Intel's announcement of 'milestone' breakthrough in optical interconnect technology presents an impressive vision of the future but it also raises questions about the chip giant's attitude to the industry's next for next-gen I/O here and now.
There's no two ways about it, Intel's announcement of a fully end-to-end photonic system where multi wavelength lasers are embedded directly on the communication silicon, represents an astonishing achievement which looks set to become an integral feature of the computers of tomorrow. Optical-based 50Gbit/s speeds raises the possibility of cables connecting HDTV displays and high speed storage alike, with none of the cable length issues of electrically based systems.
Intel has been working on the technology for some time, having unveiled Light Peak in 2009, and while the new 50Gbit/s technical is aimed at a future generation of the standard, the company has stated that it plans for the first generation of Light Peak to arrive in hardware this year with initial bandwidth of 10Gbit/s.
Intel's goal, with Light Peak, is to be a replacement for a wide array of interconnects such as SCSI, SATA, Firewire, USB, PCI-X and even HDMI and thus reduce the growth of many and varied interconnects that appear on motherboards now and in the future. What's less clear is how that might happen. The chipmaker has said publicly that it sees Light Peak as complimentary to USB 3.0, the 5Gbit/s next-gen USB which has already gained widespread support.
However USB 3.0 support in upcoming Intel chipsets is conspicuously absent. with major motherboard manufacturer Gigabyte reportedly claiming at Computex in June that it did not expect a USB 3.0-supporting Intel chipset to appear until 2012. Expressing a disappointment shared by others in the industry, Gigabyte pointed out that it had shipped more than a million USB 3.0-enable motherboards by incorporating a third-party controller from NEC.
Some of the confusion stems from the fact that Light Peak isn't strictly speaking a competing cable/plug standard because it also allows transportation of signals in protocols native to other standards such as SATA and HDMI. So you could theoretically have a USB 3.0 cable which merely used Light Peak to get the data optically between each end, raising the possibility of a 100-meter long USB 3.0 cable.
Yet this is clearly not all that Intel intends. In April this year Intel senior fellow Kevin Kahn said at the Intel Developer Forum in Beijing, that the company views Light Peak as "a logical future successor to USB 3.0," and that Intel would like to “build the last cable you'll ever need.”
Kahn also clarified Intel's earlier statements about the 2010 time frame as a time frame in which Light Peak would be available to hardware partners in later 2010 with actual systems implementing Light Peak expected to appear in early 2011. That's still a year before Intel has USB 3.0 support scheduled for Intel chipset integration if Gigabyte gripes are to be believed.
That kind of time frame strongly suggests that Intel is looking for some breathing space where it can talk up the benefits of Light Peak ahead of USB 3.0 becoming truly ubiquitous, assuming it even had the power to decide such a thing.
Without USB 3.0 built into chipsets, support requires components adding cost with the result that the standard has tended to appear in higher-end products so far. Still, Gigabyte's 1 million shipped USB 3.0-equipped motherboards shows a willingness to make it happen regardless of Intel's support.
If the delay is indeed part of a strategy then it seems misplaced. Peripheral manufacturers are chomping at the bit to roll out USB 3.0 products. The explosion in demand of high capacity personal storage and high definition video peripherals and consumer electronics has created a market that will already benefit from the ten-fold faster USB 3.0 speed.
Others aren't sitting on their hands with tech watcher Digitimes reporting that AMD had been in talks NEC regarding adding USB 3.0 support to the company's upcoming Hudson D1 southbridge chipset. That's USB 3.0 support potentially a full year ahead of Intel in the area where integration really matters, chipsets aimed at mobile computing products. Does Intel imagine that 2011 tablet PCs wont benefit from USB 3.0? Surely not.
Additionally the range of third-party USB 3.0 I/O chip availability is increasing with a reduction in unit pricing and hence a reduction in the cost of adding USB 3.0 support. With the world's largest motherboard manufacturers already on board with higher end products, it's inconceivable that they wont wish to rapidly roll out USB 3.0 to mainstream and low-end products alike. Everything from tablets to netbooks to desktops will have USB 3.0 in 2011. To say nothing of consumer electronics such as camcorders.
Light Peak may well be the future and may well replace all sorts of different shaped plugs and sockets on the back of the PC. Yes, it might ultimately even replace USB 3.0 too. However it's going to need to go through the adoption cycle with third-party support and cost reduction, not to mention a coherent marketing plan.
However as recently as a few weeks ago Intel revealed that the positioning and marketing of Light Peak was still under “internal discussions”, meanwhile there are ongoing statements that the company views Light Peak as complimentary to USB 3.0. That seems reasonable, from what we've heard so far, but is everyone at Intel on the same page?
It's high time Intel got the story straight and it could foster considerable good will by accepting that USB 3.0 is the I/O standard the industry needs right now. Accepting with deeds, not just words.
USB 3.0 has the performance to tide us over into 2011, giving Intel the time to consult with its partners as to how, where and when Light Peak can provide the data hungry world with a smooth transition into the light.