intel delays 14nm?-process problems

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

cytg111

Lifer
Mar 17, 2008
23,006
12,689
136
having ppl, even engineers, run round in circles producing nothing is obviously quite expensive, so of course Intel will keep some kind of a release schedule. On the other hand, the resource locking that intel tries to fix in haswell with TX also somewhat applies to people, if one guys is 1x effective, ten guys will not be 10x.
Without proper competition I am quite sure Intel will scale back so that the collective effectiveness of their workers are optimized relative to the market/demand that they're supplying hardware for.
But thats just my deranged brain.

edit : being on the hunt for the low power segment produces another competition vector of course (different direction, different momentum)
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
So what happened to this then:

http://forums.anandtech.com/showthread.php?t=2274477

It says Broadwell @ 14 nm is already taped out and working.

So are they having problems with high volume production?

Oregon and Arizona(new Fab42) fabs will as expected. The irish fab conversion is just 6 months delayed. The 14nm production will still be higher than the 22nm was at launch.

Broadwell is taped out. Else it couldnt be ready for Q1 2014 release.

The PCH (8 and 9 series chipset) for Haswell and Broadwell will be made on 32nm process.
 
Last edited:

Dadofamunky

Platinum Member
Jan 4, 2005
2,184
0
0
Yeah this is totally about demand. Could be a sign they are pessimistic about Airmont's ability to drive smartphone sales.

Based on the Atom-based Motorola RAZR i's pallid performance numbers against other current-gen smartphones, (http://www.anandtech.com/show/6440/google-nexus-4-review/3) I'd say you're hitting the bulls-eye.

I have to admit though, a little slow-down in Intel product cycles wouldn't break my heart.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
the only problem with 14nm is the lack of incentives (read: competition) to make it work.

Smaller is always better :)


Now if only Intel can get atom ASPs up? (I'd imagine there is much room for ASP improvement provided the Intel phone SOC allows a much smaller and more powerful phone to be built. This compared to all the myriad of competitors using basically the same TSMC process.)
 
Last edited:

Torn Mind

Lifer
Nov 25, 2012
11,557
2,624
136
i have found the source of the 14nm problems. look at the picture: Intel has replaced "Israel" with "Ireland" and thus losing 4 FAB points. (28 vs 24) :p

but really, for Intel - the only problem with 14nm is the lack of incentives (read: competition) to make it work. i doubt it even makes economical sense to build an entirely new fab for this process plus converting 2 other fabs to 14nm, think about it as diminishing returns.
but they will do it anyways.
They need their chips to reduce power consumption. They get more chips to sell from the same size wafer. Both of these things are need to keep them for them to compete against ARM CPUs.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
They need their chips to reduce power consumption. They get more chips to sell from the same size wafer. Both of these things are need to keep them for them to compete against ARM CPUs.

I'm also wondering how much Intel's process advantage can affect the form factor of the phone?

For example, Here is the Lava Xolo mainboard (Intel Penwell 32nm SOC under the DRAM) with ICs located on both sides of it. The cut out is for the battery.

http://www.chipworks.com/blog/recentteardowns/2012/05/15/inside-the-lava-xolo-intel-penwell-inside/

full-board-front.jpg


full-board-side1.jpg


1. If Intel can get the rest of the ICs either smaller or integrated onto the main SOC the logic board size should drop down a good amount. This allowing for a larger battery and/or smaller, thinner, lighter phone.

2. Or will Intel instead use its process advantage to install a larger CPU and/or GPU on the SOC while keeping the rest of the discrete ICs the same size as their competitors?

My guess is that smartphone display technology will probably play a strong role in the Intel Atom SOC design choices. (re: If we get the low power screens, it would seem logical to me that SOC integration (aka choice number 1) has a greater impact on the final form factor of the phone.)
 

guskline

Diamond Member
Apr 17, 2006
5,338
476
126
Own both Intel and Amd chips so I don't have a dog in the fight. What's the big deal? Intel has the money and expertise to make the 14nm. Are they having problems? Probably. Are they able to solve them? Most likely.

Now let's ask the same questions about AMD and TRUTHFULLY answer the last question.:colbert:
 
Last edited:

Haserath

Senior member
Sep 12, 2010
793
1
81
Own both Intel and Amd chips so I don't have a dog in the fight. What's the big deal? Intel has the money and expertise to make the 14nm. Are they having problems? Probably. Are they able to solve them? Most likely.

Now let's ask the same questions about AMD and TRUTHFULLY answer the last question.:colbert:

You mean AMD going bankrupt before 14nm?:colbert:
 

inf64

Diamond Member
Mar 11, 2011
3,680
3,943
136
They don't have any technical issues it's just that demand is lower than expected and if they delay one fab's ramp of 14-nm production it is not a sign of process node troubles.

Intel is still miles ahead of any other foundry and this won't change any time soon.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
They don't have any technical issues it's just that demand is lower than expected and if they delay one fab's ramp of 14-nm production it is not a sign of process node troubles.

Intel is still miles ahead of any other foundry and this won't change any time soon.

They actually accelerate 14nm. Since they build another expansion to the D1 complex. Plus the fab42 is an entirely new full fab.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
They don't have any technical issues it's just that demand is lower than expected and if they delay one fab's ramp of 14-nm production it is not a sign of process node troubles.

Intel is still miles ahead of any other foundry and this won't change any time soon.

I think the rumor of "delay" on the 14nm has merits, but not in the way most people think.

Ivy Bridge was "delayed" but came with more advancements than expected like the graphics, causing them to call it a Tick+. Broadwell may arrive 6 months later in late 2014/early 2015, but is rumored to have architecture changes to achieve lower power and another significant overhaul on the iGPU. Also, even PCs are increasingly adopting the Smartphone/Tablet model of much more co-operation between Intel and PC manufacturers. That will certainly add to the time.

Rather than going the traditional pace of introducing new process technology every 2 years with certain gain, they are moving to 2.5 years but ending up that much better in the end.

Imagine over 10 years.

2.5x gain per 2.5 years x 4 = 10x
2x gain per 2 years x 5 = 10x

The change might be because it is getting harder to shrink transistors, but engineers and managers are using clever ways to make the end result the same. Ultimately though, time will tell what will happen.
 
Last edited:

pablo87

Senior member
Nov 5, 2012
374
0
0
I think the rumor of "delay" on the 14nm has merits, but not in the way most people think.

Ivy Bridge was "delayed" but came with more advancements than expected like the graphics, causing them to call it a Tick+. Broadwell may arrive 6 months later in late 2014/early 2015, but is rumored to have architecture changes to achieve lower power and another significant overhaul on the iGPU. Also, even PCs are increasingly adopting the Smartphone/Tablet model of much more co-operation between Intel and PC manufacturers. That will certainly add to the time.

Rather than going the traditional pace of introducing new process technology every 2 years with certain gain, they are moving to 2.5 years but ending up that much better in the end.

Imagine over 10 years.

2.5x gain per 2.5 years x 4 = 10x
2x gain per 2 years x 5 = 10x

The change might be because it is getting harder to shrink transistors, but engineers and managers are using clever ways to make the end result the same. Ultimately though, time will tell what will happen.

More importantly, it reduces annual capex and stretches out the depreciation schedule. Which, I would not be surprised, is part and parcel of a POR to dramatically improve factory ROI from where it is today ($12B capex in 2012 on ~300MM processors shipped) back to where it was ($5.2B capex in 2010 on 280MM? processors shipped).

Let's face it, Intel has a huge process lead on AMD, increasing that lead esp. across the board, serves no purpose in x86. As for tablets and smartphones, seems the issue lies elsewhere...
 

2is

Diamond Member
Apr 8, 2012
4,281
131
106
I think at this point in time, Intel is watching ARM and the mobile market like a hawk... I doubt AMD is even on their radar.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Let's face it, Intel has a huge process lead on AMD, increasing that lead esp. across the board, serves no purpose in x86. As for tablets and smartphones, seems the issue lies elsewhere...

No, they are not. The competition against them is now purely performance/$ across its range, and they seem to be doing well there too.

But I think you misworded it there a bit. You said "it serves no purpose in x86". The correct thing to say is that "it serves no purpose in the traditional PC market."

Time to stop the denial. AMD's revenue and unit shipment breakdown shows that it ships 50/50 mobile and desktop chips. For them, Desktop is still very relevant. For Intel its not, because its 65/35 for mobile/desktop in units, and probably 70/30 for revenue. In fact, Intel probably wants to shift to mobile further, especially that they are moving their process advantage to better serve that purpose.

And in the NEW PC era, pushing process is still important. I do not think though in a relative sense they are slowing down either. Though they were later than expected with Ivy Bridge and 22nm, that was same with foundries.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
As for tablets and smartphones, seems the issue lies elsewhere...

You are probably talking about that 22nm is expected towards end of 2013.

See, unlike the PC market, where its already established and any time shift by Intel is directly felt, that's not true for mobiles. Taking a chip and making it into a fully functional smartphone takes time.

The first Medfield Smartphone came in May 2012. If Nokia hadn't abandoned MeeGo and went with Medfield, it would have been out fall of last year. The chips were available more than a year before that, which is typically what's needed by phone makers. If the Nokia deal worked, we'd have seen the 22nm chip that much earlier, which would make it early-mid 2013.

But the nature of your posts may mean I'm wasting my posts anyway so I'll stop here.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
The delay is mainly due to another 1.1mio sqf additional expansion of D1X. D1X is simply higher priority. And fits Intels latest "US jobs" plan.
 

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
Time to stop the denial. AMD's revenue and unit shipment breakdown shows that it ships 50/50 mobile and desktop chips. For them, Desktop is still very relevant.

Actually it is a lot more relevant than you are stating here.

Once you exclude Brazos and look at their big core numbers only, you get not 1:1 to but 1:2 favoring desktops. Despite the hype and the marketing gimmicks, AMD is still a desktop company.
 

pablo87

Senior member
Nov 5, 2012
374
0
0
No, they are not. The competition against them is now purely performance/$ across its range, and they seem to be doing well there too.

But I think you misworded it there a bit. You said "it serves no purpose in x86". The correct thing to say is that "it serves no purpose in the traditional PC market."

This is what I meant, yes - I also believe x86 can sur/pete ARM on performance and hopefully compete on power. Cost well that depends on the afore speculated POR.... The rest well, you lost me.:confused:
 

pablo87

Senior member
Nov 5, 2012
374
0
0
The delay is mainly due to another 1.1mio sqf additional expansion of D1X. D1X is simply higher priority. And fits Intels latest "US jobs" plan.

Most likely because its most cost effective...all this talk about us jobs by corporations, nobody actually believes this do they? Fact is, US is becoming very competitive due to the lowest energy costs in the world.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Most likely because its most cost effective...all this talk about us jobs by corporations, nobody actually believes this do they? Fact is, US is becoming very competitive due to the lowest energy costs in the world.

Energy aint the reason. But one reason could be that US wages are so low, that its starting to be on the level of development countries. Or simply centralization of its operations.

http://www.eia.gov/electricity/monthly/current_year/may2012.pdf
http://www.energy.eu/

Cheap energy by itself doesnt mean you are more competitive. Cheap energy can actually be a direct penalty in efficiency modernization.
 
Last edited:

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Energy aint the reason. But one reason could be that US wages are so low, that its starting to be on the level of development countries. Or simply centralization of its operations.

http://www.eia.gov/electricity/monthly/current_year/may2012.pdf
http://www.energy.eu/

Cheap energy by itself doesnt mean you are more competitive. Cheap energy can actually be a direct penalty in efficiency modernization.

Dang Euros don't know nothing about the benefits of cheap oil :whiste:

(j/k) But this matter is view differently in the US. or at least in parts of it.