bit-tech.net

Rumours point to problems with Intel's 10nm node

Rumours point to problems with Intel's 10nm node

Intel's 10nm Cannon Lake chips are rumoured to have run into trouble, and are being replaced by 14nm Kaby Lake parts in leaked documentation.

Intel is rumoured to have encountered further difficulties with the shrink to a 10nm process node, with leaked documentation appearing to show a new 14nm node - Kaby Lake - being launched where Cannon Lake should have been.

Intel runs a tick-tock development methodology, where it alternates between switching to a smaller process node and introducing a new microarchitecture. Each brings its own improvements, but the ever-shrinking process node is the most challenging: as component sizes get smaller, the laws of physics begin to get in the way. Back in 2013 Intel admitted that things were getting harder and that following Moore's Law - the observation turned commandment by Intel co-founder Gordon Moore that the number of transistors on a integrated circuit doubles roughly every eighteen months - was becoming a challenge and that the 14nm Broadwell parts were necessarily delayed.

Now documents claiming to be leaked from Intel suggest that the troubles of shifting to 14nm are rearing their heads again at 10nm. According to product listings released by Benchlife this week, Cannon Lake - the 10nm successor to 14nm Skylake - has been bumped in favour of a previously unannounced chip family dubbed Kaby Lake. While these parts would fit in the portion of the release cycle which should mean a process node shrink, they are claimed to instead be based on a 14nm process node - the first time Intel has deviated from its traditional tick-tock schedule, and a sign that the continued shrinkage of process nodes may be coming to an end.

Intel, as is to be expected, has refused to comment on the alleged leaks.

9 Comments

Discuss in the forums Reply
Corky42 25th June 2015, 12:37 Quote
Question if I may: From what I've read via my Google-fu quantum tunnelling starts to occur around 3nm, the accepted way of describing CPUs are 14nm, 10nm, etc, etc, that (afaik) is only the transistor size, or is it the gap?

Anyway even though a CPU is described as 10nm are there actually smaller components inside the CPU, smaller enough to experience quantum tunnelling?
Hustler 25th June 2015, 12:49 Quote
Hmm, these 'Kaby Lake' have a 'Broadwell' like smell about them.
Alexg 25th June 2015, 13:37 Quote
In reply to Corky42 - Yes there are smaller parts, 14nm is the transistor size. The tunneling is a particular issue for the oxide layers used to provide insulation on the gate of the transistor. One of the reasons silicon is good chips is it has a solid stable oxide however these layers are now so thin that electrons can tunnel across them and leak out of the chip, essentially resulting in it going from charged to uncharged or 0 - 1 over time. The silicon-dioxide is being replaced by alternative materials to help mitigate this but it becomes more and more of an issue and harder to fix as the chip gets smaller.
wolfticket 25th June 2015, 13:57 Quote
Quote:
Originally Posted by Hustler
Hmm, these 'Kaby Lake' have a 'Broadwell' like smell about them.
Fraudwell™
silk186 25th June 2015, 14:33 Quote
don't we hear the same thing with every node?
play_boy_2000 25th June 2015, 18:05 Quote
I just wish tsmc or glofo could figure out something sub-20nm, to move the rest of the semicon industry along.
jrs77 26th June 2015, 16:49 Quote
If we look at the problems they allready have with the current 14nm process pushing out decent yields, than it's no surprise really that the yet smaller process has it's problems.
amagriva 9th August 2015, 18:08 Quote
Quote:
Originally Posted by Hustler
Hmm, these 'Kaby Lake' have a 'Broadwell' like smell about them.
Fraudwell™

Zap!well anyone?
rollo 10th August 2015, 10:46 Quote
Id imagine Intel are fast approaching the point where Cost to make will outwiegh the benefits if we are not already at that point.

If it costs them several $bil to develop and sort 10nm they have to make back at least twice that for it to be worthwhile.

Skylake is a nice chip and has good stock improvements over Sandybridge but are they enough for people to spend the cash upgrading. Id personally just overclock it a bit and get similar bang for buck.
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums