and take a look ahead to see what the future has in store for USB 3.0.

2007: Initial announcement

In September 2007, at the Intel Developer Forum, Intel’s Pat Gelsinger?announced?the forthcoming development of USB 3.0 to succeed USB 2.0. The new standard, we were told, would feature an optical fiber link to supplement the four copper wires which had sufficed for all prior USB connections, boosting speed to 5.0 Gbps. Because the four copper wires were still the same, cable length limitations would remain substantially the same, and, though this wasn’t mentioned, USB’s limited power transmission capabilities would likely follow suit.?

These developments spurred a firestorm of critical attention on Ars, with particular emphasis on the lack of any attention to solving USB 2.0’s high CPU usage and low power transfer specs, which permit a mere 2.5W (500mA of 5V DC), and to the cost and potential fragility of optical fiber. Presumably, someone with more pull than Ars forum denizens was equally upset, because the next time USB 3.0 surfaced, some changes had occurred.

2008: Revisions, xHCI infighting, and final specification

In January of 2008, the physical connectors of what would become USB 3.0 were?unveiled?at CES. The optical fiber had been replaced with copper, and a novel system developed to allow backwards compatibility.? The five new pins were situated deeper in the connector than the legacy pins, allowing the deeper new connector to connect the extra pins, while legacy plugs in new sockets, or new plugs in legacy sockets, would use only the original four.? B-style connectors have also been changed to carry more pins, in a way that will allow USB 2.0-styled plugs to fit the new ports, but not vice versa.? The importance of this should be limited, since most B-style ports are already mated with suitable cables.

The USB-IF also announced the name of the new speed mode, to complement Low Speed, Full Speed, and High Speed: Super Speed.? The name raised ridicule and heckles, although it’s hard to picture what else the standards body could have done (short of jumping to Ludicrous Speed). Super Speed is moderately dignified, clearly faster than High Speed, and consistent with prior naming. Of course, the stage for these difficulties was set with USB 2.0 in 2000 when the confusion of High Speed and Full Speed began, or even with USB 1.0 in the 90s, when named speed modes were introduced.

In June, AMD and NVIDIA?raised ire?about Intel’s development of the Extensible Host Controller Interface, the de facto standard on which all USB 3.0 host controllers are expected to be based. Intel had stepped into the void in the USB 1.0 era by developing a single controller, the WHCI, and the chipmaker released it under a royalty-free license to ensure compatibility with all devices. Intel reprised the feat with USB 2.0’s EHCI, and was moving ahead with the USB 3.0xHCI. The other chipset vendors alleged that Intel was engaging in an anticompetitive practice by first developing its own chipsets with draft versions of the xHCI, and only months later releasing the finished controller to other chipset vendors, giving Santa Clara a six- to nine-month lead on Sunnyvale and NVIDIA. Intel responded by saying that it would be unwise and irresponsible to release an unfinished specification and that the xHCI and chipset teams were separate development processes.

When Intel?released?the finished xHCI in August, the rift seemed to have been mostly healed, as NVIDIA, NEC, Dell, and Microsoft all swore fealty to the new de facto standard and pledged to base their USB 3.0 controllers, devices, and drivers on the xHCI. Mere days later, a USB host controller was?demonstrated?at IDF, where it sustained transfer speeds of 396 MBps, some 90% of the theoretical bulk transfer speed of the new interface.

In November, the final specification was?released, bringing some new refinements, but also some disappointment. The CPU overhead of the protocol would remain relatively high.? However, maximum power will increase from 2.5W to 4.5W (900mA), and power-sipping devices on USB 3.0 buses will be even more ascetic than their ancestors, taking advantage of a new idle power mode to draw less juice for standby operation. An early slew of devices is expected in 2010.

2009: Controllers and interoperability

Since then, a number of controllers have been announced.?In May of 2009, NEC announced the first standalone USB host controller, which is expected to find its way into add-in cards from many vendors.? Presumably based upon the xHCI, the ?PD720200 is a PCIe to USB 3.0 bridge capable of 4.8Gbps speeds, offered in bulk for a mere $15 per chip. NEC promised that the new controller would feature a lower CPU overhead than USB 2.0, and that it would churn out 1 million of the new chip per month by September.

Also in May of 2009, the USB Developers Conference had brisk traffic in USB demos, with a particular emphasis on interoperability.? A drive controller from Fujitsu mated with an xHCI implementation from Fresco Logic.? NEC showed off its xHCI implementation mated with a USB SATA bridge from LucidPort.? These interoperation demos showed rising confidence in the new protocol, as the pieces necessary for a launch to market begin to fall rapidly into place.

This month, LucidPort formally announced its USB SATA bridge, which would allow external hard disks and external enclosures to easily carry a USB 3.0 interface.? The new controller has been tested at up to 244 MBps using the Windows Mass Storage Controller driver, a decent chunk of theoretical performance but nowhere near the Intel IDF demonstration.? Using a custom driver dubbed USB-Attached SCSI, though, the LucidPort controller attains 336 MBps, underscoring the importance of improved drivers to the performance of USB 3.0.

Also this month, ASUS first announced and then canceled its, and the world’s, first USB 3.0-supporting motherboard, the P6X58 Premium.? ASUS is frequently the first to feature new technologies on motherboards; its premium line incorporate third party controllers for lots of features, and in their time were first or near-first to introduce onboard SATA, RAID, SAS, and other goodies.? This board, though, was cancelled before being launched, without real explanation.? It’s most likely that ASUS is waiting for a better chipset or software refinements to launch USB 3.0 products; they may well have run into trouble with an early controller chip.

2009-2010: Devices and adapters begin to arrive

With controllers shipping to manufacturers, devices should begin arriving in early 2010 or possibly even late 2009.? The first devices to drop will be host controllers sporting PCIe interfaces, of course.? Such controllers probably won’t bottleneck themselves by using PCIe x1 slots, since a PCIe 2.0 x1 interface has the same theoretical bandwidth as a USB bus, and would thus impose a performance hit.? A PCIe 1.0 x1 slot has less than half the bandwidth of a USB 3.0 bus, so it’s likely that add-in USB 3.0 cards will use the PCIe x4 interface.

After this, hard disks and mobile docks will become available. Hard disks have been available which exceed the transfer rate of USB 2.0 since at least 2002, and since 2006, all but the slowest hard disks have been able to saturate it. Even modern laptop drives can saturate USB 2.0.? USB 3.0 will allow massive increases in external drive performance, and drives will once again differentiate themselves on performance. It’s even possible that external SSDs will rise in prominence once USB 3.0 allows them to shine at their true performance potential.

The extra power may have some notable effects. It is possible but unlikely that 3.5″ bus-powered drives will emerge. This event would depend on finding a way to make a desktop drive run on 4.5W, which may or may not be possible. With virtual certainty, external DVD and BD drives will become faster as more power allows faster disc spinning and better lasers.? The available plethora of coffee warmers, joybuzzers, rocket launchers, and vibrators will also make exotic (erotic?) and innovative use of the increased power budget.

At some point in this era, device drivers and operating system support will arrive. The Linux kernel, OSS developers brag, will be the first to support USB 3.0, with the launch of kernel 2.6.31 this September. USB 2.0 had to wait for the launch of Windows XP for native support in Windows, but it’s likely that Windows 7 will receive USB 3.0 support via an update rather than at launch. It’s unlikely that Snow Leopard will support USB 3.0 at launch, but Apple is likely to introduce USB 3.0 across its lineup in 2010, and the successor OS will almost certainly support USB 3.0. With operating system support, one more roadblock to easy consumer adoption will be pushed aside.

2010-2013: Proliferation and necessity

All these devices will occupy a profitable but obscure niche until USB 3.0 begins to arrive in desktop and laptop chipsets, which won’t happen until the middle of 2010 at least.? Due to their xHCI advantage and general ahead-of-the-curve development cycle, it’s likely that Intel will be the first to ship compatible chipsets, although the lead, if it exists will be a matter of months at most. By this point, as USB 3.0-supporting chipsets will have become the norm and supplies of supporting chips for devices ramp up, penetration will ramp up dramatically.? In 2010, projections run, USB 3.0 will be a minority privilege, but by 2012 45% of all laptops shipped will feature USB 3.0, IDC projects.? In 2013, it’s likely that USB 3.0 will attain a dominant market share of 80% or so.

At this point, between 2011 and 2013, the vast majority of USB devices which require neither the speed nor the novel features of the new interface will begin to transition.? USB 3.0 support will gradually become first a selling point and then a matter of course for millions of printers, mice and keyboards, joysticks, webcams, LCD modules, bluetooth, WiFi, and mobile broadband dongles, and cheap hard disks.? By the time the transition to USB 3.0 on host systems is complete, peripherals should be in the fullest swing of their own migration.? Certain devices, like mice which benefit from the thinner cables allowed under USB 1.1, and devices targeted to the absolute lowest possible cost, won’t migrate.