are we going to see the first terahertz cpu in the future

[Glenwing]

Basically, when you get up to these frequencies, the speed of electricity actually matters.  Light travels 300,000,000 meters in 1 second.  So, if you have a 1GHz CPU, doing 1,000,000,000 cycles in 1 second, light can travel 0.3m (30cm) before the next CPU operation occurs.  In a 10GHz CPU, light could only travel 3cm before the next CPU cycle occurs.  That means the maximum distance (so, diagonal) on a 10GHz CPU die can theoretically be no greater than 3cm, less in practice, but perhaps it could be stretched a bit with some clever engineering, ensuring that one side of the CPU is never required to signal to the opposite corner or something... It would be a whole new level (nightmare) in CPU architecting...  Anyway, the limit would be reduced to 1.5cm for a 20GHz CPU, 3mm for a 100GHz CPU, and so on.  So, to increase clock speeds we would be limited to a very small die size, and we would basically rely on further transistor gate shrinks to be able to fit more transistors in that maximum area.  Transistors as well though, are approaching the scale where the size of atoms restricts any further gate shrinks, as well as electron leakage.  Carbon atoms such as in graphene are smaller than Silicon atoms, so using graphene would allow for further shrinkage than Silicon, beyond the 1nm barrier, but the minimum size limit would only be pushed a little further back until we are restrained by the size of Carbon atoms too.

 

At that point, the entire foundation of computer technology would need a successor to advance any further, no longer based on semiconductor transistors.  Such as quantum computing.  Or black hole computing That's a fun one...

[Mentalguy]

But what about those processors that would use light instead of electricity to transfer data? I can't remember what they're called, photon processors? Anyway, like I said, it's a processor which would use light to process data instead of electricity. That would eliminate crosstalk interference, and heat output. And would enable use of more light beams parallel. That's all I remember, it's a concept, but a very plausable one.

read glens post.

 

At that point, the entire foundation of computer technology would need a successor to advance any further, no longer based on semiconductor transistors.  Such as quantum computing.  Or black hole computing That's a fun one...

that is pretty much it

[iWearKiltz]

we'll be using multiple activities in the future, so i'd reckon that for commercial use, more cores and possibly less clock speeds to save power

[Eroda]

http://www.technologyreview.com/view/518426/how-to-save-the-troubled-graphene-transistor/

 

took me a second to find it again but 400ghz transistors  very simple transistors but still i don't see silicon being viable much after 2020 so they are going to have find something to replace it eventually

[Dsgreek]

My guess is that we will have quantum computers before we even get near to those figures.

We are so close to making a breakthrough in Quantum computing. My mum works at the UNSW and she's always like QUANTUM COMPUTERS ARE THE FUTURE ( yet she knows like nothing about hardware in pcs )

But yeah we are close to a breakthrough

[SSL]

Don't forget that as CPU speed increases, the memory-CPU performance gap also increases. Throwing more on-die cache at the problem won't help because that would require a larger die.

[Gob]

Basically, when you get up to these frequencies, the speed of electricity actually matters.  Light travels 300,000,000 meters in 1 second.  So, if you have a 1GHz CPU, doing 1,000,000,000 cycles in 1 second, light can travel 0.3m (30cm) before the next CPU operation occurs.  In a 10GHz CPU, light could only travel 3cm before the next CPU cycle occurs.  That means the maximum distance (so, diagonal) on a 10GHz CPU die can theoretically be no greater than 3cm, less in practice, but perhaps it could be stretched a bit with some clever engineering, ensuring that one side of the CPU is never required to signal to the opposite corner or something... It would be a whole new level (nightmare) in CPU architecting...  Anyway, the limit would be reduced to 1.5cm for a 20GHz CPU, 3mm for a 100GHz CPU, and so on.  So, to increase clock speeds we would be limited to a very small die size, and we would basically rely on further transistor gate shrinks to be able to fit more transistors in that maximum area.  Transistors as well though, are approaching the scale where the size of atoms restricts any further gate shrinks, as well as electron leakage.  Carbon atoms such as in graphene are smaller than Silicon atoms, so using graphene would allow for further shrinkage than Silicon, beyond the 1nm barrier, but the minimum size limit would only be pushed a little further back until we are restrained by the size of Carbon atoms too.

 

At that point, the entire foundation of computer technology would need a successor to advance any further, no longer based on semiconductor transistors.  Such as quantum computing.  Or black hole computing That's a fun one...

 

Further, it is not solely the size of the atoms that constrain the size of the classical transistor. At smaller and smaller transistor sizes, the distance between the source and the drain of the transistor is so small that it gets harder and harder to control the flow of current due to quantum tunneling.

 

However, I have heard of some promising research using photonics (much faster clock speeds, less electromagnetic cross talk, significantly lower temperatures, etc).

 

We are so close to making a breakthrough in Quantum computing. My mum works at the UNSW and she's always like QUANTUM COMPUTERS ARE THE FUTURE ( yet she knows like nothing about hardware in pcs )

But yeah we are close to a breakthrough

 

Quantum computers are not a replacement of classical computers. In fact, in heavily serialized loads, a quantum computer will perform far worse than a classical one. Quantum computers are built for parallelism and that is where they shine (understatement).