Intel is going to announce "a new technology" tomorrow, and the online buzz reckons it will release 'Light Peak', its alternative to USB. Certainly CNet thinks so, citing the usual "unnamed source familiar with the event", who says Intel will indeed put the spotlight on Light Peak tomorrow. And the interweb has been buzzing …
10Gbps is a major flaw
I work at a shop where we develop 3Gbps video interconnect networks hardware. While we have the same bugs you'd find in another other equipment out there (software bugs, manufacturing flaws etc...) our #1 money losing bug is the 3Gbps connection itself.
To achieve 3Gbps you need to convince a signal on a wire to turn on and off 3 billion times a second and make it so that the device receiving the signal is able to synchronize to that signal and decode it. To use the term "the signal travels over wires..." is very accurate in an odd sense. The wires used to carry this signal don't specifically hold the signal in place, but behave more like the tracks of a mag-lev train. The signal flows along the wires which provide a guideline in a sense that the signals should follows. It's not like you're 100% guaranteed that the signal will just happily stay wrapped up in the wires themselves.
In short, if the wire turns around a radius too tightly, the signal will behave like a race car going too fast around a track and fly right off the side into the crowds.
To design hardware able to carry these signals, the engineers have to consider countless variables covering everything from the wires themselves to shape of the leads on the connectors. 3Gbps is a nightmare. I mean a true disaster. Since we originally learned how to exceed basic reflection related issues on lines, getting up past 1.5Gbps while still using parts large enough to place on the circuit board without a microscope has been a true disaster.
Along comes 10Gbps circuits. I mean, these are just not even funny. Companies like Realtek (possibly the worlds biggest Ethernet chip maker) hasn't touched this one yet and probably with good cause. It's just scary. Companies like Intel who can afford to replace 100,000 cards with bugs if they have to are much different than other companies who don't have the means. To suggest 10Gbps serial connections is rocket science would be doing the engineers who actually make 10Gbps work an injustice. It's actually harder.
It'll be interesting to see how we move forward today. My guess is, USB 3 will suffer many of the same problems that USB 1 and 2 did in the beginning where packet loss and device connectivity failures were the norm. Lightpeak is going to be more like Firewire, you know that Apple engineered standard that never really quite worked right. It was good enough to keep a few users happy, but the protocol and overall design was too damn complex for most engineers to actually implement correctly.
Well, as an early adopter, I'm looking forward to trying it for local area networking (if it's possible since my home server has a sustained disk read speed of 2 gigabytes per second), but I'm doubting that it will work right for a few years to come.
P.S. - Keep in mind that 10Gbps is 4 times as fast as current PCIe implementations, so DO NOT purchase and add-on card that is only PCIe 1x for lightpeak or USB3 unless you're happy with 1/4 or 1/2 the performance respectively.
Horses for courses
Light peak sounds more like next generation FireWire than USB. USB 3.0 is a mess for manufacturers as backwards compatibility means both a USB 2 and USB connection and children with sufficiently nimble fingers for the cabling are increasingly rare. Then, of course, there is the question as to whether USB is really suited for high-speed data as well whilst being able to deal with a plethora of devices with 1 or 2 kbps required by keyboards and mice. For the truly *serial* connections Bluetooth with some kind of NFC-based pairing is probably more attractive.
Goodbye USB, we won't really miss you.
Other than video kit it's getting to the point where all of this bandwidth is unusued anyway. Sure you can buy kit today with USB3 sockets, but most of them are bolted on to a USB2 controller at one or both ends of the wire. All a waste of time, and a mostly unecessary means to try and pull "look shiny" over the eyes of the consumer and charge more for the same thing.
Don't get me wrong - the need for this will come with time, but it isn't now. And besides, I don't want to be tied into yet another hardare standard which will inevitably get tied to only one or two chipset providers; that's a route to lockin, getting tied to intel-only mobo's, etc.
USB and more widely adopted standards all the way please.
- Product round-up Six of the best gaming keyboard and mouse combos
- China building SUPERSONIC SUBMARINE that travels in a BUBBLE
- Boffins attempt to prove the UNIVERSE IS JUST A HOLOGRAM
- Review Raspberry Pi B+: PHWOAR, get a load of those pins
- Linux turns 23 and Linus Torvalds celebrates as only he can