[SemiAccurate] Intel kills off the 10nm process

[Kenmitch]

Seeing how Intel is godlike I don't believe it one bit. /s

 

[moinmoin]

"the 10nm process as proposed by Intel"
Even without reading past the paywall I'm sure this is the most important qualifier. The node, being in development, was being tweaked all the time anyway. If business planning at Intel is not a complete mess they started some alternative developments to fall back on, one of which may get the "10nm" name tag in the public while the original development gets the "knifing" internally. If they do make any cancellation public they'll announce it as a "necessary" part of "technological progress" and give the replacing node a "7nm" name tag or some such.

 

[IntelUser2000]

"the 10nm process as proposed by Intel"
Even without reading past the paywall I'm sure this is the most important qualifier.

Sounds like a logical conclusion.

They should have done this the first time. They had this idiotic focus on density, which is likely why have this mess. Their main Core line doesn't benefit from the density anyway. It's always for the others like with Small(Atom) cores and maybe FPGA.

They sacrificed bread-and-butter to chase a dream. It's usually how companies end up dying. You have a functional main business, but a serious contender arises. Then the management announces a plan to address the competition. They put too much resources into doing that and make brain-dead decisions that hurt their main business. Business decline accelerates, resulting in more brain-dead decisions.

 

[jpiniero]

They should have done this the first time. They had this idiotic focus on density, which is likely why have this mess. Their main Core line doesn't benefit from the density anyway.

It greatly impacts margins, which is a big deal.

 

[IntelUser2000]

It greatly impacts margins, which is a big deal.

*Only* for non-Core lines. Do you not read the whole post? It's a bit rude. Only Airmont gained the density advantages claimed for 14nm, not Broadwell.

 

[arandomguy]

I wonder if for future plans they will officially move to less aggressive node jumps with more iterative follow up improvements similar both TSMC and Samsung. They're already somewhat doing this with 14nm, effectively a process-->optimization->optimization cycle.

It's like that proverb about how slow and steady winning out.

 

[NTMBK]

Sounds like a logical conclusion.

They should have done this the first time. They had this idiotic focus on density, which is likely why have this mess. Their main Core line doesn't benefit from the density anyway. It's always for the others like with Small(Atom) cores and maybe FPGA.

They sacrificed bread-and-butter to chase a dream. It's usually how companies end up dying. You have a functional main business, but a serious contender arises. Then the management announces a plan to address the competition. They put too much resources into doing that and make brain-dead decisions that hurt their main business. Business decline accelerates, resulting in more brain-dead decisions.

Except Intel's "bread and butter" also includes extremely lucrative server chips. Higher density = more cores per chip. Without higher density they're stuck doing dumb stuff like 2x28 core dies on 14nm...

 

[IntelUser2000]

Except Intel's "bread and butter" also includes extremely lucrative server chips. Higher density = more cores per chip. Without higher density they're stuck doing dumb stuff like 2x28 core dies on 14nm...

You are not reading what I said as a whole either. Their *Core* line did not benefit from the 14nm density claims. The decrease in size was in line with predecessors, not dramatically better(as was with Airmont).

 

[LTC8K6]

"the 10nm process as proposed by Intel"
Even without reading past the paywall I'm sure this is the most important qualifier. The node, being in development, was being tweaked all the time anyway. If business planning at Intel is not a complete mess they started some alternative developments to fall back on, one of which may get the "10nm" name tag in the public while the original development gets the "knifing" internally. If they do make any cancellation public they'll announce it as a "necessary" part of "technological progress" and give the replacing node a "7nm" name tag or some such.

So, as suggested a couple times before, they might have gone to 10nm+ or ++, and now they might just call it 7nm?

What they may have killed is the original 10nm node, after producing that one mobile chip?

 

[IntelUser2000]

Their Core line(courtesy of Hans de Vries and Etain05): https://forums.anandtech.com/thread...p-cpus-anandtech.2554913/page-5#post-39604243

In line with 50-55% size in a Tick core. Go analyze it far back if you wish. Check Penryn vs Merom. Westmere vs Nehalem. Ivy Bridge vs Sandy Bridge.

Their Atom line: https://www.hardware.fr/news/14169/airmont-64-plus-petit-que-silvermont.html

A much greater 64% reduction, a density increase of 2.78x.

 

[Kenmitch]

Except Intel's "bread and butter" also includes extremely lucrative server chips. Higher density = more cores per chip. Without higher density they're stuck doing dumb stuff like 2x28 core dies on 14nm...

Higher density also means more heat in a smaller place. Not like the high end chips run cool enough to throw in some extra cores without paying the piper.

 

[Andrei.]

People, people.

If Intel's struggling immensely on the 10nm process, do you think they'll be better off skipping to a 7nm process? Technology improvements are iterative.

They are not. You build a PDK for a given process. If that doesn't work, you scrap it because the changes to make it work will no longer be compatible, and you need a new PDK.

Intel's purported changes to 10nm are big enough to warrant a new PDK, *if* they are recycling it, and certainly so if they are moving onto what was meant to be 7nm in the pipeline.

 

[Ottonomous]

They restructured (more segmented ) their tech, manufacturing and supply, restated holiday 2019. Its looking like a strong 12/14nm push to then

 

[rbk123]

Another reason why they are increasing their 14nm capacity - to carry them over until they get either 10 or 7 actually working. If this is true, it just means it'll be later rather than sooner, which makes sense since increasing 14nm capacity will also take a year or so to start reaping benefits.

 

[zinfamous]

If this is true, then their official announcement, only a few weeks/month ago, of "10nm being right on time" seems rather suspicious. I'd wait until we get actual word from Intel.

 

[Dayman1225]

Well, Intel has now publicly denied this via their Twitter:

Media reports published today that Intel is ending work on the 10nm process are untrue. We are making good progress on 10nm. Yields are improving consistent with the timeline we shared during our last earnings report

 

[DanNeely]

People, people.

If Intel's struggling immensely on the 10nm process, do you think they'll be better off skipping to a 7nm process? Technology improvements are iterative.

Intel's 10nm was allegedly using a number of quad patterning steps to get feature sizes that Samsung/TSMC's roadmap had as not doable without EUV and that it was the quad patterning that never worked out as planned. The same rumors also said that for their 7nm process, Intel would replace all of those quad pattern steps with EUV. If so, by getting rid of the quad patterning steps 7nm would be free of what broke 10nm.

The biggest caveat with all of this is that SemiAccurate's kept any details they've learned about Intel's troubles behind their $1k paywall, and the only source I'd seen outside of SA that had any details about what was supposed to be wrong is one that's known to report relatively sketchy quality rumors at times.

https://wccftech.com/analysis-about-intels-10nm-process/

 

[jpiniero]

Well, Intel has now publicly denied this via their Twitter:

Of course nothing would stop them from calling another 14 nm revision 10 nm, with or without any kind of shrink. Without seeing the whole details from SA, yeah, it's a bit murky.

 

[AkulaMD]

People, people.

If Intel's struggling immensely on the 10nm process, do you think they'll be better off skipping to a 7nm process? Technology improvements are iterative.

If they indeed skipping to "7nm", perhaps it is just a rename of the 10nm to justify another delay. Any possibility?

 

[LTC8K6]

If they indeed skipping to "7nm", perhaps it is just a rename of the 10nm to justify another delay. Any possibility?

Since they are not skipping 10nm, no.

 

[AkulaMD]

Since they are not skipping 10nm, no.

Or perhaps they could just call another iteration of 14nm as the 10nm?

 

[PotatoWithEarsOnSide]

They did release 10nm on time; that 8123u or whatever it was called. A CPU that only one OEM took, for kickbacks of course, which was the worst performing CPU produced in the past 5 years. 10nm is working exactly as intended.

If this is true, then their official announcement, only a few weeks/month ago, of "10nm being right on time" seems rather suspicious. I'd wait until we get actual word from Intel.

 

[.vodka]

Or perhaps they could just call another iteration of 14nm as the 10nm?

People like techinsights could quickly take one of these hypothetical chips apart and do their measurements only to find parameters that match a 14nm derivative, not a 10nm class process.

It would backfire spectacularly.

We'll see what happens to the 10nm meme in three days when Q3 results are published and Intel gets to answer questions from their faithful investors on that matter, once again.

 

[jpiniero]

People like techinsights could quickly take one of these hypothetical chips apart and do their measurements only to find parameters that match a 14nm derivative, not a 10nm class process.

That's true, but also that's a year+ from now. I doubt they will admit anything of that sort during the earnings call, but will later at some point.

 

[Spartak]

Semi-accurate has never been fully accurate hence the name. Burying it behind such a massive paywall enables them to stir up the hype without the need to deliver the meat. Those silly 5 investors won't share that 'knowledge' or make a dent in the stock market.

FWIW people speculating about skipping 10nm for 7nm clearly have no clue about the consequences and the ambitious specifications of 7nm. It's more dense then TSMC's 3nm process based on what is known right now about both processes.

If you can't jump 5ft you won't solve that by jumping 10ft.

What is likely at play here is Intel relaxing the m2 pitch from 36 to 40 or the m1/2 from 40/36 to both 44 to avoid the quad patterning of those two metal pitches. Depending on what their double patterning can handle it will either be 40 or 44nm. TSMC can handle 40 on their 7nm process but it's really unclear why Intel went with 36 in the first place.

Charlie sells that as 'dropping 10nm' which is simply hyperbole from an attention whore.