Wtf is the "nm" process with GPUs? [SOLVED]

[Tim Drake]

Can someone explain the nm process to me?

I don't understand it at all..

 

Like people say the smaller the better but idek what it is.

[Aphala]

The space between transistors on a silcone die.

 

Smaller measn faster jumps in the gaps, but also means they get hotter.  (Edit: Gets cooler).

 

OR Band Gap.

 

https://en.wikipedia.org/wiki/Band_gap

 

Things like 14nm beat the current 28nm dies and would be VASTY superior but eventually gets to a point where the gap becomes too small and the electrons jump randomly, then Quantum Physics come into play but that's another story.

[Masada02]

nm stands for nanometer. Basically is the precision that the manufacturing process can hold while making the chips.

[AlwaysFSX]

It's how big gates and stuff are. The smaller they are the less time it takes for electricity to move through it -> faster.

[Tim Drake]

The space between transistors on a silcone die.

 

Smaller measn faster jumps in the gaps, but also means they get hotter.

 

OR Band Gap.

 

https://en.wikipedia.org/wiki/Band_gap

 

 

nm stands for nanometer. Basically is the precision that the manufacturing process can hold while making the chips.

 

 

It's how big gates and stuff are. The smaller they are the less time it takes for electricity to move through it -> faster.

Thnx gais

[GER_T4IGA]

The space between transistors on a silcone die.

 

Smaller measn faster jumps in the gaps, but also means they get hotter.

 

OR Band Gap.

 

https://en.wikipedia.org/wiki/Band_gap

 

Things like 14nm beat the current 28nm dies and would be VASTY superior but eventually gets to a point where the gap becomes too small and the electrons jump randomly, then Quantum Physics come into play but that's another story.

Wrong. Smaller means cooler. Ask any studied engineer.

It results in shorter reaction times of transistors and therefore shorter times in which your circuit shorts--> and heats up. It is to complicated for me to explain it in english because this is not my native language. But trust me this is it...

[Tim Drake]

Wrong. Smaller means cooler. Ask any studied engineer.

That's what I thought considering Maxwell was much cooler than anything AMD had because of the nm process, right?

[Masada02]

That's what I thought considering Maxwell was much cooler than anything AMD had because of the nm process, right?

Maxwell is on the same nm process as AMD's cards.

[Aphala]

Wrong. Smaller means cooler. Ask any studied engineer.

 

Really? I though small jumps with electrons would increase the temps on the small dies?

 

Or am I just having a Friday melt down lol

[Delusional Phil]

Can someone explain the nm process to me?

I don't understand it at all..

 

Like people say the smaller the better but idek what it is.

This video clears it up pretty well ... 

 

 

Also GPU cores use the 28nm transistors because it's different from your CPU ... It has hundreds , even thousands of cores and you certainly wouldn't want then to be that small with many cores around it ...

 

Don't worry though , 14nm is coming this year or next year ( I think it's with Arctic Island and Pascal )...

[Masada02]

Really? I though small jumps with electrons would increase the temps on the small dies?

 

Or am I just having a Friday melt down lol

 

The smaller process means it takes less energy to complete the circuit, but because the chip is fundamentally smaller it heats up faster with the same amount of energy flowing through it.

[GER_T4IGA]

That's what I thought considering Maxwell was much cooler than anything AMD had because of the nm process, right?

Yes exactly,

smaller means less capacity in the transistors and faster reversing of polarities resulting in less time for the circuit that it is shorted and less heat... OK it was easier than I thought

[GER_T4IGA]

Maxwell is on the same nm process as AMD's cards.

NO

Just ... no

GCN is totaly different fromw what NVIDIA does and GCN 1.2, the one currently used is 22nm if I am not mistakeing.

[ybriK]

That's what I thought considering Maxwell was much cooler than anything AMD had because of the nm process, right?

No both Nvidia and AMD have been using the 28nm process since 2012.

[GER_T4IGA]

Really? I though small jumps with electrons would increase the temps on the small dies?

 

Or am I just having a Friday melt down lol

What do you mean by jumping electrons?

[GER_T4IGA]

No both Nvidia and AMD have been using the 28nm process since 2012.

DERP...

sry I brainfarted again.

[Masada02]

NO

Just ... no

GCN is totaly different fromw what NVIDIA does and GCN 1.2, the one currently used is 22nm if I am not mistakeing.

I didn't say the core design was the same, but that the precision level of the manufacturing of the cores was the same. They are both at 28nm.

[GER_T4IGA]

The smaller process means it takes less energy to complete the circuit, but because the chip is fundamentally smaller it heats up faster with the same amount of energy flowing through it.

If all architecture was using the same amount of power you would be right but you are not.

The power needed to run a circuit decreases with size and therefore you assumption is not correct.

[GER_T4IGA]

I didn't say the core design was the same, but that the precision level of the manufacturing of the cores was the same. They are both at 28nm.

Yes I already admitted that. I am sorry for presenting wrong facts in this case. I can't remember what my brain was doing then... msut have been something stupid sry

[ybriK]

I don't know the full details but the efficient memory compression scheme is what makes maxwell run generally cooler lower TDP and lower power consumption compared to AMD. Something to do with the delta color compression from what I remember when maxwell was announced.

[Masada02]

If all architecture was using the same amount of power you would be right but you are not.

The power needed to run a circuit decreases with size and therefore you assumption is not correct.

We both just said the same thing, lol.

 

100W running a chip at 28nm will be converted to the same temperature increase as 100W at 14nm, but because there is less material used in the smaller process it takes less time for it to increase. Unless I have forgotten something from Thermodynamics I believe this is how it works.

[Aphala]

What do you mean by jumping electrons?

 

There are bridges between gates which eletrons hop over, the small those gaps get the easier it is for "loose" eletrons to jump without being told too.

 

The small these Die processes become the harder it is to control them although we can up until it hits 7 to 4nm> we should be okay after that It moves onto Quantum mechanics which is a whole other ball game.

[Aphala]

We both just said the same thing, lol.

 

100W running a chip at 28nm will be converted to the same temperature increase as 100W at 14nm, but because there is less material used in the smaller process it takes less time for it to increase. Unless I have forgotten something from Thermodynamics I believe this is how it works.

Like induction heating? 

[Masada02]

Like induction heating? 

 

I'm not really sure that they are similar. Induction heating uses an electrically produced magnetic field to heat metals, but with cpus its the electricity flowing through the metal itself that creates the heat. I'm just kind of generalizing, but if you think of the chip as 1 big resistor all of that energy that the cpu draws in to make calculations gets converted to heat. So if 100W is converted, the chip with the smaller amount of material will have a higher energy density and will get hotter faster, all with the same total amount of heat being produced.

 

Hopefully I'm not fully talking out of my ass, someone please correct me if I'm wrong.

[Aphala]

I'm not really sure that they are similar. Induction heating uses an electrically produced magnetic field to heat metals, but with cpus its the electricity flowing through the metal itself that creates the heat. I'm just kind of generalizing, but if you think of the chip as 1 big resistor all of that energy that the cpu draws in to make calculations gets converted to heat. So if 100W is converted, the chip with the smaller amount of material will have a higher energy density and will get hotter faster, all with the same total amount of heat being produced.

 

Hopefully I'm not fully talking out of my ass, someone please correct me if I'm wrong.

Yeah no idea why I said induction heating...lol

 

But yeah, I hope the new 14/16nm process will give the 2012 one a kick in the stones.