Good article, and rather exciting potential, but
'Do the math'?
'Do the maths'
IBM last month claimed a breakthrough in photonics – the practice of using light pulses rather than electrons to quickly send signals in chips. Big Blue claimed its boffins had successfully designed and tested a fully integrated wavelength multiplexed silicon photonics chip which would be capable of turning out 100Gbps optical …
This post has been deleted by its author
Interesting, but why are they concentrating on Silicon, rather than moving to Gallium Arsenide (GaAs)? GaAs is faster than Silicon. Plus, GaAs is a direct band-gap material, which means that it can produce photons directly from electron transitions, which Silicon mostly doesn't.
I think the main reason is there is SO MUCH silicon technology, experience, and fab facilities it is easy to make complex chips from it, whereas GaAs has been generally kept for the fastest of products where you don't have the same density of components but need higher speed. GaAs is more radiation-hard than silicon, but also less tolerant to heat. I'm sure others with more knowledge can provide a better informed answer though.
"GaAs is faster than Silicon. "
Not it's not. Electrons move faster in GaAs. Holes (+ve charge carriers critical for CMOS) move slower.
SiC is better and IIRC so is SiGe.
But the core issue is that effectively you're needing two chips in the same package.
The real break through with Silicon Photonics is a)Making Si emit light in the first place. b) Incorporating those structures into chip mfg process straightforward enough to use in a conventional production line.
Being a bit behind the head hurty field of nanoscale circuitry... isn't one of the advantages of GaAs that the circuits can be made smaller and use less power than the equivalent silicon based circuitry.
On the other hand, the production techniques for GaAs really aren't anywhere near the level of refinenent of silicon. Possibly because it's more expensive to start off with and doesn't get any cheaper at any point.
@IBM is a strange company
seems their management want to move more and more into consultancy where there is huge profit and less expense like in research and development. Hopefully we will hear more of their R&D efforts coming to the foreground & maybe the other giants will pay attention and start balancing their income by R&D too.
Ah WebSphere! A wonderful bunch of products. Some good some excellent some totalt shite.
Kept me in gainful employ for the last 15 years (Not WAS I might add).
Most of it works pretty well. MQ is a proper message queueing system that is far better than that toy called MSMQ. Well I'm biased because that it what I do for a living.
I do agree that sometimes they (IBM that it) clearly suck with their product decisions and directions.
There's nothing new about 4 x 25G WDM 100G interfaces, we already have 100G LR4 interfaces which will do exactly what they describe in the article, but they are very expensive and consume power & space. Presumably IBM have managed to produce a much smaller and lower power interface which will help a lot. I suspect it's still on a separate chip. If one could get true 'on-chip' optical 100G WDM interconnects that really would be something.
So 140 bytes * 10.5 to get the filesize of images gives you 1.4K per image. You ain't going to get much image in there: Icons maybe; but nothing as large as -for example- an El Reg article image*.
*Totally taking the piss there***
***Why hasn't W3C come up with a sarcasm font yet? They've finished HTML5...it's not like they're busy or anything.
We've already got 25 Gbps transceivers, and have 100 Gbps networking based on four of them working together. Sure, it is notable they've shrunk it down so all four wavelengths can come from a single transceiver, rather than four transceivers, but it is still four 25 Gbps signals underneath the hood.
Surely if you can now get silicon chips that produce light this could be the ideal backpane for a laser phoshpor TV? Rear projection versions of this have been developed where a laser is scanned across the back of a screen using a chip of micro mirrors, but these screens are bulky. Having a tiny silicon laser for each sub pixel could solve that.