News Intel CFO: Our 10nm Will Be Less Profitable than 22nm

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,620
136
Is this the first time Intel has been this clear about the state of their 10nm node as well as having lost their process node leadership?


“Look, this just isn't going to be the best node that Intel has ever had. It's going to be less productive than 14 [nanometer], less productive than 22 [nanometer] … The fact is, like I said, it isn't going to be as strong a node as people would expect from 14nm or what they'll see in 7nm.”

"we feel like we're starting to see the acceleration on the process side that we have been talking about to get back to, you know, parity in the 7nm generation and regain leadership in the 5nm [generation]"


Bonus: Intel expected AMD to grab more of the server market sooner.
"I think we, we've said we expect to see stronger competitive dynamics in the second half of this year. But what we've seen, we actually thought we would see some of that a little bit sooner. And what we've seen is, again, very strong demand for our leading edge products. ... But we do see rising competition, we plan for that, as we talked about our forecast, that they'll be share impacts that we'll have to absorb within that. And then as we look at our product roadmap over time, we think again, we start to present an even more compelling competitive position as we go into 7nm and 5nm."
 

Markeyse

Member
Feb 9, 2020
112
13
41
They Will Recover though.

Note Intel produce most their own silicone, and most designs are the monolithic designs, which is harder and more expensive to produce than AMD's modular chiplet design. If Xe is any indication, Intel will start adapting that model for 7nm as well.
 

BigDaveX

Senior member
Jun 12, 2014
440
216
116
They Will Recover though.

Note Intel produce most their own silicone, and most designs are the monolithic designs, which is harder and more expensive to produce than AMD's modular chiplet design. If Xe is any indication, Intel will start adapting that model for 7nm as well.
It's tempting to write off their 10nm node as an aberration, but if you go back the last decade-and-a-half or so, an uncomfortable pattern starts to emerge among Intel's processes:

90nm - Decent at lower clockspeeds, complete redacted at anything higher than about 3GHz.
65nm - Awesome.
45nm - Equally awesome.
32nm - Slow start, only being used for Gulftown and the CPU die on the early i3s, but was doing pretty nicely by the time Sandy Bridge showed up.
22nm - A few teething troubles, with Ivy Bridge being paper launched and some chips dying very quickly from over-volting, but otherwise not too bad at all.
14nm - Major issues at first, forcing Intel to focus on 2C mobile Broadwells and a smattering of Xeons/HEDT chips, and resulting in Skylake launching over a year late.
10nm - EPIC FAIL.

They've certainly been doing better than AMD/GloFo - who never managed any better than an otherwise very good 45nm process that was "only" about 15 months later to the market than its Intel counterpart - over the same time period, but in retrospect it seems like 10nm has just been the breaking point for a series of issues over the last decade-and-a-half, rather some anomaly that'll be forgotten once we get to 7nm.

Profanity (even abbreviated)
is not allowed in the tech forums.

AT Mod Usandthem
 
Last edited by a moderator:

OriAr

Member
Feb 1, 2019
63
35
91
10nm was probably the climax of years of mismanagement in Intel's manufacturing group combining for a perfect storm of mess.
Hadn't Murthy ordered the timeout 2 years ago chances are Intel would be still up in deep redacted with it (Instead of things actually getting somewhat better now and looking to be finally OK by the end of 2021).
Intel scaling back 7nm and being open about 10nm's failings suggests to me that they learned quite the lesson there, and personally I get the feeling things are going well with 7nm so far.
I'd be very hesitant to count out Intel outside the leading edge node race, all it'll take is one TSMC redacted up and Intel will regain the lead.

Profanity isn't allowed
in the tech forums.

AT Mod Usandthem
 
Last edited by a moderator:
  • Like
Reactions: pcp7

jpiniero

Lifer
Oct 1, 2010
14,511
5,159
136
Skylake wasn't a year late. It was more or less released when it was supposed to be. Broadwell had basically a year delay though.

4770K was launched on June 2, 2013
4790K was launched on June 2, 2014
6700K was launched on August 5, 2015
 
  • Like
Reactions: coercitiv

OriAr

Member
Feb 1, 2019
63
35
91
Skylake wasn't a year late. It was more or less released when it was supposed to be. Broadwell had basically a year delay though.

4770K was launched on June 2, 2013
4790K was launched on June 2, 2014
6700K was launched on August 5, 2015
Broadwell was such a low volume it didn't really exist for desktop.
 

Markeyse

Member
Feb 9, 2020
112
13
41
It's tempting to write off their 10nm node as an aberration, but if you go back the last decade-and-a-half or so, an uncomfortable pattern starts to emerge among Intel's processes:

90nm - Decent at lower clockspeeds, complete clusterf**k at anything higher than about 3GHz.
65nm - Awesome.
45nm - Equally awesome.
32nm - Slow start, only being used for Gulftown and the CPU die on the early i3s, but was doing pretty nicely by the time Sandy Bridge showed up.
22nm - A few teething troubles, with Ivy Bridge being paper launched and some chips dying very quickly from over-volting, but otherwise not too bad at all.
14nm - Major issues at first, forcing Intel to focus on 2C mobile Broadwells and a smattering of Xeons/HEDT chips, and resulting in Skylake launching over a year late.
10nm - EPIC FAIL.

They've certainly been doing better than AMD/GloFo - who never managed any better than an otherwise very good 45nm process that was "only" about 15 months later to the market than its Intel counterpart - over the same time period, but in retrospect it seems like 10nm has just been the breaking point for a series of issues over the last decade-and-a-half, rather some anomaly that'll be forgotten once we get to 7nm.


100% agree with you. I think a lot of people forget that Intel is such a big ship, that we soo think they are suppose to just do the best, but there is a lot of things to consider for them, and a lot of Markets. AMD is a fraction of what Intel is, and they was able to take the performance crown this past year as far as the HEDT and some of the datacenter markets are concerned. They also have GloFo since they been fabless for a while, while Intel produce the majority of their products. I think now having Raja Koduri, and chip guru Jim Keller, a lot of things will change with Intel, but we won't really see that until next year. But man isn't it some exciting times we in for tech, and especially CPU tech.
 

Markeyse

Member
Feb 9, 2020
112
13
41
10nm was probably the climax of years of mismanagement in Intel's manufacturing group combining for a perfect storm of mess.
Hadn't Murthy ordered the timeout 2 years ago chances are Intel would be still up in deep redacted with it (Instead of things actually getting somewhat better now and looking to be finally OK by the end of 2021).
Intel scaling back 7nm and being open about 10nm's failings suggests to me that they learned quite the lesson there, and personally I get the feeling things are going well with 7nm so far.
I'd be very hesitant to count out Intel outside the leading edge node race, all it'll take is one TSMC redacted up and Intel will regain the lead.

Completely agree! And there is more pressure on Intel than most other tech companies. They produce a wider variety of products. I mean Networking, FPGA, CPU, GPU soon, and ASIC/AI. And more product categories than that even. But you see what they are doing with OneAPI, and you see what they are doing with 3D Stacking and Foveros, and what they about to do with their CPU and GPU. It is going to be great. I bet in a year or two we won't be complaining anymore!

Skylake wasn't a year late. It was more or less released when it was supposed to be. Broadwell had basically a year delay though.

4770K was launched on June 2, 2013
4790K was launched on June 2, 2014
6700K was launched on August 5, 2015

Some stuff was delayed though.
 
Last edited by a moderator:
  • Like
Reactions: pcp7

Markeyse

Member
Feb 9, 2020
112
13
41
And lets consider one thing. All of AMD Ryzen chips aren't 7nm, and not all of their chips have everything 7nm when they are considered to be. I think the cores are 7nm, but the I/O is like 12 or 14nm if I'm not mistaken.
 

Doug S

Platinum Member
Feb 8, 2020
2,203
3,407
136
10nm was probably the climax of years of mismanagement in Intel's manufacturing group combining for a perfect storm of mess.
Hadn't Murthy ordered the timeout 2 years ago chances are Intel would be still up in deep redacted with it (Instead of things actually getting somewhat better now and looking to be finally OK by the end of 2021).
Intel scaling back 7nm and being open about 10nm's failings suggests to me that they learned quite the lesson there, and personally I get the feeling things are going well with 7nm so far.
I'd be very hesitant to count out Intel outside the leading edge node race, all it'll take is one TSMC redacted up and Intel will regain the lead.


Sure, but TSMC has showed no signs of redacted up in any way. They were taping out 5nm designs for "a major customer" (i.e. Apple) last September so mass production of 5nm with the full EUV stack starting this summer looks to be a done deal. Even accounting for how what Intel calls 7nm is roughly equivalent to what TSMC calls 5nm, TSMC would have to hit a multi year stumbling block for Intel to catch up. And that assumes Intel was able to return to their two year cadence, which they haven't maintained for nearly a decade now.

Intel is beginning to face the same reality that AMD faced a while back that caused them to spin off GloFo, that RISC vendors like Sun and HP faced even further back which caused them to stop fabbing their own chips. Every process generation gets more expensive to develop and deploy, but Intel no longer increases annual sales at a double digit clip to stay ahead of that reality. They will not have their own fabs by the end of this decade.

If Intel was unable to deliver 10nm literally for YEARS, I don't see why that should give anyone confidence they will keep their promises about 7nm. Remember that 14nm was delayed a fair amount too - they covered that up by retroactively changing the schedules they had shared with customers showing earlier dates so they could claim it was only a few months late. These issues have been building for a long time, and it is very unlikely they'll be able to sweep all that away and go back to perfect execution on a two year cadence.

Profanity is not allowed
in the tech forums.

AT Mod Usandthem
 
Last edited by a moderator:

jpiniero

Lifer
Oct 1, 2010
14,511
5,159
136
Sure, but TSMC has showed no signs of redacted up in any way. They were taping out 5nm designs for "a major customer" (i.e. Apple) last September so mass production of 5nm with the full EUV stack starting this summer looks to be a done deal. Even accounting for how what Intel calls 7nm is roughly equivalent to what TSMC calls 5nm, TSMC would have to hit a multi year stumbling block for Intel to catch up. And that assumes Intel was able to return to their two year cadence, which they haven't maintained for nearly a decade now.

But AMD is saying now that they won't ship any 5 nm CPUs until 2022. What TSMC does isn't entirely relevant in this case. It's just tough to say whether it's realistic Intel can actually release Granite Rapids in 2022.

Intel is beginning to face the same reality that AMD faced a while back that caused them to spin off GloFo

AMD had to spin off the fabs because they spent way too much on ATI.
 

dmens

Platinum Member
Mar 18, 2005
2,271
917
136
Since Intel is using EUV (and chiplets presumably) they have a better chance of 7 nm actually working out. But you are talking about two years from now at a min.

So all they need to do is:
- Catch up to a foundry competitor that has at least a year of a head start on a process that is unprecedented in its difficulty to achieve HVM
- Overrule an army of powerful careerists in middle management who are dead set against doing anything new
- Come up with better designs while suffering the biggest brain drain in the company's history
- Maintain revenue while fighting against attacks on every single market it is engaged in

Piece of cake. :grinning:
 
  • Like
Reactions: scannall and Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,483
14,434
136
- Overrule an army of powerful careerists in middle management who are dead set against doing anything new
Actually, this one line can also help them. The idiot middle managment of many large companies have managers that say "we can't use anything but Intel CPUs in our datacenter, the only real CPUs".

At least thats what I heard from our management back in 2002-2006 when Opteron was kicking their hiney and power and performance vs P4 Xeons.
 

dmens

Platinum Member
Mar 18, 2005
2,271
917
136
Actually, this one line can also help them. The idiot middle managment of many large companies have managers that say "we can't use anything but Intel CPUs in our datacenter, the only real CPUs".

At least thats what I heard from our management back in 2002-2006 when Opteron was kicking their hiney and power and performance vs P4 Xeons.

That is assuming the newcomers actually win the political battles. Don't hold your breath.