What happens when transistors can't physically get smaller?

What happens when transistors can't physically get smaller?
What do we do? Is there any way to overcome the limits?

Here's a stupid idea- accelerate a computer to relativistic speeds to induce time dilation

Other urls found in this thread:

youtube.com/watch?v=UvluuAIiA50
twitter.com/NSFWRedditGif

I hope you realize electrons already move at relativistic speeds and time dilation does effect them on a very insignificant scale

The key word here is relativistic. Entropy is only less effective in relation to the electrons and silicon gates themselves, not the computer on a macroscopic scale. Time Dilation is only of real interest to people going at high velocity and when observing distant fast objects. High-school dropout faggot

What we can do is deepen out of order pipelines, increase core count, and then some. A last resort would be to increase die surface size but this would cause exponential increase in heat output.

Interesting

make bigger chips dum*ass

What I wonder is how it's possible to make the transistors that small. Even a 32nm process, how do you manipulate things that small?

Read up on photolithography. A UV mask is etched into a silicon waver essentially. This is why its cheap to mass produce semiconductor ICs but costs billions of dollars in R&D and equipment to make them smaller. Its a high initial investment but a guaranteed reward.

I think they shine light through a sort of cast.

Electrical engineering student here, this is false. Your average electron is moving with an average speed in centimeters per second. What's moving at "relativistic speeds" are the electric fields that excite said electrons. The thing that's stopping you isn't the speed of the fields, or the electrons -- but the size of the fields you generate when you switch things on and off at extremely high frequency. Your processor runs at 2.5GHz, but that's just about the ONLY thing in your computer running at 2.5GHz. All of the interconnects and buses on the motherboard, all of the memory and storage, and basically everything that's not literally in the silicon chip is limited to lower frequencies to prevent transmission line effects from getting out of hand. At 2.5GHz your wavelength is small enough that you can't just make 2cm long copper trace on the PCB without having to deal with all kinds of shit.

That's why moving away from electric computers and towards optical computers is going to free us up in the future. It's the same reason fiber optic cables are orders of magnitude better than copper cables for internet bandwidth. Losses are smaller, capacity is higher, and most importantly light doesn't generate magnetic field as it travels through a glass tube, and even if it did it doesn't interfere with other glass tubes because magnetic fields don't induce light in glass like they induce current in copper.

Global Foundries has a great entry-level video on it
youtube.com/watch?v=UvluuAIiA50

quantum gates

What what point does quantum computing become a viable alternative to shrinking the die?

Any recommended reading/watching for optical CPUs?

The moment that keeping a computer at 2 degrees above absolute zero becomes economically viable.

Quantum computing is a meme, even if they weren't stupidly infeasible they are still very specialized and not suitable for general compute.

That's basically how computers work on Star Trek, only they use warp fields instead of actually accelerating it.

None I know off the top of my head. Optical is still pretty far off and everything I know about it has been heard second-hand from professors teaching us about the electrical circuits and their inherent drawbacks. While the interconnects are all fine and dandy, how are you supposed to go about storing information in the "optical domain"? You can't have a CPU without first having registers and flip-flops and other basic logical units. What does a single stored bit of light look like? How do you interact with it and change it without having to jump through conversions back to the slow electrical domain? There's not much point in making all of your lines optical if you still have to use electrical relays.

doesn't work, because of "speed of light limitations" etc

Light is just a specific range of EM radiation. Why not UV or X-ray for example? :D

There is a significant difference between an EM field and an EM radiation. It's why you can put current through any old wire and make a magnet, but if you want to put current through a wire to make an antenna you have to get down and dirty with wave analysis.

his get says otherwise

No it's literally the same thing.
If you put alternating current through a straight wire it creates a dipole field, which results in EM radiation that looks like this.
Of course there are better ways to do this, but it still works if you do it like this.

"the cloud"

This is what I'm talking about it's not just "a straight wire".

basically we're heading towards the AI

it's an obvious sign

We improve the algorithmic side of the things, so stuff doesn't take fuckhuge computations per second.

Get a load of this guy lol, he is seriously advocating optimising code. I guess you expect people to actually be proficient in their jobs too? Software is racist and sexist towards POC and *women.

We will make them out of graphene!

Ironic shitposting is lazy shitposting