View Feed
group-icon
Coffee Room
Discuss anything here - everything that you wish to discuss with fellow engineers.
12933 Members
Join this group to post and comment.
Jeffrey Arulraj
Jeffrey Arulraj • Jul 5, 2013

1 THz processor energy requirements

https://www.crazyengineers.com/threa...by-samsung-nvidia-qualcomm.68860/#post-282470

Check the comments of this thread and you will find Abhishek Rawal and Megoing into another heated row in CE

Well this is not the first and it will not be the last time here as well

Back to topic What in your heads seem to be the limitations or the requirements which must be met by a 1THz processor
Abhishek Rawal
Abhishek Rawal • Jul 5, 2013
I think major question is, Is it possible to embed 1 THz clock-speed processor in a single mobile-sized die ?
I don't think it's possible until Nanotechnology development in semiconductors becomes very mature.
We need something that can replace transistors ,i guess 😛
Jeffrey Arulraj
Jeffrey Arulraj • Jul 5, 2013
Heard of Quantum dot FET

Here the gate length is reduced a lot to an extent of .1nm

The actual gate length is due to a single phosporous atom instead of a channel The movement of electrons from the source to the drain in the transistor is made successfully by this single phoshorous atom itself

So one of your problem is solved

Next question ??????😛😛
Abhishek Rawal
Abhishek Rawal • Jul 5, 2013
Honestly, I don't see any logic in discussing about this topic.
In 2020, you'll re-read this thread & compare the 2020's processor's clock-speed, and you will find that there's not a single palm-sized die developed which can handle even 1 THz of clock-speed.
You're talking here about clock-speed & not just a memory. 1024 GHz in 2020,a dream!
Not being pessimist, but a realist.

PS : I am out of this thread. If we discussing about current technology, then I am in.Sorry, but this makes no sense.
silverscorpion
silverscorpion • Jul 7, 2013
I refer you to this 9 year old article - "The free lunch is over"

The article primarily deals with traditional processors and not mobile processors. But the trends and the general message from the article applies also to mobile processors. And the interesting thing is, this trend is still valid..

[​IMG]
From the graph, we see that the number of transistors in the processor increases, as expected. That is what the Moore's law says. Ok. But the clock speed, power dissipation and the performance are all stabilizing and not increasing anymore. This is because, the number of transistors doesnt directly translate into performance. Each time you increase the clock speed, you need to increase the voltage level for the processors, and power has a square relationship with the voltage. This means, if you double the clock speed, you quadruple the power consumed. This alone is a big problem, especially in mobile processors.

That is what led to multi-core processors. Let's say you have a processor with 1 MHz clock, and it uses 10 W power. If you increase the clock speed to 2 MHz (which means it's capable of 2 MFLOPS), the power usage will go to 40 W. Instead, if you get another similar processor and connect them together, you just need 20 W power, and you still get 2 MFLOPS performance. This is the basic idea, why multi-core has become very common. It's the way to go, since we cant do much more with single core processors anymore.

So, parallel and concurrent software is how we get more and more performance these days.. Hardware improvements are there, no doubt, but those are not nearly anywhere near wat we used to get in the 90's and 2000's. Today, we just put more cores in a processor and tie together many such processors, and write parallel programs for whatever we want to do.. So, ultimately, even if a 1THz processor is realized at some point in time, it won't be using a single core. It will be a multi-core or many-core processor which has lots of smaller cores with smaller clock speeds.

Share this content on your social channels -