Silicon photonics to take over CPU Industry?

[JoshHolloway]

Hello all. I am finishing up my first year of a CE/EE major. I want to go into IC design; primarily for CPU's. I was wondering what you guys think that CPU's will use photonics extensively within the next 8-10 years.
If the average CPU is photonic based within the next ten years, shouldn't I specialize in photonics? Because if 8-10 years from now everyting uses photonic technology in computers, then by the time I get my Masters degree I would be working with this technology directly.
I don't want to do research (scientific research), I really want to work in the industry. So do you think it would be wise to consider specializing in photonics?

 

[berkeman]

None of our ASIC houses are saying anything yet about photonics replacing electronics. Are Intel and IBM making some announcements or something?

 

[JoshHolloway]

I just mean in the next ten years. It looks like intel is pursuing this pretty intensely:

http://www.intel.com/technology/silicon/sp/index.htm

 

[dlgoff]

Well. Photonics has been around since the 70s and research is just now finding photonic uses of silicon (as far as wave quides etc). You might want to find a way to get Photonics Spectra magazine to see where things are going is this area.

Regards

 

[luckycharms]

Most current photonics research is for the telecomm industry. Basically, the industry would like to go to an "all-optical" network. Currently, the light signal comes in on a fiber, hits a photodetector, and gets converted to an electrical signal and all the logic and routing operations are performed electronically. Then the signal is converted back to an optical signal by using a laser to re-emit the light pulse down the next fiber.

These optical-electronic-optical (OEO) signal conversions tend to be rate-limiting in the overall speed of the network. So, there is great interest in developing all-optical (OOO) logic and routing capability. I'm speaking of huge 10 and 40 Gbit/s routers operated by the network providers for the Internet backbone, not your Linksys home router.

Intel is doing research in this area, but I don't think it is as a replacement for your Pentium chips. Silca/Silicon systems usually run on 1300 or 1550 nm wavelengths, so the devices tend to be fundamentally limited to feature sizes on that order. Compare that to the extremely small sizes of current electronic transistors, just tens of nanometers. With photonics, you can gain speed, but you lose hugely in density.

Electronics isn't going anywhere in CPUs. I suspect that Intel is looking at photonics because a) they have the fab manufacturing resources to move easily into telecomm and b) they have some potential CPU uses for photonics maybe as interconnects or bus between multiprocessors and c) they like to keep current with anything going on with silicon research.

All that being said, photonics is a great area to get into. Just don't think of it as necessarily an electronics replacement. And it's not just Si, there are many III-V semiconductor and organic-based systems that have extremely intersting photonic applications.

 

[CrusaderSean]

I don't think you'll see photons replacing electrons in intergrated circuits. they tried this before (as in trying to route photons) and it didn't work too well. main reason is light likes to go straight. if you want to build a realistic cpu, you're going to have a lot of "corners" and light will not be guided or they're radiate a lot. another main factor is silicon based electronics is just much more cost efficient. technology is ultimately based on costs. I don't forsee photonics replacing electronics.

 

[Nothing000]

I don't forsee photonics replacing electronics.

Seriously? I think that cost will eventually come down once we master sillicon based lasers.

 

[marlon]

Hello all. I am finishing up my first year of a CE/EE major. I want to go into IC design; primarily for CPU's. I was wondering what you guys think that CPU's will use photonics extensively within the next 8-10 years.

Forget about photonics for now. The industry is currently working on replacing the polysilicon gate by a metal gate and the SiO2 dielectric by a high k dielectric. After that, which is a job that is far from done, the carbon nanotubes (CNT's) seem to be very promising (to say the least).

marlon

ps : look at the Roadmap on the INTEL website