10nm CPUs to not launch, replaced by more 14nm!

W

Weeth

Gawd
Joined
Sep 7, 2011
Messages
662
No Cannonlake

Wow. There must be some serious rumblings behind closed doors at Intel. I would bet my bottom nickel that they have run into a brick yield wall and that those problems are definitely tied in to the LGA Broadwell vapor launch!
 
Wow, what a lazy, wrong and hyperbolic story.

The original source: http://benchlife.info/cannon-lake-postpone-and-kaby-lake-will-replace-skylake-in-2016-06232015/

The information shows that 14nm Kaby Lake is coming in 2016. That's it. The author speculates that means Canon Lake is delayed, but offers nothing to support it. Ridiculous sites are taking that speculation as "Canon Lake is indefinitely delayed". Idiotic, as usual.

There's an investor conference call in less than 4 weeks that would confirm or deny any delays, and IDF is 2 months away if anything changes after next month.
 
A lot of influential tech sites have picked up the "rumor" so only time (or an official Intel statement) will tell the tale. I'd love to be on that investor conference call and ask "what happened to the LGA Broadwell launch?"
 
Weeth said:
A lot of influential tech sites have picked up the "rumor" so only time (or an official Intel statement) will tell the tale. I'd love to be on that investor conference call and ask "what happened to the LGA Broadwell launch?"
Click to expand...

I don't know what influential sites you are talking about, but the only rumors I put credence to are the ones picked up and talked about on Anand's in regards to Intel.

Rumor could be right or not, but right now I have no reason to believe it until the conference in 4 weeks.
 
Digital Trends
Extreme Tech
Tech Spot
Softpedia News
Tech Report
bit-tech
TweakTown
CPU-World

Ok, not all of these are sterling reputation sites and not all are calling it doomsday for Cannonlake, but the "rumor" is well spread. I don't see anything on Anand yet, and that may very well substantiate your belief in that site's credibility. I've always found Anand to be considerably more authoritative than Tom's for example. BTW, Tom's is silent on this as well. So the whole thing very well may just be a misunderstanding mixed in with a healthy helping of hyperbole.
 
Weeth said:
Digital Trends
Extreme Tech
Tech Spot
Softpedia News
Tech Report
bit-tech
TweakTown
CPU-World

Ok, not all of these are sterling reputation sites and not all are calling it doomsday for Cannonlake, but the "rumor" is well spread. I don't see anything on Anand yet, and that may very well substantiate your belief in that site's credibility. I've always found Anand to be considerably more authoritative than Tom's for example. BTW, Tom's is silent on this as well. So the whole thing very well may just be a misunderstanding mixed in with a healthy helping of hyperbole.
Click to expand...

You know, I didn't even think to check Tom's, I'll keep on eye there to see if they pick it up as well. I'll stick to my plan with Broadwell for now, if this rumor is indeed true I'll just scrap my plans.
 
Trimlock said:
You know, I didn't even think to check Tom's, I'll keep on eye there to see if they pick it up as well. I'll stick to my plan with Broadwell for now, if this rumor is indeed true I'll just scrap my plans.
Click to expand...

It's almost midnight Eastern time and both Tom's and Anand have no info on the Cannonlake situation at all. It will be interesting to see what if anything they have to say in the morning. I'm hoping that it is all a big mistake since I would really hate to see CPU manufacturing hit a wall at 14nm!
 
Weeth said:
It's almost midnight Eastern time and both Tom's and Anand have no info on the Cannonlake situation at all. It will be interesting to see what if anything they have to say in the morning. I'm hoping that it is all a big mistake since I would really hate to see CPU manufacturing hit a wall at 14nm!
Click to expand...

Its more than likely an extrapolation from some information released a few days ago, causing the rumor sites to run wild with it.
 
they do have to run into a wall eventually...I mean...one would think

how many atoms does it take to make a transistor?
 
Trimlock said:
Technically all you need is 1.
Click to expand...

I would think at least 3, NPN or PNP doping

plus one to bond those 3 to and hold them in place, that may take more, I'm not a chemist

a few more to connect the 3 leads out to the rest of the circuit, I'm going to guess 7 as a theoretical minimum, and I doubt we ever achieve that on the material plane

edit/ hmmm, I see what you mean http://www.purdue.edu/newsroom/research/2012/120219KlimeckAtom.html, but that doesn't seem anywhere close
 
Weeth said:
Ok, not all of these are sterling reputation sites and not all are calling it doomsday for Cannonlake, but the "rumor" is well spread. I don't see anything on Anand yet, and that may very well substantiate your belief in that site's credibility.
Click to expand...
That's really the problem with rumors in general. It gets repeated broken telephone game style until no one is aware of the original source of the rumor, or what it actually said. Compounding the problem in this case is that it's in Mandarin, so the people repeating it are unlikely to even read a lazy google translation.

There is one source, and the only information it has is that another 14nm processor is coming in 2016. The author speculates that means Canon Lake is delayed. That's not unreasonable speculation, but it's still mere speculation from someone apparently not directly involved in knowing.

Where people are getting any kind of idea that there's a source actually stating Canon Lake is delayed or 10nm is indefinitely on hold is basing that on nothing. As I mentioned above, there's an investor conference call in a few weeks and that's a topic that will certainly come up, and IDF is in 2 months. It's not like people are going to be blindsided by any delay since there is ISSCC, 2 IDFs and 4 investor conference calls every year. :p

If I can make my own wild speculation, 10nm is a pivotal process. The upcoming tock on 10nm will have a big problem for the follow up shrink at 7nm. Given how things have shaken up at 10nm (Canon Lake replaced Skymont, now Kaby Lake is following Skylake and a possible delay for Canon Lake), there may have been strategy changes for upcoming architectures causing potential delays. New materials may be integrated at 10nm to smooth the transition and development. That and who knows what else are delay risks.
 
jojo69 said:
they do have to run into a wall eventually...I mean...one would think

how many atoms does it take to make a transistor?
Click to expand...

That's true, but I don't think the wall is at 10nm. Not that I'm an expert of course, just base off what I've seen. Some time ago, TSMC announces that they will begin 10nm in 2016, with 7nm planned for the future. I think the industry seem confident with 10nm thus far.
 
jojo69 said:
I would think at least 3, NPN or PNP doping plus one to bond those 3 to and hold them in place, that may take more, I'm not a chemist a few more to connect the 3 leads out to the rest of the circuit, I'm going to guess 7 as a theoretical minimum, and I doubt we ever achieve that on the material plane edit/ hmmm, I see what you mean http://www.purdue.edu/newsroom/research/2012/120219KlimeckAtom.html, but that doesn't seem anywhere close
Click to expand...

If you are talking about transistors operating in absolute zero single atoms are OK. For normal temperature ranges one need many atoms to deal with thermal noise. Several nanometers is the limit then.

But even more serious is economic limit. The push for smaller transistors is motivated by lowering cost of producing devices. This is just now approaching the limit, it is said there is no real gain in going to 10nm from 14nm. Thus 10nm may be the last ore one before the last generation.
 
I read that Intel has plans down to 6nm, beyond that is still sci fi
 
Weeth said:
No Cannonlake

Wow. There must be some serious rumblings behind closed doors at Intel. I would bet my bottom nickel that they have run into a brick yield wall and that those problems are definitely tied in to the LGA Broadwell vapor launch!
Click to expand...

When you don't have competition, why improve?
 
I'm posting this from the year 2043. We're still using 14nm. It's over guys. Enjoy what you're using now, because we're still using Skylake processors. In fact, I have a friend who keeps saying that there's no reason to upgrade from Sandy Bridge -- for the past 30 years.
 
If Intel is delayed by 1 year that means TSMC is delayed by 5. Watch them spin their fables lol.

Anywho, at this point new nodes don't really matter. We aren't getting massive performance boosts anyways to get our e-peens all worked up over the latest and greatest.
 
notarat said:
When you don't have competition, why improve?
Click to expand...
I think there's plenty of competition in the low power SoC segment, and smaller process along with more efficient architecture will help Intel compete there.
 
notarat said:
When you don't have competition, why improve?
Click to expand...

Those who have a processor wont buy another if there is no point.
The need for progression stalls and a large market disintegrates.
 
My best guess is that doing N multi-patterning where N > X isn't economically viable any longer and they're waiting for EUVL to get its shit together.
 
Liger88 said:
Can't honestly think it's worse than a chip named "Braswell" lol.
Click to expand...

2moa79x.jpg
 
Nenu said:
Those who have a processor wont buy another if there is no point.
The need for progression stalls and a large market disintegrates.
Click to expand...

Ive been on my 2600k since they came out in what, Jan 2011? I have never stayed on 1 CPU for longer than 2 years..... Will most likely be nearly 5 or more years before I upgrade too.

While there is no competition from AMD, there is also no reason to upgrade.
 
doz said:
Ive been on my 2600k since they came out in what, Jan 2011? I have never stayed on 1 CPU for longer than 2 years..... Will most likely be nearly 5 or more years before I upgrade too.

While there is no competition from AMD, there is also no reason to upgrade.
Click to expand...

Yep, when progression is slow, game devs dont push the boat out.
At the moment its not like GPUs where you can add more cores and get better CPU performance, so they are reliant on significantly faster CPUs coming before they increase the workload.

DX12 should help the situation and as long as CPUs with more cores and high enough IPC/clock speed are affordable, we will see higher CPU usage and a forward moving market.
AMD might be able to compete too.
 
As soon as I get some $$ in about 6 months I will be upgrading. Not because I want to but because I have to. My hardware is slowly failing and the only thing I haven't replaced is the motherboard, CPU and GPU (and I am fairly certain that the mobo is bad and it is a socket 1366 and you know how expensive the old used ones are going for...) And I used to be one of those people who upgraded every year too.
 
I'm still on the 2600K, and I want to upgrade if Intel can bring something great out. I don't care if it's 14nm or 10nm. Just want the performance to be much better than it is.

Of course, with the stagnant CPU market lately, I've been using the upgrade funds for other projects.... :)
 
Lol, I was wondering why TSMC was still making 28nm chips in 2015. I remember them touting 20nm or so a few years back.

I'm still running on Westmere i7-980...

I'm more interested in the ramifications for mobile chips though (if true).
 
Ur_Mom said:
I'm still on the 2600K, and I want to upgrade if Intel can bring something great out. I don't care if it's 14nm or 10nm. Just want the performance to be much better than it is.

Of course, with the stagnant CPU market lately, I've been using the upgrade funds for other projects.... :)
Click to expand...

My 2nd system is 2600K and if I have to be perfectly honest there is absolutely no need for a PC that is more powerful or newer for 99.5% of all the things I do on it. My 5960X is total overkill for the vast majority of the uses I apply it to and again to be perfectly honest, sometimes I wish I hadn't bought it and spent that big wad of money on more beer! :D
 
Weeth said:
My 2nd system is 2600K and if I have to be perfectly honest there is absolutely no need for a PC that is more powerful or newer for 99.5% of all the things I do on it. My 5960X is total overkill for the vast majority of the uses I apply it to and again to be perfectly honest, sometimes I wish I hadn't bought it and spent that big wad of money on more beer! :D
Click to expand...

That's funny as one of my other projects was getting into home brew beer. :)
 
Weeth said:
My 2nd system is 2600K and if I have to be perfectly honest there is absolutely no need for a PC that is more powerful or newer for 99.5% of all the things I do on it. My 5960X is total overkill for the vast majority of the uses I apply it to and again to be perfectly honest, sometimes I wish I hadn't bought it and spent that big wad of money on more beer! :D
Click to expand...

As you grow up and get older you start to realize that the less often you upgrade your cpu, the more significant the gains become between upgrades.
 
ZodaEX said:
As you grow up and get older you start to realize that the less often you upgrade your cpu, the more significant the gains become between upgrades.
Click to expand...

Er... I'm 58 years old... but I do see what you're saying! :)
 
doz said:
Ive been on my 2600k since they came out in what, Jan 2011? I have never stayed on 1 CPU for longer than 2 years..... Will most likely be nearly 5 or more years before I upgrade too.

While there is no competition from AMD, there is also no reason to upgrade.
Click to expand...

Same here. Longest CPU cycle for me EVER. As speculated I think we have 1, maybe 2 shrinks left before silicon is played out. That's a frightening thought. We may see a period of relative stagnation - kinda like we've seen with TSMC on the GPU front, with their failure to launch a viable high performance node smaller than 28nm year after year after year, leading to more and more designs on the same process node.

Hopefully after the 'end of silicon' we will come out the other side with a new paradigm that's not silicon transistor based, and it's a return to Moore's law for a while while people figure out how to optimize the new paradigm (talking years or even decades out, of course)
 
You must log in or register to reply here.
Back
Top