Intel Turbo-Charging Transistors

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Intel has reached a milestone in its quest to make transistors switch ever faster while using less energy, by integrating a high-k gate with a compound semiconductor transistor. Details were presented this week at the International Electron Devices Meeting (IEDM). Intel has been researching the possibility of replacing the silicon channel of the transistor by a compound semiconductor material such as indium gallium arsenide (InGaAs).

Up until recently, such transistors used a Schottky gate with no gate dielectric, and were subjected to large gate leakage. Intel has now identified and integrated a high-k gate dielectric to reduce leakage with these so-called QWFETs (quantum well field effect transistors). The prototype device was fabricated on a silicon wafer substrate, pointing towards eventual process synergy with the existing silicon infrastructure. By using a high-k dielectric, gate leakage for short channel devices was reduced by 1000x compared with a Schottky gate, while the electrical oxide thickness was reduced by 33%, leading to higher switching speeds, which in turn leads to improved chip performance.
 
So we won't need a cooler with it's own starting caps for the fan on the next chip Intel releases? :D
 
lol, I actually know what the article is talking about since i took the device physics courses at my uni!!:p
 
By 1000x? That seems like an embellishment. Does this mean that a CPU core based on this technology would only need 0.0013V core?
 
no. It doesn't. First off the 1000x was comparing the QWFET with a Schotky gate not a normal transistor. Second gate leakage current is electricity running through a transistor when it's switched off. It affects power consumption, particularly at idle when your chip isn't drawing power to do work.
 
well leakage of 0.1 compared to 0.0001 is like 1000x

but good news for us because it will allow chip to be a bit cooler or them able to use higher clock speeds that makes it faster.
 
well leakage of 0.1 compared to 0.0001 is like 1000x

but good news for us because it will allow chip to be a bit cooler or them able to use higher clock speeds that makes it faster.

Which would just bring them back to the original temp, lol.
 
well leakage of 0.1 compared to 0.0001 is like 1000x

They don't mean a 1000x reduction in voltage, just a 1000x reduction in the waste gate leakage. it would still require a typical, if slightly less, voltage to operate the transistor.
 
Great, but how are they going to solve the fabrication problem? We'd all be using that single-atom transistor from a while back if they could get the failure rates down to the 10^-9 to 10^-12 that you need in order to actually build anything with it that's of any use outside a research lab.
 
Great, but how are they going to solve the fabrication problem? We'd all be using that single-atom transistor from a while back if they could get the failure rates down to the 10^-9 to 10^-12 that you need in order to actually build anything with it that's of any use outside a research lab.

I didn't read anything about yields in the article. Care to elaborate?
 
I didn't read anything about yields in the article. Care to elaborate?

You didn't read anything about yields, which is pretty much my point. Making one transistor isn't terribly useful these days, you need to make a great many of them at once, repeatedly, with extremely few errors. A modern CPU uses billions of transistors for each chip produced.
 
transistors based on optical technology are already there, why are they stalling?
 
Back
Top