Alternative CPU construction

[racolvin]

This is just a curiosity question on my part, I make no representation of being a materials genius 🙂

Years ago, I helped a gf with a college paper about GaAs (gallium arsenide) IC technology and its bugged me ever since. There seemed to be numerous reasons why using GaAs instead of silicon to make a CPU made perfect sense: lower power consumption, faster processing speeds, etc.

So my question to you that are WAY smarter than me about this sort of thing is: Why was this passed over? With all the talk about power consumption, subsequent heat dissapation issues, speed "ceilings", etc with todays silicon CPU's, why is the GaAs technology not a good way to extend the existing paradigm that much further, rather than going with multi-core and the like?

Not that I'm against multi-core chip mind you, I'm just curious for the discussion 🙂

R

 

[Matthias99]

So my question to you that are WAY smarter than me about this sort of thing is: Why was this passed over? With all the talk about power consumption, subsequent heat dissapation issues, speed "ceilings", etc with todays silicon CPU's, why is the GaAs technology not a good way to extend the existing paradigm that much further, rather than going with multi-core and the like?

The last numbers I heard quoted (this was maybe 5 years ago, and I have no idea where) were that chips on GaAs cost roughly 100x as much as on silicon (between higher materials costs, and most expensive/custom-made manufacturing equipment that is needed). Maybe this has changed, but until the stuff gets cheap, it won't be widely used.

 

[racolvin]

Originally posted by: Matthias99

The last numbers I heard quoted (this was maybe 5 years ago, and I have no idea where) were that chips on GaAs cost roughly 100x as much as on silicon (between higher materials costs, and most expensive/custom-made manufacturing equipment that is needed). Maybe this has changed, but until the stuff gets cheap, it won't be widely used.

Ahhh, cost. Yea, a 100x multiplier over silicon would tend to slow things down in the adoption of a different technology wouldn't it? :beer:

 

[f95toli]

The reason is that GaAs is not made of silicon.
If there is choice between silicon and another material the choice will almost always be silicon, even if the silicon processing is much more complicated.
The reason is simply that almost everying (factories, process lines,machines etc) is Si based and that commercial Si-processing is cheap.

So basically alternative materials are used only in applications were it is impossible to use Si for some fundamental reason; GaAs is used in high-speed circuits (GaAs tranistors are rather common), GaN in optoelectronics, InP is circuits which require VERY high speeds, and there are also components based on SiC and SiGe.

 

[soccerballtux]

Yes its too expensive. However we use it for SOI technology. When poored over a Germanium lattice, the Si atoms allign themselves to some extent with the lattice. This allows for greater electron flow in the Si. So we sorta get some of the benefits already.

Why not use gold and overclock?

 

[CTho9305]

I've been told that GaAs has really poor hole mobility, which means p-devices would be slow.

 

[montag451]

Aren't they using that stuff in the latest super-conductors at -270K?