Suppose you took a processor, and made an exact replica, only 100 times larger. Would it run faster or slower?

Suppose you took a processor, and made an exact replica, only 100 times larger. Would it run faster or slower?

Assuming your motherboard would also be 100 times larger of course. Other components can be regular size.

Other urls found in this thread:

youtube.com/watch?v=9JRLCBb7qK8
en.wikipedia.org/wiki/Speed_of_electricity
twitter.com/SFWRedditVideos

Slower

>More transisters
>Bigger instruction pool
>More logic gates
>Bigger die
>Slower

it wouldn't work at all. processors rely on quantum properties that only exist at very small scales.

There's an interesting video about that actually:

youtube.com/watch?v=9JRLCBb7qK8

>processors rely on quantum properties
They work around quantum properties, they don't rely on them. Transistors work on any scale

Slower.
Signals take longer to travel.
That is, given that they can cover the distance at all, with those minimal voltages.

>Signals take longer to travel
Ultimate lol

>electrons can move instantaneously

the difference would be miniscule but still technically slower

Slower, mostly because of massively increased gate capacitance. Look up the MOnSter 6502 for an actual example. That's an old CPU rebuilt with discrete parts that runs 10 times slower than the original.

DO YOU KNOW WHAT LIGHT SPEED IS
DO YOU

tips fedora physicucks btfo?

>Electrons move at the speed of light
Wew lad

Are you retarded?

Even at 1% of C it is still ridiculously fast. Difference at such a tiny scale is negligible

en.wikipedia.org/wiki/Speed_of_electricity
>In copper at 60 Hz, v ≈ {\displaystyle v\approx } {\displaystyle v\approx } 3.2 m/s.

negligible yes, but still technically slower over a longer distance like i said.

>bigger die
larger power requirement
>larger power
larger heat
>larger heat
less efficiency


>bigger die
higher latency
>higher latency
slower intra-communications

>exact replica
Same number of everything, user.

The larger distances between components makes the processor slower because the speed of light does not change.

>copper
What is this, the industrial revolution?

I saw your /sci/ thread, you nigger. Bigger transistors take more time to get stable, so you would have to greatly downclock it, in like orders of magnitude.

Difference between few nanometer produces 10-15% performance gains per generation.

100X difference would produce 1500x speed difference.

>exact replica, only 100 times larger

And for those who don't have a degree in areas related to computer engineering, please abstain from spouting non educated guesses. Better yet, go back to Sup Forums.

...

>copper
this isn't used anymore grandpa

well I'm not finding any reliable sources for conductivity values of doped silicon but even if it's a few orders of magnitude higher than copper it's still not anywhere near a meaningful fraction of c like suggested

Probably wouldn't run at all. The doping and geometry for a transistor changes dramatically as the transistor changes size. There is an awful lot of shit that had to be changed to get from the 486 size to Ryzen size. You wouldn't just blindly scale a house to make a skyscraper.

I would if I could

Larger MOSFETs = higher gate capacitance. Gate capacitance limits signal and clock rise and fall times, which limits your maximum operating frequency.

You guys are right, electricity moves so quickly that the added distance would have no effect unless we were trying to do something stupid like get the entire circuit synchronised at an absurdly high frequency. I mean, who ever heard of a chip running at several billion hertz? We'd need to find some shorthand way of writing that shit, like gigahertz. Man, could you imagine a chip running at like, five gigahertz? That'd be insane. That's the point where, like, the speed of electricity and signal propagation would become an issue for the chip. Luckily we don't need to consider that in modern chips at all, no sir.

This, p-n junctions in transistors being 100 times larger would have a massive effect, basic scaling laws would indicate that operating voltage would scale to square root of the dimension which in this case means 10 times the voltage, capacitive effects from ridiculously long leads and from the CMOS themselves would ruin any high frequency operation anyways

>exact replica
wouldn't run at all
you'd have to redesign the entire transmission line wiring network inside since the electromagnetic interactions of the whole damn thing is field simulated prior to layouting where any mismatch could have a catastrophic impact on the amplitude/phase/noise of the processed signal

but in general, assuming you do proper workarounds, this not quite, but this the only correct answers ITT

electromagnetic waves do travel at the corresponding speed of light in a given medium, in this case the silicon substrate i.e. 8.5E8 m/s

retards

>grab myself a ryzen
>have it flown in by military helicopter to be placed carefully on my giant fucking motherboard that spans wisconsin
>order Hellman's extra large mayonnaise jar
>hell yeah coolant
>drive nebraska to flip the on switch
>forgot to plug in my stock fan
>the entire east coast goes up in flames

I'd say it'd run faster, but I don't anything to say why.

As some other anons here already mentioned to some detail, too much size leads to desynchronization because the electrons wouldn't be able to reach other parts of the chip fast enough. Big-ass chips like Epyc are multiple dies that are buffered by infinity fabric, desync can't happen because each one runs independently and only communicates with the others if it needs to. Monolith chips from Intel are similar but more tightly integrated, with each core connected to either a ring bus or mesh that buffers the data. Massive chips like Volta can exist because GPUs are a "one-way" device that doesn't care about feedback on the data it's outputting, and so can afford to be high-latency in many aspects.

light is electromagnetic waves U moron. Also electrons move pretty slowly (i dont remember exact number but we are talking about few cm or m/s not higher) BUT their effect = signal travels thru medium w/ speed close to c.

Also higher capacitance does not necessary make transistors slower but you need more charge to open them so if you need to keep speeds youll need more Amps as C=I.t. 100times larger transistors would draw 100times more power at same speed. Thats also the point of modern die shrinkage.

>namecalling
explains your ignorance but I'll still enlighten those who are indeed willing to learn something ITT
signals ARE electromagnetic waves and nothing else, the drift velocity of electrons on the other hand is meaningless unless we're talking about direct current i.e. zero frequency

>higher capacitance does not necessary make transistors slower
wrong, learn to electronics desu

As an EE, can confirm this guy knows his shit. Rest of the thread are CS kids that think they are smart but just can't into electricity.

>speed of light
CPUs still use electricity to transfer data.

this guy knows what he's talking about

>0.001% yields
Great idea, OP.

And the electric field expands at light speed.

Yes and that's why RAM is slower than the CPU's memory

> (You)
>>namecalling
>explains your ignorance but I'll still enlighten those who are indeed willing to learn something ITT
>signals ARE electromagnetic waves and nothing else, the drift velocity of electrons on the other hand is meaningless unless we're talking about direct current i.e. zero frequency
1. EM Wave can be signal but it is not implied. Smoke can be signal too.
2. Logical levels (0,1 for example) are defined by voltage. It is 5V for TTL logic for example. This is signal carrier in processors not em wave lol. Voltage level.
The speed we are talking about isnt speed of EM wave but field propagation which is nearly instaneous (well close to c). t. Electronic engineer

>>higher capacitance does not necessary make transistors slower
>wrong, learn to electronics desu
Make statement w/o any proof. Neat

What does the speed of the electrons have to do with this?