Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/intel-has-serious-problems-with-10nm-yields.12223/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Intel has serious problems with 10nm yields

lefty

Active member
Some interesting facts emerged during a recent Susquehanna conference with Charlie Demerjian. We all know how badly 10nm was delayed, but Charlie maintains that even now the yields are not at production levels and probably never will be. For this reason, Charlie maintains that 10nm laptops will only make up a tiny percentage of total sales (much less than 5%) for the next couple of years.
Intel had 3 major problems with 10nm: COAG, cobalt and 36nm metal pitch. COAG was the reason that the Cannon lake CPU did not have a functioning iGPU. Intel had to ultimately remove COAG. The second problem was cobalt, which Intel were able to fix somehow. The third problem was that the 36nm metal pitch requires SAQP and is hard to achieve without using EUV. TSMC's 7nm, for example, only has a pitch of 40nm. I'm am guessing that it's this aggressive pitch is what is causing their current yield problems.
 
Last edited:
Charlie demerijan isn't a credible. He was claiming layoffs were imminent that never happened.

Intel's a $70 billion company so they're doing alright. Long term 10 nm yields maybe a problem but it's a product with an excessive demand. How large will profits be will come down to if the 14 nm node will still be sellable in the future. Sources inside Intel had said they will most likely need to consolidate in the future. It isn't really a secret.
 
Charlie previously said Intel 10nm was cancelled, right? Yield is an interesting metric. It's very secretive so someone is breaking an NDA or they are outright lying (usually competitors or shorts).

Yield also changes from one month to the next so yield news to me is always questionable. I have been inside an NDA on many different projects where yield was "leaked" and it was never right. Seriously, I mean never.

The big yield tell is revenue/margin so at some point in time we will know the truth, like what happened with 14nm yield when BK choked on his words.

And you got this from an AMD stock chat on reddit? :eek:
 
There are no secrets for any length of time. Follow the money and this can be done many, many ways. Who is being hired, what equipment is being purchased, satellite views of parking lots, materials being purchased, careful reading of financial filings, freight shipments(also viewable by satellite or Google). A master strategist can put many disparate pieces together to get at what is really going on. Many of these people are masters of the game and very wealthy. I have also heard of drones being used. Two truckers I worked for who really knew how to work bars, got over on IBM and made a serious amount of money, one of my more interesting adventures that I personally had my hands in. I have met a few, but they almost never tip their hand.
 
Charlie previously said Intel 10nm was cancelled, right?
You pick the one thing he said was wrong and ignore the dozens of things he said that were correct.

And you got this from an AMD stock chat on reddit? :eek:
No, that is a transcript from Susquehanna conference with Charlie. Also, that reddit group breaks a lot of reliable info before the rest of the internet knows anything about it.

Well, many people decide they don't like Charlie, because of his abrasive attitude, but he has reliable sources in Intel and OEM partners and has been uncannily accurate in the past.
 
You pick the one thing he said was wrong and ignore the dozens of things he said that were correct.

No, that is a transcript from Susquehanna conference with Charlie. Also, that reddit group breaks a lot of reliable info before the rest of the internet knows anything about it.

Well, many people decide they don't like Charlie, because of his abrasive attitude, but he has reliable sources in Intel and OEM partners and has been uncannily accurate in the past.

Hi Lefty,

I have a bit of history with Charlie starting before SemiWiki.com. In fact, Charlie is one of the reasons I started SemiWiki and I even used the same forum software he did. This was before Charlie and all of the other forums (AnandTech) started blogging. We all used the same forum software and added blogging. I started blogging in 2009 when the whole TSMC 40nm yield problems started. I worked closely with ATI on a 40nm project so I actually knew what the 40nm GPU yield problem was first hand.

Charlie has broken some good stories, he has also misinformed people but he does not talk about that. To put it bluntly, he throws things against the wall to see what sticks. When he is right the world will know but when he is wrong you will not hear about it. I only mention the Intel 10nm canceling story because we are talking about Intel 10nm. There have been many others in regards to yield and the foundries.

And to imply that Charlie has more reliable sources at Intel than I do is incorrect. Please remember that most of us at SemiWiki are working semiconductor professionals so our "reliable sources" are actually people we work with. I also have the SemiWiki advantage. When I publish something on SemiWiki that is wrong I hear about it from our audience (publicly but mostly private).

Not a big deal, just my experience, but one thing I learned from Charlie is that when I am wrong it is best to admit it rather than bury it. I have seen Charlie at conferences and receptions over the years. The first time was at CES is Las Vegas at the GlobalFoundries party. My first impression was that he is a beady eyed little weasel. Now I just find him disingenuous. Here is the blog from the party. Be kind, it was one of the first blogs I posted on SemiWiki:

https://semiwiki.com/semiconductor-manufacturers/globalfoundries/320-ces-globalfoundries-party-pics/

Interesting to note, Jason Gorss, Technology Communications Manager for GlobalFoundries pictured in the blog, now works for Intel as do many other former GF employees so it really is a small semiconductor world.

Back to Intel 10nm, to be fair both Charlie and I have been wrong about Intel 10nm. Me on the plus side him on the minus side. The full story will be out at some point in time and I will definitely write about it. Until then let's just agree to disagree that Intel has "serious 10nm yield problems".

D.A.N.
 
Last edited:
> And to imply that Charlie has more reliable sources at Intel than I do is incorrect.
No, I did not imply that!

Anyway, I duly note what you say - that Intel has no yield problems - and I believe you have good sources. Yet, there is something strange going on, because Intel is releasing a lot of 14nm products throughout 2020 - comet lake, rocket lake. With the switch from 22nm to 14nm nothing like that happened. Also, this year they are releasing 2 server products almost simultaneously, 14nm cooper lake and 10nm Ice lake. They have never done this before. Poor yielding 10nm would be a possible explanation.
 
> And to imply that Charlie has more reliable sources at Intel than I do is incorrect.
No, I did not imply that!

Anyway, I duly note what you say - that Intel has no yield problems - and I believe you have good sources. Yet, there is something strange going on, because Intel is releasing a lot of 14nm products throughout 2020 - comet lake, rocket lake. With the switch from 22nm to 14nm nothing like that happened. Also, this year they are releasing 2 server products almost simultaneously, 14nm cooper lake and 10nm Ice lake. They have never done this before. Poor yielding 10nm would be a possible explanation.

Well either Charlie is lying or Bob Swan and the folks at IEDM last month are lying. I certainly hope it's Charlie.

I don't track Intel products, I stay at the process level. My current position is that I believe Intel when they say 10nm yield problems have past and they are ramping up production. Saying "Intel has never done this before" does not carry a lot of weight with me since Intel has been in uncharted waters since the 10nm debacle.

The next big conference is SPIE Advanced Lithography at the end of February so we will do another Intel yield honor check then.
 
Intel can sell as many at 10nm as they can make and I assume they can make a lot.

It's selling at the 14 nm node that will be harder.
 
Charlie Demerjian is fantasizing about Iranian terrorist attacking Trump property.


"Anyone want to bet that #Iran's response to Trump is to go after Trump properties worldwide? Trump ignored the emoluments clause, now is he going to ask for government resources to defend his private properties? It's going to be fun. "

This iranian "general" killed 600 servicemen, it's a good thing he's dead.
 
I believe that Charlie Demerjian can be just another internet personna saying whatever he wants. Writing words on the keyboard and blogging is free, right? So the whole story here should not be about Charlie or not. It should be about Intel and whether its responses demonstrate a coherent story with its marketing announcements. Unfortunately, the whole 10nm story, from its beginning, is so problematic for Intel, that I believe it will take years to undo the damage that has been been done. If it was not a company with the size (and pockets) of Intel and a company with a product portfolio that cannot just be replaced from one day to the other, we would be having a totally different discussion here.

I think it is wrong to write something like "we track and discuss processes here, not products" because Intel is an IDM and not a foundry. So, the products that are released, tell the story of the processes that they are made. At this moment, we dont have any of the highest performance systems yet made from 10nm, we still dont have desktop parts, we still dont have server parts and unfortunately the mobile parts that we have is a mixed bag, since they are limited and 14nm parts are released along with 10nm parts (I was quite surprised that the latest highest performance 10th gen i7 mobile processor announced just in Q3 19 is not made in 10nm but 14nm. This seemed very bizarre to me as the 10nm parts were supposed to be the ones that brought higher core counts and performance and yet the only 6 core part is 14nm and the highest performance (clock) 4 core part is again 14nm : https://ark.intel.com/content/www/u...10th-generation-intel-core-i7-processors.html ).

Anyway, let's give the benefit of doubt to Intel and see what will be announced in the next couple of months. This will be the strongest indication of whether things are actually resolved with the 10nm process or issues remain.
 
Back
Top