Moore's Law

>Moore's Law

More like Moore's Theorem am I right ?

Or are chip makers going to bust out the top secret better than silicone tech?

Other urls found in this thread:

youtube.com/watch?v=msVuCEs8Ydo
physics.stackexchange.com/questions/2175/is-it-possible-for-information-to-be-transmitted-faster-than-light-by-using-a-ri
en.wikipedia.org/wiki/Drift_velocity
twitter.com/NSFWRedditGif

Well the idea is that alternative computational paradigms will emerge around the same time we exhaust the possibilities for 2D photolithography on silicone dies. 3D computing and quantum computing are the two leading potential trajectories for future computing hardware development.

We are at

>IBM z13 storage controller has more transistors than the cpu itself
whut ?

>doesn't know about DARPA chips
One day you'll be Sup Forums, keep working hard.

>theorem
you mean conjecture. theorems are proven
>silicone
wrong valley, that's San Fernando

>AMD stops being competitive because they couldn't generate revenue from selling due to Intel shady partnerships
>suddenly moore's law "stops"/stagnates for nearly a decade

>years later

>AMD pulls one up on Intel and pulls together a glued CPU stack
>Intel suddenly rushes to release 6c as mainstream
>AMD releases monster 16 cores 32 threads for consumers
>moore's law resumes

What did they mean by this?

I could see virtualisation becoming more of a thing with improved technologies. Imagine your phone was more of just a screen/remote controls/receiver, processing via satellite or terrestrial station, meaning you don't need all the processing in your pocket, and you can request/pay for resources as required. So when there's no more compact space; virtualise.

Can a 1nm processor at 4Ghz calculate more shit per second than a 10nm processor at 4Ghz when they both use the same microarchitecture(?). So the 1nm chip is just the shrinked down version of the bigger one. As I understand it, it would perform exactly the same but just use less power, am I wrong?
If this is true, would the only performance increase come from clock speed increase (excluding shit like creating a better microarchitecture or speed up through more cores for paralleliseable work)?

Other question: What if I create a 1nm chip that's as big as the 10nm chip but they both run at 4Ghz. Can I use the additional space for anything but more cores (won't increase single threaded performance)?
I'm asking this because I'm not sure what inherent perfomance benefit denser dies give without clockspeed increase.

I suppose it's time for standard PCs to physically start changing, the same way we aren't today using upgraded IBM 360 or PLATO 3 systems.

They wont use the same microarchitecture.

Of course.
I searched my question and someone said what I expected:
>It doesn't really, not by itself. With a die shrink, you can either make the same thing with less silicon (higher yields, more chips per wafer, so overall lower production cost, and lower power consumption per chip) or you can put more transistors in the same amount of silicon, improving performance.

that makes sense to me. What I'm not sure about is, is a 1nm ALU inherently faster than a 10nm ALU or would it only be faster thanks to higher clock speed (that the 1nm arch will probably allow)?

no one wants to run Crysis on a phone. Technology is already becoming good and cheap enough for everybody.

x times more transistors on the same surface...

Ok, so that would mean if I can use let's say 5 times as many transistors for a single core and maybe cache, I will get a definite performance increase without having to up the clockspeed, right?

1nm likely isn't possible (for mass production) without some crazy engineering shit

of course, it's the (amount of transistors) x (the frequency) that determines the performance of a processor (for the same architecture and number of cores).

in fact it's the number of transistors that follow moore's law, not CPU clocks...

a pentium 4 prescott could run up to 3.8 GHz but only got 169.000.000 transistors.

a ryzen CPU runs about the same clock (4GHz) but got almost 5 billions transistors.... hence the performance gap.

just make the cpu bigger lmao

they could but would cause heating issues (you would need a higher voltage, higher electron current through much more metal=>Joule effect) and moreover, less profit...

It would dramatically increase the electrical resistance of the CPU, and the power consumption.
Also, silicium isn't exactly cheap...

Not to mention lightspeed lag. The higher the clock speed, the more limited to total size of a core is. If it takes your electrical signals too long to reach one side of the chip from the other, you don't have a working chip.

Latency

Lmao just slap more cores

>lightspeed lag
Isn't electricity supposed to be instantaneous?

nothing is instantaneous dude...

The way I see it, If I were to push a vase across the room using a long metal bar, there wouldn't be a delay between my hand movement and the vase moving, same goes for electricity.

Do you know why speed of light is written as "c" in equations? Speed of Causality.
youtube.com/watch?v=msVuCEs8Ydo

It takes time for the atoms to push eachother, or in the case of electricity, the electrons
Nothing can surpass the speed of light

It's not really a good analogy, current is a flow of electrons through the silicone crystal, not a transfer of mechanical force...

>silicone
*silicon lol

there is a delay, a "push" takes time to propagate through a material

physics.stackexchange.com/questions/2175/is-it-possible-for-information-to-be-transmitted-faster-than-light-by-using-a-ri

this holds true for everything, including electricity
en.wikipedia.org/wiki/Drift_velocity

>there wouldn't be a delay between my hand movement and the vase moving

In fact depending on the elasticity of the materials the vase and the rod are made of, their inertia, and the action of frictional forces there would be a delay (a time needed for their acceleration). a delay that a human wouldn't notice but it would exist.

mechanical force is the interaction of electrostatic fields, whereas electricity is the flow of electrons that is caused by an electric field (the field exerts force on the electrons)

in both cases there is a limit to how fast the object being acted on (electrons or metal bar) can move from one place to another

I'm pretty sure that with a rod long enough and a high speed camera we could actually witness the delay

This is the closest to Moore's original formulation in the thread, but still not quite right. It was number of transistors per cost

Probably not, it'd be virtually immeasurable visually. Kinda like the definition of a meter, you'd have to use similar techniques to "watch" such a minute interaction occur.

...

then again, with a rod long and flexible enough you would probably see the deformation propagating through the metal before pushing the vase

Just used it as an example.

Thanks.

People who cite Moore's law a lot ignore Wirth's law much to their detriment.

Flexible materials would compromise the experiment, wouldn't they? You'd have to load it before it would start imparting force, even if it was only a minute bit.

Well no material is infinitely stiff, a flexible rod would just enhance the delay for experimental purpose...

And don't forget quantum tunneling could be observed at that level so it would be worthless unless you cool it to 0+k temps

or a very heavy vase if you prefer.

Interesting to hear the opinions of people like Carmack on this, he's said he thinks we'll get an order of magnitude out of computing but probably not 2. So maybe another 1000x faster.

Near as I can tell Intel have road mapped technology down to about 5nm where not only materials but the quantum effects of shrinking become significant barriers.

Not Intel per se but the guys who make the machines capable of that for Intel.
Even if we get those processors in 4 years time I'm not sure where can we go from there.

>silicone
Retard

That's completely wrong. Your push will travel through the metal rod at the speed of sound in the rod.

but what if my rod is made out of whatever exotic matter is at the center of a black hole?

As far as modern physics understands, nothing goes faster than the speed of light in a useful way. A lot of extremely smart people have spent a very long time trying to find exceptions to this, and examining in detail cases where it looks like superluminal signaling might be happening.

This has lead us to very interesting discoveries, deepening our understanding of the universe, but nothing has straight up violated it. As you might be familiar, there are cases where objects too far apart to communicate slower than the speed of light have correlated states. Nonetheless, the laws of the universe seem to conspire to prevent us from actually getting faster-than light communication out of it.

I'd absolutely love if we discovered FTL, but the safe bet is that we won't, and the speed of light is somehow a fundamental part of the universe. Like in Conway's Game of life: although there are tricks that make it appear like you can signal faster than one cell per tick, you simply can't. It's hard-coded into the most fundamental rules of that universe.

Physicists obviously can't say it's "impossible", since they always defer to observation. If someone gets FTL working, it doesn't matter what all our (now-dated) math says is impossible: the universe says otherwise, and it's the job of physicists to model the observable truth of the universe to make predictions. But it really seems to be one of the most fundamental truths in physics, that had continued to hold even in the face of extremely strange situations and observations.

Just mainframe things