Ai Dreams Forum
Software & Hardware => General Hardware Talk => Topic started by: Art on June 11, 2018, 01:14:24 pm
-
In computers of course...
https://www.youtube.com/watch?v=1FsKfoQxqFE (https://www.youtube.com/watch?v=1FsKfoQxqFE)
-
Ive got an idea for a super computer myself.
I got the idea by accidentally looking at how they make cd's.
When they press cd's dvds, even old school vinyl, they actually get all the music data in one hit off the injection moulder! thats the equivilent of having 4 gigbits (for the dvds) all at the same time in a single copy operation!
The computers we are on now, only put through 32 bits at once, our harddrives are so big, but the compution inlet is very small. theres something that someone could fix there?
-
It almost sounds as if you're comparing Storage space (GigaBytes...not Gigabits), to data access via Fiber Optic data access/transfer. Huge difference.
The CD's and DVD's will give way to Streaming of data for music, video and dare I say Gaming. In fact, a lot of people believe that those Sony Playstations and XBOX's will soon be something akin to the electric typewriter on a dusty shelf in the second-hand store.
Online streaming data access is where it's going to be. Storage will be more cloud-based as a lot of it already is. Why would a person even need to keep all their "private" information locally...on their personal computer in their room at home? - Really?
The Need for Speed will be our ultimate undoing.
Faster A.I. - Faster Robotics - Facial Tracking, Movement throughout your day, etc. Blink...you're already there! O0
-
@ranch, what about a huge stamp with many trillions of pins that pound a electrical stamp sensor below many times per second that send signals down pin lines?.......just an idea......
-
@ranch, what about a huge stamp with many trillions of pins that pound a electrical stamp sensor below many times per second that send signals down pin lines?.......just an idea......
Yes, thats the idea, and it all goes through at once.
But it would be so hard putting down 44 trillion pins, I have to try and clean sweep it some how, with some homogeonous surface.
Just think a piece of paper can have as many dots as you want on it (a piece of paper is in the zotta bits i think :) ), u just have to read it off somehow.
Im actually running an emulator which is doing pattern matching that could be run on it, just need to think.... how the hell do i get 44 trillion parallel apartments? and it isnt easy...
-
Ranch, I've been thinking about that idea some time ago.
Suppose we have a function that takes a 3d mesh as input and gives a bitmap as output (classic 3d rendering function). Theoretically, it should be possible to design an one cycle hardware circuit with input and output wide enough to hold the input and output data, that would adjust output according to input. I think that would represent a great accelerator, but to design such a circuit, we should have some kind of compiler from a language in which the function is described, to hardware transistor setup. That would assume having some automatic transistor assembler, but I think it could work.
But there still might be a few itches about representing program loops.
-
Nice idea ivan.
Well, factories that make 2TB hard drives make all the bits on a disc and sets them to 0s...
-
Ranch, I've been thinking about that idea some time ago.
Suppose we have a function that takes a 3d mesh as input and gives a bitmap as output (classic 3d rendering function). Theoretically, it should be possible to design an one cycle hardware circuit with input and output wide enough to hold the input and output data, that would adjust output according to input. I think that would represent a great accelerator, but to design such a circuit, we should have some kind of compiler from a language in which the function is described, to hardware transistor setup. That would assume having some automatic transistor assembler, but I think it could work.
But there still might be a few itches about representing program loops.
One cycle hardware is thrilling and spoiling. makes you less interested in normal computer hardware.
Thats ok, but ive got my own idea about this. (putting a program on it.)
I think you could put the program on the "paper" itself, and you keep space left for computation, if u were charging backwards and forwards continually on the paper, you could actually "propagate" the data across and it slowly turns into the output, when its finishes filling. it also allows for constant data to be represented at the start position, and it gets to stay there because its got separate space for output, input and intermediate.
the current setup im thinking about, with a grid of capacitors, is theres 2 layers, and they both via differently, so you get different overlaps, and it "ands" as it goes through, and it "ors" as it fills across, so ive got and + or, and thats a unified logical combination, thatll make any fixed program.
so its like a hardisk, but some apartments turn on together - and thats what causes the computation to be possible.
So in this setup its not one cycle, it has to get through the propagation steps, but then its free to release the frame once it gets there.
-
Nice idea ivan.
Well, factories that make 2TB hard drives make all the bits on a disc and sets them to 0s...
yes. i wonder if they can actually start them off with data instantaniously as well. somethings "fishy" here me'thinks.
-
Wait but even if you access tons of data like that, the computation using math still must happen in the billions of transistors in a processor...
-
Wait but even if you access tons of data like that, the computation using math still must happen in the billions of transistors in a processor...
you needent have an alu, a schedulor, or any cache. they are just there optionally.
All you really need to do is carry out truth tables. and if you have AND + OR, or ANY + INVERT, you can put the alu on it, an alu is just truth tables (really anything is.)
... and it are not that big....
If you want to carry out logic, which can be any program imaginable, u just need (or could be) AND + OR, and u can do any program. it needent be transistors, im planning on doing logic with just capacitors. logic takes ANY form, you can make a *mechanical* AND gate if you want.
Any component has some logical effect, needent be a transistor, their use isnt really for logic, its to turn a quiet current into a loud current. logic itself doesnt require amplification, it stays at the same amp throughout the process.
and again... you needent have transistors to amplify either. if you put a big cap after a small cap, they either charge together fully, or not at all. (GO FARADAY!) just like a transistor, anyway.
And u can also get a similar effect out of a mechanical relay.
-
ok i'm lost
can u draw it?
I got left behind at this - a rod with a plate at the end and on the bottom of the plate is trillions of little prick pins....... xD
truth tables ok i kinda get what u mean but really not seeing the device....please draw it in detail (and not too crypto ranch lol!)
I did think of something....instead of a HDD head searching & retrieving 1 bit at a time, why not it look like a tooth brush and retrieve a sector at the same instant? All you do is make sure the computer saves the whole same program in 1 area instead of being scrambled around separated. ?
-
Logic is an inherent property of the universe.
All you need to create a ‘computer’ is a pile of stones and a set of rules.
Think... abacus.
:)
-
yep. thinking only electricity can do it is bad thinking created by the times.
logic to me comes to a motion transfer, so if up and down go in, and left and right go out, logic has happened. it has "transformed."
-
@ Korrelan - The Abacus maker walks up to an abacus owner and hands him one bead. The owner looks at it then at the guy and asks, "What's this?"
The maker says, "It's the latest Update!"
===============================
OK...Years ago there was talk that a Japanese company had been experimenting with 3D storage technology for memory.
Imagine one Period/Dot at the end of this sentence.
Now knowing where that Dot is located, place another Dot directly on top of the previous one.
Then repeat this several more times but noting the relative position of each Dot that's in a particular Row and Column and Vertical position and the data that they contain (since a single dot can hold hundreds of bits of data).
Can one imagine how much data could be stored on a single postage stamp in this 3D configuration? Oh yeah...you bet, a lot!!
There's 3D X-Point and some NNANDS etc. but I think they were talking about something far better and faster.
Interesting concept and maybe we're not that far away.
I remember 'bubble memory' way back in the day so there!
-
"OK...Years ago there was talk that a Japanese company had been experimenting with 3D storage technology for memory.
Imagine one Period/Dot at the end of this sentence.
Now knowing where that Dot is located, place another Dot directly on top of the previous one.
Then repeat this several more times but noting the relative position of each Dot that's in a particular Row and Column and Vertical position and the data that they contain (since a single dot can hold hundreds of bits of data).
Can one imagine how much data could be stored on a single postage stamp in this 3D configuration? Oh yeah...you bet, a lot!!"
That doesn't make sense...so you mean that 1 dot in the 3D block has "more" "qubits" of data because the machine looks at the dots in its rows that relate to it or um.....like killing 2 birds with 1 stone? Explain...
-
There was really no mention as to the ultimate length, width or height of the 3D memory device. Only that theoretically such things could be possible whether or not the Japanese or any other creative manufacturing country could do it or could have already done it. Just exploring some possibilities.
-
I remember 'bubble memory' way back in the day so there! <-from art san. (remembering poor mr lap with a beautiful chinese tree next to a lemon tree lay where he die.)
"Bubble memory" well indeed yes, it will create a super computer a single cm^3, with only 1 micrometre the tiny bubbles apart. But more minification is not required, because we are exchanging hertz for an extra register multiplier! (the bubble space from the bottom to the brim of the glass.)
We will gain ~1,000,000,000,000,000 parallel registers at a 60hz program finish-frame rate