[SemiAccurate] Intel kills off the 10nm process

[TheELF]

Intel now has to sell bigger chips to satisfy users. Xeons with dual 28 core dies, 8 core desktops, quad core ultrabooks. Everything needs a bigger chip, which means fewer chips per wafer, which means more wafers, which means tighter margins and a capacity crunch.

Yeah but with 10nm, now that's gonna be a chip where 2 cores will be enough for anything....right?!Is that your argument here?
10nm is harder to make so it's more expensive and intel would still need to put the same amount of cores into each product,even with 40% more product per waffer I doubt the production cost would be cheaper for a few years.

 

[DrMrLordX]

And what exactly would they need MORE 14nm lines for?Only increased demand.
No company is stupid enough to switch to a new production that they aren't ready for,they had hoped to have 10nm ready by now so they switched everything to 10nm....because hope?!?!

What you're describing, actually happened. They also increased "demand" for 14nm by moving chipsets to 14nm (away from 22nm and older nodes) and agreeing to fab a bunch of modems for Apple. Low margin modems. Ooops. There were supposed to be working 10nm lines to take the pressure off the 14nm fabs. Coffeelake was never supposed to happen. Cascade Lake was never supposed to happen. Nor was Whiskey Lake.

You keep acting like demand is going up for Intel chips, but their sales numbers would indicate something different.

 

[epsilon84]

What you're describing, actually happened. They also increased "demand" for 14nm by moving chipsets to 14nm (away from 22nm and older nodes) and agreeing to fab a bunch of modems for Apple. Low margin modems. Ooops. There were supposed to be working 10nm lines to take the pressure off the 14nm fabs. Coffeelake was never supposed to happen. Cascade Lake was never supposed to happen. Nor was Whiskey Lake.

You keep acting like demand is going up for Intel chips, but their sales numbers would indicate something different.

Sales numbers? You mean revenue? It was actually a record 3Q for Intel in terms of revenue:
https://www.intc.com/investor-relat...d-Quarter-2018-Financial-Results/default.aspx

"Record quarterly revenue of $19.2 billion, up 19 percent year-over-year (YoY); data-centric* revenue grew 22 percent and PC-centric revenue grew 16 percent."

 

[Arachnotronic]

What you're describing, actually happened. They also increased "demand" for 14nm by moving chipsets to 14nm (away from 22nm and older nodes) and agreeing to fab a bunch of modems for Apple. Low margin modems. Ooops. There were supposed to be working 10nm lines to take the pressure off the 14nm fabs. Coffeelake was never supposed to happen. Cascade Lake was never supposed to happen. Nor was Whiskey Lake.

You keep acting like demand is going up for Intel chips, but their sales numbers would indicate something different.

Demand has been going up for Intel chips. They're guiding to $71.2B in sales this year, up from their original projection of $65B.

The sheer number of extra chips that Intel has to produce to meet that $6.2B extra in demand is nontrivial.

 

[Nothingness]

And what exactly would they need MORE 14nm lines for?Only increased demand.
No company is stupid enough to switch to a new production that they aren't ready for,they had hoped to have 10nm ready by now so they switched everything to 10nm....because hope?!?!

The increased demand would have been directed to 10nm products if they had been there. They're not there and that's why Intel has to increase 14nm production capacity. That's all I said. Not that demand hadn't increased, neither that they should have switched to a process that isn't delivering good yields. And if you don't get that planning fabs is done years before the process is ready, there's not much I can do for you.

 

[Nothingness]

Demand has been going up for Intel chips. They're guiding to $71.2B in sales this year, up from their original projection of $65B.

The sheer number of extra chips that Intel has to produce to meet that $6.2B extra in demand is nontrivial.

I'm convinced most of these extra chips should have been 10nm, and that if they had been their sales would have been even better due to enough capacity.

I wonder how much extra capacity Intel typically have. I can't believe they have just enough capacity to fill what they predict the market will ask for.

 

[epsilon84]

Demand has been going up for Intel chips. They're guiding to $71.2B in sales this year, up from their original projection of $65B.

The sheer number of extra chips that Intel has to produce to meet that $6.2B extra in demand is nontrivial.

That is true. $6.2B actually exceeds AMDs yearly revenue which was $5.3B for 2017. Just some perspective there.

Intel will have to endure the consequences of their 10nm mess, and that may yet show in future profit/revenue reports. But for the moment they are still raking in record sales despite being stuck at 14nm, so obviously the demand is still there, and is in fact causing them problems in terms of meeting that extra demand.

 

[DrMrLordX]

Demand has been going up for Intel chips. They're guiding to $71.2B in sales this year, up from their original projection of $65B.

The sheer number of extra chips that Intel has to produce to meet that $6.2B extra in demand is nontrivial.

And how many of those chips are replacements being shipped due to Meltdown/Spectre?

 

[Arachnotronic]

And how many of those chips are replacements being shipped due to Meltdown/Spectre?

Do you think Intel's strong Q1, Q2 and Q3 perfomance were due to shipping unreleased Cascade Lake chips? Or do you think the return to growth in the PC market is because your avg Joe is buying new laptops with Spectre/Meltdown mitigations?

 

[jpiniero]

Or do you think the return to growth in the PC market is because your avg Joe is buying new laptops with Spectre/Meltdown mitigations?

Corporate might be however.

 

[NostaSeronx]

What would be worse?

1. Intel killing 10-nm and skipping over it to 7-nm. (While, calling it 10-nm in its 193i form)
2. GlobalFoundries killing 14LPP/12LP and replacing it all with 12HPC/12HD bulk FinFETs. (While, having no FinFET node planned to succeeded them) ((Further, 12HPC and 12HD are derived from 14HP, but as stated are bulk nodes!))

 

[Arachnotronic]

Corporate might be however.

Was Whiskey Lake widely available in Q1? Q2? Are there even vPro SKUs of Whiskey Lake out now?

 

[jpiniero]

Was Whiskey Lake widely available in Q1? Q2? Are there even vPro SKUs of Whiskey Lake out now?

No on both. Maybe there was buying to get up to date to ensure they would get fixes at least?

 

[Markfw]

And what exactly would they need MORE 14nm lines for?Only increased demand.
No company is stupid enough to switch to a new production that they aren't ready for,they had hoped to have 10nm ready by now so they switched everything to 10nm....because hope?!?!

First, my comment is not directed at TheELF, but the context.

This is almost funny. The "No company is stupid enough" and "Intel" in the same sentence.

Yes, they were that stupid with 10nm

 

[Arachnotronic]

No on both. Maybe there was buying to get up to date to ensure they would get fixes at least?

That's a stretch.

 

[IntelUser2000]

Demand has been going up for Intel chips. They're guiding to $71.2B in sales this year, up from their original projection of $65B.

I wouldn't be so quick to say its entirely due to demand. They were able to increase ASPs by introducing new chips and moving more customers to higher end chips in addition to raising prices across the board.

In desktops particular, the volume dropped 5% but ASPs rose by an incredible 10%. I'd even have AMD being competitive bringing excitement to the enthusiast market having waterfall effects for the whole sector. It's not entirely due to that, but they were forced to improve by upping core counts at a rapid pace.

 

[jpiniero]

That's a stretch.

True. I could see people in IT using it as an excuse to upgrade antiquated laptops but not enough to get an 8% vol gain. The volume increase in the data center was definitely helped by Smeltdown tho.

 

[IntelUser2000]

Who cares about the volume and ASPs though, really. It doesn't affect us at all, unless you are a shareholder.

We should be solely focused on tech. I'm excited they decided to reveal their future plans. It's when companies keep quiet people should be worried.

Like when Apple told recently that they are no longer going to reveal unit sales in terms of volume. That's about the same time when suppliers heavily involved with Apple announced they had to cut forecasts because volume is going to decline. Also latest iPhone sales not doing so well.

I bet their revenue isn't going to suffer though. Lower volume but higher ASPs do that.

 

[DrMrLordX]

Do you think Intel's strong Q1, Q2 and Q3 perfomance were due to shipping unreleased Cascade Lake chips? Or do you think the return to growth in the PC market is because your avg Joe is buying new laptops with Spectre/Meltdown mitigations?

ODM buyers like Google and Amazon were getting Skylake-SP months earlier than anyone else. So if they had an "oops" moment and realized that they couldn't get max performance out of their Intel hardware anymore (remember, Google at least was on the cutting edge of testing fixes for Meltdown and Spectre), you gotta think they tapped the same channels for replacements. And I think of the three reported quarters for 2018, Q3 is the most likely one to reflect shipments of replacement parts. Possibly at a discount, but obviously not a very big one.

In any case I think we're seeing a lot of artificial pumping of 14nm "demand": people buying stuff because they have to (Intel won't sell them anything else). Nobody wants to see Intel continue releasing 14nm products. Sure, Intel is making money now, but very soon people are going to go sour grapes unless Intel makes major changes to their product lineup.

 

[Yotsugi]

ODM buyers like Google and Amazon were getting Skylake-SP months earlier than anyone else. So if they had an "oops" moment and realized that they couldn't get max performance out of their Intel hardware anymore (remember, Google at least was on the cutting edge of testing fixes for Meltdown and Spectre), you gotta think they tapped the same channels for replacements. And I think of the three reported quarters for 2018, Q3 is the most likely one to reflect shipments of replacement parts. Possibly at a discount, but obviously not a very big one.

In any case I think we're seeing a lot of artificial pumping of 14nm "demand": people buying stuff because they have to (Intel won't sell them anything else). Nobody wants to see Intel continue releasing 14nm products. Sure, Intel is making money now, but very soon people are going to go sour grapes unless Intel makes major changes to their product lineup.

Early shipment for CSL-SP is happening now (as in now-now, right now).
Volume is Q1 2019.

 

[DrMrLordX]

Early shipment for CSL-SP is happening now (as in now-now, right now).
Volume is Q1 2019.

Cascade Lake-SP or AP?

 

[NTMBK]

Do you think Intel's strong Q1, Q2 and Q3 perfomance were due to shipping unreleased Cascade Lake chips? Or do you think the return to growth in the PC market is because your avg Joe is buying new laptops with Spectre/Meltdown mitigations?

Intel added two cores to pretty much every chip in their lineup, finally giving Avg. Joe a solid reason to replace his Sandy Bridge dual-core laptop!

 

[Yotsugi]

Cascade Lake-SP or AP?

-SP.
-AP is dunno when.

 

[TheELF]

First, my comment is not directed at TheELF, but the context.

This is almost funny. The "No company is stupid enough" and "Intel" in the same sentence.

Yes, they were that stupid with 10nm

Looking at intel's arch day they announced a lot of products on 10nm which pretty much confirms that there is no problem with their 10nm production,so yes I guess they switched a lot of production over to 10nm because they made new products and they prefer to use the available volume on these new products since 14nm CPU sales are still more than fine.
So yeah I guess one could say that they have a problem with making enough volume on 10nm.

 

[Yotsugi]

which pretty much confirms that there is no problem with their 10nm production

Hence why it's EOY 2019 HVM for smallass chips.
All while TSMC is busy shipping >100mm^2 A12X in volume.