Sunday, May 15, 2011

Why not above 4 Ghz??????? TERAHERTZ still A DrEaM???

Over the years scientists have been trying hard on semiconductors - (microprocessors and memory) to increase the performance and speed simultaneously.The number of transistors in a single chip has gone up to 731 million in recent i7 processors; all the way from 2300 in Intel's 4004. This has been achieved only  because of the reduction of the size of these transistors which increases the problems faced by the scientists.

When one transistor works in conjunction with a million others, the chip can deal with huge amount of data at a high speed.This is possible due to the property of Silicon which of course fails at very small size arising due to the  issues regarding power consumption, heating etc. George Hanson states in his book "Fundamentals of Nanoelectronics" : "With the current technologies, if the rate of increasing device density were to continue, microprocessor would sooner be producing more heat per square centimeter than the surface of the sun". Thus scientists turn to solutions other than silicon thus entirely moving away from the CMOS technology. Graphene throws in some light here due to its peculiar chemistry and higher rate of electron flow. Though Graphene transistors can be commercially produced; digital signals cannot be processed using these until now. This is due to the zero bandgap of Graphene which prevents it from being turned on and off.

The processor speed frequency seems to have a ceiling though we have hyper threading and multiple cores. A possible and evident reason is that the transistors cannot get faster anymore. We know that the higher the strength of the electric field the faster the transistor would switch. But now the gate isolator underneath the gate on a 45nm process is just about 0.9nm in thickness - just a little more than a Silicon dioxide molecule. Jonathan Kang - CPU Architect on "The simpler matter is that the easy scaling we've had in the past where every shrink in transistor size would lead to faster transistors is over". Thus these calls for a shift towards newer materials.

Another hurdle in increasing speed would be the surrounding ecosystems which is an inevitable bottleneck. This can be within the component itself or the surrounding components in the process chain. Even the bus transport width can be a challenge here. Thus researches in these areas are inevitable. Its high time that the technology we aim today is producing exascale supercomputer which can perform MILLION TRILLION calculations a second. Though the figure sounds fictional like a GAZILLION let us realize it is very very far from the unmeasurable which is INFINITY!!

Thus Hardware developers really have a future ahead!! Why don't you be one among them?