Intel's Cannonlake CPUs will use the companies 14nm process

[WYP]

Intel's Cannonlake CPUs will use the companies 14nm process, breaking their new Process, Architecture, Optimisation model after just one generation of products.



Read more on Intel breaking their Process, Architecture, Optimisation model with Cannonlake.

 

[SpencerUk]

Sound's like Intel are more than just a little concerned over AMD's incoming products given the press releases from them of late.

 

[Avet]

Maybe they know something we don't...

 

[RobM]

small increase again

 

[SeekaX]

That cycle was going to break at some point anyways, we probably won't go much smaller than 10nm.

 

[WYP]

That cycle was going to break at some point anyways, we probably won't go much smaller than 10nm.

Aye, nobody seems optimistic about going under 7nm, seems to be where the limit is at.

Even then it seems increasingly difficult to get to that point. AMD, Nvidia and Intel will need to work on their architectures more than usual, as they can't rely on process shrinks for performance gains.

 

[NeverBackDown]

Unless it's a new architecture it won't be much better.
They probably stalled 10nm due to capacity limit's

 

[AngryGoldfish]

Do they really rely on process shrinks that much? Obviously after a while you have to shrink the die to make any tangible gains, but Nvidia managed to squeeze a lot of efficiency out of 28nm. 28nm may have outstayed its welcome by then, but it still worked out, at least for them.

 

[AlienALX]

Do they really rely on process shrinks that much?

Yup. It's a free lunch (launch) dude. All you do is take your existing tech, shrink it, clock it higher due to lower volts etc and then sell it as a brand new range (see also Nvidia 10 series, Maxwell on speed).

They will continue doing it for as long as they can because a brand new tech is expensive and takes years. Look how long it's taken AMD to do Zen. About five years now?

 

[NeverBackDown]

Yup. It's a free lunch (launch) dude. All you do is take your existing tech, shrink it, clock it higher due to lower volts etc and then sell it as a brand new range (see also Nvidia 10 series, Maxwell on speed).

They will continue doing it for as long as they can because a brand new tech is expensive and takes years. Look how long it's taken AMD to do Zen. About five years now?

Doesn't always work that way. When moving to a smaller node, you have really two basic options if you don't change your architecture much. You can 1) increase clocks slightly(emphasis on slightly) and in turn since everything is smaller(aka more transistor/sq) and can switch faster, that will improve performance. 2) Or you can keep the same performance, but take advantage of the fact that you have more transistor density, which means you can have the same amount as before but in a smaller area, which decreases overall power consumption.
And no, Pascal moved to a FinFet process. Which is much better at everything compared to the old planar process. It's not really comparable besides the fact it's nearly the same thing but clocked higher and more efficient. But that's only due to FinFet. This would be a classic example of what you are talking about. Smaller node, better performance, enhanced further because they changed the underlying architecture there's is built upon from the start. Which leads to improvements in every area and combines 1 and 2 I mentioned before

 

[Tripp]

Aye, nobody seems optimistic about going under 7nm, seems to be where the limit is at.

Even then it seems increasingly difficult to get to that point. AMD, Nvidia and Intel will need to work on their architectures more than usual, as they can't rely on process shrinks for performance gains.

Personally I think that's more due to money tbh, and the cost to do 7nm and below and yield that would be had,I don't think it would be cost effective.... and that's why they're not optimistic about it.

On another note I watch a compelling video talking about the troubles at Intel at the moment, and what was said had good footing for the opinions, with the fact of lay offs and top executives jumping ship ect. Intel have been caught off guard in my opinion (based off what info and conjecture I have at hand), and have become some complacent. They have let there guard down and in turn given AMD the opportunity to make a palatable comeback, and are scared of the market share they're going to loose.

I think this is nothing but good for us consumers, an wish AMD all the best in the near future

 

[NeverBackDown]

It's not that cost is preventing below 7nm. It's physics. 7nm is the farthest we can go with silicon. After that weird things start to happen and you can no longer predict what the atoms are going to do, aka no longer process anything.. you start getting into quantum physics. That's why other technology is being researched to start on a different atom(s).

 

[Zoot]

Well... one can be sure that it's now unlikely Cannonlake will be delayed.