Your Thoughts on Cannon Lake?

[WC4L420]

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

Cannonlake is said to have 10nm transistors.

My big question is will 7nm be the stopping point after this?
Or will intel go all the way to a 5nm die shrink?

Skylake is only 14nm and just released. The theory is that 5nm is the smallest transistor we can make before electrons don't move the proper way in the circuits.


And for that matter does anyone have a skylake CPU? I like the box art for them personally.

[initiaLiSeD]

Yeah, once you go below much below 5nm the wave function of the electron is bigger than the gate so it starts to behave more like a neuron than a transistor. A litre of neurons takes about 25 years to program and after that it can't be trusted and begins to deteriorate, rapidly. So unless the transistors are separated (reducing computational density) we are approaching the physical limits of electronics, so maybe it's time to start manipulating the spin of the electrons rather than the flow of electrons, that should get us through the next 20 years or so of miniaturisation before going full quantum computation.

[Azza ☠]

Tick/Tock development cycle.

[WC4L420]

Opprinnelig skrevet av Azza ☠:

Tick/Tock development cycle.

They actually are messing that up with Kaby Lake. It's a tick, following Skylakes tick, then Cannon Lake is the tock to Kaby lake.

Opprinnelig skrevet av initiaLiSeD:

Yeah, once you go below much below 5nm the wave function of the electron is bigger than the gate so it starts to behave more like a neuron than a transistor. A litre of neurons takes about 25 years to program and after that it can't be trusted and begins to deteriorate, rapidly. So unless the transistors are separated (reducing computational density) we are approaching the physical limits of electronics, so maybe it's time to start manipulating the spin of the electrons rather than the flow of electrons, that should get us through the next 20 years or so of miniaturisation before going full quantum computation.

I actually read somewhere a while ago once we hit the physical limit (i.e 5nm) that CPU developers were going to start expandning the size of the chip themselves. Either A) by moving vertically, most likely. Or B) by moving horizontally, not as likely, but still very possible.

I mean imagine if they went horizontally, they would eventually make the socket so big, the whole motherboard is consumed by the CPU. M-ITX boards would cease existance. <That would really suck.

[[AU] Tabris:DarkPeace]

Making CPUs over 600mm^2 is very costly, so they'll probably die stack once they get past 5nm initial designs.

[WC4L420]

Opprinnelig skrevet av TabrisDarkPeace:

Making CPUs over 600mm^2 is very costly, so they'll probably die stack once they get past 5nm initial designs.

But what will that mean for termaldynamics? (spelling?)
I mean i get by that time we'll have really low temps, but wouldn't stacking the die present a heat problem for the die below the upper one?

[[AU] Tabris:DarkPeace]

They'll just use integrated layers like they do now.

Maybe 20 layers of lithography or so.

[Bad 💀 Motha]

I'm sure they'll find a work-around as tech improves.

It wasn't that long ago when we were looking at 2TB drives and face with Motherboards that had that 2.2TB limitation. And the industry fearing mechanical HDD density methods might be reaching a breaking point; as you can see, dead wrong.

[WC4L420]

Opprinnelig skrevet av Bad-Motha:

I'm sure they'll find a work-around as tech improves.

It wasn't that long ago when we were looking at 2TB drives and face with Motherboards that had that 2.2TB limitation. And the industry fearing mechanical HDD density methods might be reaching a breaking point; as you can see, dead wrong.

Maybe Seagates Heat-Assisted Magnetic Recording will be the solution to that.

[initiaLiSeD]

HDD has been running into limits for a while, SSD has come along to fix that. Spintronics, photonics or quantum holography will come along and replace electronics eventually.

[Andrius227]

I think making cpus such small is kinda pointless. There are other, cheaper ways to increase computing power, by making cpu's bigger for example or by just using multiple cpu's on the same board. Many servers already use multiple cpus so why cant regular pc's do that ? I guess the only problem is that most games or programs just dont support multiple cpu's.