IBM Reaches Graphene Milestone, Makes Much Faster Transistor 155 Gigahertz!

http://news.softpedia.com/news/IBM-...one-Makes-Much-Faster-Transistor-193986.shtml

Now this is a FAST processor indeed 155 GHz!
Intel, AMD, can you keep up?

 

Damn that's fast. Is that only one core?

 

Yes really really fast

About if it has only one core, I'm not sure actually. But at that speed 155 GHz I can accept that it may only have one core.

If I get that processor I wouldn't need to buy a new one in 20 years at least, if at all

 

A transistor is NOT a "core". A transistor is just the basic building block of any electronic circuit. A CPU is a complex device that contains hundreds of millions (maybe billions) of transistors.

What the news article is saying is that graphene (an allotrope of carbon) in nanostructure form was used instead of silicon and/or germanium (as is used today) to make a transistor that can perform operations (i.e. oscillate) at a frequency of 155GHz (This was not previously practically possible due to various parasitic and stray wire capacitances that arise at such frequencies, but I'm getting way too technical here ).

 

So if you combine all those transistors, will the theoretical limit be much higher than 155GHz?

 

Doubt it. 155GHz relates to the frequency of the input signal (frequency in general means number of times something occurs per second). In such kinds of circuits, operations are done via various "functional blocks" - sets of electronic circuits specialized for doing some kind of task (e.g. math calculations or display render, etc.).

Let's say you have one transistor which takes an input signal and (theoretically) outputs at 155GHz. In any functional blocks the transistors are cascaded (i.e. many transistors are "pooled" together to achieve various characteristics). However, the other transistors also have internal resistances and capacitances which oppose the motion of the signal through the circuit.

As the total opposition to the signal increases with the number of transistors, less useful signal reaches the output (in transistor theory we are concerned with what comes out finally as compared to the electrical signal we give to the transistor at the beginning).

A typical transistor frequency response graph is shown below. The "frequency response graph" is a graph drawn of "voltage gain" (output voltage vs. input voltage) and frequency of the signal.

http://img96.imageshack.us/img96/1554/p73ws.jpg

You can see that the gain reduces at the highest frequencies. As the number of internal capacitances increase, the highest frequency at which gain will reduce will also decrease. Therefore, combining all those capacitors means the whole circuit will have to work at something less than 155GHz if one wants to have maximum efficiency of operation.

Note that current transistors have their gains reduced at about 40GHz, i.e. they can operate only upto something about that frequency. However, no core so far has reached 40GHz, the best we can see is around 3.6-3.8GHz. So if we consider that the graphene transistors can work at speeds of upto 3.8x faster than current transistors, then we can expect a 3-4 times rise in the speed of cores. So practically, if progress goes fast enough we will be looking at 12-15GHz cores.

Of course, there's a lot more to be discovered as far as graphene transistors go. Specially how to make them work as digital circuits......(It may take 10-20 years, but I'm being pessimistic)

(Sorry if this went way too technical, I tried to explain as simply as possible)

 

It is a bit over-technical, but no need to be sorry.

 

Yes, but very interesting at the same time. Thank you Firecat.

 

Firecat provides some nice info

However unlike Many analog circuits, digital circuits & computing don't rely on the gain/bandwith equation/limitations, as there is NO amplification of signals. The transistors only work as switches, therefore they are able to turn on/off at a higher rate.

I agree though, that whatever highest fT = frequency of transition, is stated in the data sheet, it is a maximum "possible" under only certain circumstances. So in reality it might not be achievable.

As the article itself states

So even for digital switching = NO gain, at those speeds they are subject to internal/design/manufacturing limitations. But it's still a milestone, & they always manage to integrate new ideas into products, sooner or later

 

Well, theoretically you could say that no amplification means output = input, so basically gain factor is 1. Anyway, what we need is whether you get some output or no output. At high frequencies you get to "no useful output" state anyway and that's what causes digital electronic circuits to have frequency limitations (I think).

Indeed, the processes have yet to catch up with theory

Also way too many things have to be determined now for these transistors to be used even in limited analog circuits, because IMO essentially the graphene transistors are a "new device" (similar to how we compare the old vacuum tubes to current electronics and so on).