Instead of spending billions of dollars on making transistors

the speed of light is too slow

>>59797351
I dont understand

>>59797329
Then yields would be lower and the prices would increase. You can't just make a huge chip and expect to not have a shit ton of defective units.

>>59797379
Im saying transistor size would stay the same, just add moore of them?

>>59797329
Stupid questions thread plz.

>>59797329
If you increase the chip's size, you increase the distances and the signals, which are limited to the speed of light like everything else, take longer to be transmitted. At this moment you lose the benefits of the extra transistors.

That's a good question op idk

>>59797428
Ahh okey, that explains it, thanks!

>>59797351
>>59797428
Do you realise how fast the speed of light is, idiots?

>>59797441
I'm not an Intel engineer so please double-check or ask reddit at least

>>59797329
If the chips are larger the cost will be larger.
Do you know how much a printed wafer costs? Hundred large at best. Good luck affording the equivalent of 200≈500 cpu dies and powering that madness. It will take over 10kw/h to idle.

We need denser 3d circuitry and possibly more efficient substrates like super conductivity and quantum processing.

The future has a lot on its horizon. FPGA coprocessing integration into the cpu will be a huge thing for CPUs. Basically task can be accelerated on-the-fly like a dedicated ASIC for algorithmic processes.

>>59797460
>electric signals can traverse silicon at infinite speeds
no

>>59797460
Do you realize that it's the engineers who create chips that say that, and that at the ridiculous scales of today's CPUs the speed of light can matter ?

Also we aren't transmitting signals on the chip with light right now, it's all electricity which is much slower. Please go learn something.

>>59797460
Actually electricity doesn't travel at the speed of light it is slower. Ion transfer is fast but not as fast as photons.

Have you ever heard of latency?

>>59797476
Actually I don't think that the wafer costs that much, it's the initial R&D and the facility building which cost and have to be compensated. Then it's all benefits.
Adding 2 cm to the chip size wouldn't make the price explode. It's a more fundamental problem, or else supercomputers would have 1m2 chips already.

File: 1490027180944-pol.png (132KB, 396x385px) Image search: [Google]

132KB, 396x385px

>>59797528
My point is what is the difference of 1 big cpu which will have very low yelds verses many CPUs with exponentially higher yelds and set them up as a super computer?

It isn't practically achievable. Also the technology has reached it's apex. There are quantum technologies, photonic computation, superconductivity transfer technologies, and architectural advancements. We are reaching size limits with silicone but other materials have the potential to get much much much smaller then 7nm.

>>59797528
Stop talking out of your ass.
First of all, wafers aren't a large portion of the cost directly but the problem is that chips have yields. So, certain top of the line chips might only have 15 out of 100 working chips made. The larger the chip physically is, the lower the yields will become with everything else the same. So, now your chips have become much, much more expensive for a small increase in space.
Second, larger chips pose design issues - larger chips will physically put out more heat and clock speeds will go down due to resistance increasing.

Someone check my math, but I think @ 1GHz, an electron can only go about 2.2 cm?

If you consider all the twists and turns the electron needs to go through, it'd mean the chip would have to be smaller still.

File: 1486709241417-pol.jpg (273KB, 904x711px) Image search: [Google]

273KB, 904x711px

>>59797528
Yields for currently CPUs bring the wafer cost just above 100k (dependent). If you make the wafer 1 large die do you understand how much 1 working wafer would cost? Millions easily due to extremely low yields. How will you power the beast? How will you design the architecture?

This isn't practical.

>>59797460
It's already a problem (or rather, engineering challenge) on current motherboards

Because it's better to make dedicated cores for specific tasks inside the chip than to have general purpose bullshit that will throttle the whole system.

File: image.jpg (37KB, 642x432px) Image search: [Google]

37KB, 642x432px

>>59797329
because economy of scale.

>>59797497
>>59797511
Do you realise that while electrons move slow (few centimeters a second) but the effect of voltage/current happens almost immediately at the speed of light? I hope its bait.

>>59797400
>just add moore
Which is the reason why failure rates increase in the first place. More transistors = more points of failure

>>59797528
Supercomouters already have relatively huge chips, they can do that since the client is willing to pay millions for it. You can't do that for a mass produced product.