[Verge] Intel revenues down 5%

[Ajay]

What did Penryn have to do with capacity and yield issues?

Intel rarely faces the issues that Qualcomm and others faces right now. Or the same issue that TSMC also had at 40nm and before. GloFo is another utter disaster for example.

Also as a small note, the year of the 32nm shortage. Intel grew 20.7%.

GF seems to be pushing really hard right now. I think they know it's do or die as a foundry unless they get 14XM up and running fast. I think they need to beat TSMC to 14nm to be able to stay in the game (and execute that node very well).

The agreement with ARM could keep them in the game, but they'll need a miracle to provide a higher performance 14nm node to attract something like AMD's GFX business (and offer AMD some performance value post Excavator, which they'll need if they stay in the x86 biz).

 

[ShintaiDK]

GF seems to be pushing really hard right now. I think they know it's do or die as a foundry unless they get 14XM up and running fast. I think they need to beat TSMC to 14nm to be able to stay in the game (and execute that node very well).

The agreement with ARM could keep them in the game, but they'll need a miracle to provide a higher performance 14nm node to attract something like AMD's GFX business (and offer AMD some performance value post Excavator, which they'll need if they stay in the x86 biz).

Right now TSMC doesnt list 14nm. Only CLN16FF (16nm).

Global Foundries list 14Xm as a 2014 product. But I think its worth noticing they list 28nm as a 2011 product too. Factory roadmaps and actual products is 2 very different things 😉
Its also worth noticing that regular CPUs, GPUs etc certainly wont be made on 14XM. GloFo even stats that its for customers comign from the LPM nodes. SoCs at TSMC for example is made on the HPM node.

 

[Idontcare]

For reference in case not everyone has seen it.

TSMC taps ARM's V8 on road to 16-nm FinFET

 

[Olikan]

so...TSMC and GloFo are in crazy race hum?

 

[ShintaiDK]

so...TSMC and GloFo are in crazy race hum?

Remember its factory roadmaps tho. Usually +1 year on actual products.

Also TSMC and GloFo seems to target completely different customers.

http://www.eetimes.com/design/eda-design/4397983/Globalfoundries-14-nm-is--low-shrink--node

Glofos 14XM actually looks rather dissapointing arcording to this. Little to no size reduction compared to 20nm.

 

[Ben90]

Funny, if I was planning to buy a new machine I'd try do it before Win 8 launches and I'm stuck with it.

http://compare.ebay.com/like/190735581993?var=lv&ltyp=AllFixedPriceItemTypes&var=sbar

 

[Olikan]

Remember its factory roadmaps tho. Usually +1 year on actual products.

Also TSMC and GloFo seems to target completely different customers.

http://www.eetimes.com/design/eda-design/4397983/Globalfoundries-14-nm-is--low-shrink--node

Glofos 14XM actually looks rather dissapointing arcording to this. Little to no size reduction compared to 20nm.

well, not that much dissapointing if you compare to competition...
http://semiaccurate.com/forums/showpost.php?p=168953&postcount=11

 

[Ajay]

Remember its factory roadmaps tho. Usually +1 year on actual products.

Also TSMC and GloFo seems to target completely different customers.

http://www.eetimes.com/design/eda-design/4397983/Globalfoundries-14-nm-is--low-shrink--node

Glofos 14XM actually looks rather dissapointing arcording to this. Little to no size reduction compared to 20nm.

Looks like a reasonable compromise. Better performance, but without a significant decrease in die size - but easy migration from 20nm designs. It's a mixed bag and I don't know how it will play out. At least GF is fighting for business, it will benefit foundry customers in the short run. I also noted that GF will need to follow on with a high performance process quickly to get higher margin contracts, unless they plan to put volume ahead of margins.

 

[Arachnotronic]

Intel has fabs down for upgrading to 14nm production. I see nothing in this that will affect their manufacturing advantage.

Idling fab capacity != slowing R&D. Intel short sellers/haters like to distort facts.

 

[Blandge]

Intel's Atom SoCs meant for mobile are on a cadence a year or two after their laptop lines. This means whenever Intel brings in 14nm, their Atom SoCs will follow by a year or two.

Remember this from MWC 2012? Keep in mind Intel's roadmaps set dates for actual product launches, not process readiness. Atom catches up to Core at 14nm.

 

[exar333]

Remember this from MWC 2012? Keep in mind Intel's roadmaps set dates for actual product launches, not process readiness. Atom catches up to Core at 14nm.

Yeah. I think we will start to see large volumes of 22nm Atoms next year and things really kick-up with 14nm.

With Intel idling some 22nm capacity, I can only see them really accelerating their Atom dev to be able to take advantage of this capacity ASAP. Once they do, they should start shipping in huge volumes, especially when they start to moce to 14nm in early 2014 with their desktop products.

 

[Lonyo]

Assuming they have 22nm Atom designs ready, it wouldn't make any sense to idle 22nm capacity when they are competing in a power and size focused world and most current Atoms are still 32nm.

 

[pelov]

Remember this from MWC 2012? Keep in mind Intel's roadmaps set dates for actual product launches, not process readiness. Atom catches up to Core at 14nm.

The Atoms are still slated for a Q4 2013 release with products only trickling in early Q1 2014.

I don't understand how some of you guys aren't seeing the gap here. There's obviously a year-ish delay between Atoms and the other 22nm products. How does that not affect their fab advantage? Of course it does. Whether this has to do with fabs or just a slow development cycle and Intel not being ready for the low end (the more likely answer) doesn't matter. If the products aren't available then the readiness of the process doesn't equate into an advantage.

 

[Blandge]

Assuming they have 22nm Atom designs ready, it wouldn't make any sense to idle 22nm capacity when they are competing in a power and size focused world and most current Atoms are still 32nm.

Atom uses a the low power 22nm process, and the "idle" capacity they have is the high performance process. They cannot simply manufacture Atom on this excess 22nm headroom.

 

[Blandge]

The Atoms are still slated for a Q4 2013 release with products only trickling in early Q1 2014.

I don't understand how some of you guys aren't seeing the gap here. There's obviously a year-ish delay between Atoms and the other 22nm products. How does that not affect their fab advantage? Of course it does. Whether this has to do with fabs or just a slow development cycle and Intel not being ready for the low end (the more likely answer) doesn't matter. If the products aren't available then the readiness of the process doesn't equate into an advantage.

And I am talking about 14nm, not 22nm.

There is usually a 2 year gap between process shrinks. Tick (Proccess Shrink) Tock (Microarchitectural Advancement) Tick Tock.

For Core products 2012 Tick(22nm Ivy Bridge) 2013 Tock(22nm Haswell) 2014 Tick (14nm Broadwell)

For Atom products 2012 Tick (32nm Medfield/Clover Trail) 2013 Tock (22nm Atom) 2014 Tick (14nm Atom)

Notice there is a 1 year cadence for process shrinks from 32nm to 14nm. This is how that 1 year gap will be closed. 22nm Ivy Bridge is released in 2012, 22nm Atom is released in 2013 (1 year gap), 14nm Broadwell is released in 2014, 14nm Atom is released in 2014 (< 1 year gap).

 

[pelov]

Notice there is a 1 year cadence for process shrinks from 32nm to 14nm. This is how that 1 year gap will be closed. 22nm Ivy Bridge is released in 2012, 22nm Atom is released in 2013 (1 year gap), 14nm Broadwell is released in 2014, 14nm Atom is released in 2014 (< 1 year gap).

I'm not doubting they'll have to accelerate their atom line. In fact that was my point. Still, people, and you until you clarified, were claiming the 22nm chips were hitting 2013 when they're quite clearly not and that gap is still very much there into 2013 and 2014, although to a lesser extent in the latter year.

It's a sign of Intel transitioning to mobile in both design and the accompanied low-power focused fab process, but denying that ARM won't be right there with them during those 2 calendar years is a little silly (again, you clarified that you at least understand that but quite clearly there are people in this thread that don't)

-- since we're focusing on Intel and their ramping and transitioning it would be useful to give a comparison with another fab.

TSMC taped out 28nm in Q1/Q2 2011 and had their 28nm running by Q1 2012. Their delay was also roughly what Intel experienced at the 22nm node, which was about a month.

 

[Blandge]

I'm not doubting they'll have to accelerate their atom line. In fact that was my point. Still, people, and you until you clarified, were claiming the 22nm chips were hitting 2013 when they're quite clearly not and that gap is still very much there into 2013 and 2014, although to a lesser extent in the latter year. )

The 2013 deadline for 22nm Atom is target at the 2013 holiday season.

It's a sign of Intel transitioning to mobile in both design and the accompanied low-power focused fab process, but denying that ARM won't be right there with them during those 2 calendar years is a little silly (again, you clarified that you at least understand that but quite clearly there are people in this thread that don't)

Both TSMC and GF have a history of very slow ramps, while Intel has a history of fast ramps. Assuming that TSMC and GF are able to begin bulk production of 14nm chips in 2014 as they claim, it would not be in any meaningful quanities. And that's a pretty generous assumption considering they are both accellerating their roadmaps by a year AND pushing finFETs for the first time. I'd be willing to bet that both will struggle immensely to produce double digit yields in the first half of 2015.

This is of course purely speculation on my part.

 

[pelov]

The 2013 deadline for 22nm Atom is target at the 2013 holiday season.

According to Anand, the likely date is Q4/Q1 with products in Q1.

And you're neglecting something here with respect to Intel. While GloFo and TSMC will have customers paying them for their wafers, Intel is its own customer thus its absolutely pivotal that they do well in mobile and have consistent growth, but more importantly enough growth and money coming in to keep these pretty fabs to themselves. They don't just have to do well, they have to outpace the ARM space, and Samsung in particular, at an accelerating pace. Fabs don't get cheaper, they get exponentially more expensive as you drop down in node sizes. In order for Intel to keep it to themselves, they must outpace the competition at the same rate, or close to it, that they did with AMD.

A down quarter is the last thing Intel needs, and with Q4's outlook not looking any better, I'm not as hopeful as you are.

 

[Blandge]

According to Anand, the likely date is Q4/Q1 with products in Q1.

Intel is clearly aiming for holiday season 2013. Whether or not they actually succeed is what is in question.

A down quarter is the last thing Intel needs, and with Q4's outlook not looking any better, I'm not as hopeful as you are.

Intel is a mature company that has weathered worse storms than a couple lackluster quarters.

 

[pelov]

Intel is a mature company that has weathered worse storms than a couple lackluster quarters.

Against a single competitor with poor fab execution. That single competitor is irrelevant.

Intel needs to do not just well, but by a very large margin BETTER than anything in the ARM space, and that includes the fabs tied to all of the other competitors. Posting quarterly profits that are half of what they posted this time last year when they should be doing better isn't a good start.

 

[Khato]

According to Anand, the likely date is Q4/Q1 with products in Q1.

And you're neglecting something here with respect to Intel. While GloFo and TSMC will have customers paying them for their wafers, Intel is its own customer thus its absolutely pivotal that they do well in mobile and have consistent growth, but more importantly enough growth and money coming in to keep these pretty fabs to themselves. They don't just have to do well, they have to outpace the ARM space, and Samsung in particular, at an accelerating pace. Fabs don't get cheaper, they get exponentially more expensive as you drop down in node sizes. In order for Intel to keep it to themselves, they must outpace the competition at the same rate, or close to it, that they did with AMD.

A down quarter is the last thing Intel needs, and with Q4's outlook not looking any better, I'm not as hopeful as you are.

I believe the only date we've seen for a 22nm Atom is Valleyview no? That's the one with an expected Q4 2013 PRQ, which likely means Q1 2014 product introductions. Valleyview is the tablet SoC, so if we have a similar gap between the 22nm smartphone and tablet SoCs as we did on 32nm then we could see the smartphone product in the first half of the year.

As for the doom and gloom... you're entitled to your opinion, and no one should question the fact that it is a possible outcome. Just like there could be some research discovery that invalidates silicon based semiconductors and upturns the entire industry. Even shrinking, Intel's current markets are more than adequate to continue supporting its current R&D and fab expenditures for 4+ years. But it's not just that - why exactly do you think that they are fabricating FPGA's for Achronix and Tabula? If you look at some of those TSMC announcements for when they ship product on a new node to customers, it's almost always FPGA's because they're one of the 'easiest' designs to start with. Intel is putting together the necessary expertise and toolset to open its fabs to outside companies if it needs to.

Regardless, the mobile space could easily go either way, all we can currently do is speculate... and since we don't have adequate information available that speculation is pretty meaningless.

 

[Blandge]

Against a single competitor with poor fab execution. That single competitor is irrelevant.

Intel needs to do not just well, but by a very large margin BETTER than anything in the ARM space, and that includes the fabs tied to all of the other competitors. Posting quarterly profits that are half of what they posted this time last year when they should be doing better isn't a good start.

I think you're forgetting that Intel's history dates back further than 5-6 years. Motorola and IBM were pushed out of the desktop industry after PPC lost traction due to manufacturing and performance issues in the early to mid 2000s.

Intel's been around for almost have a century, and it's had more competitors than AMD.

 

[Phynaz]

osting quarterly profits that are half of what they posted this time last year when they should be doing better isn't a good start.

Let's see what the profits of their competitors come out to be. Want to bet that Intel beats them all?

 

[Nemesis 1]

The Atoms are still slated for a Q4 2013 release with products only trickling in early Q1 2014.

I don't understand how some of you guys aren't seeing the gap here. There's obviously a year-ish delay between Atoms and the other 22nm products. How does that not affect their fab advantage? Of course it does. Whether this has to do with fabs or just a slow development cycle and Intel not being ready for the low end (the more likely answer) doesn't matter. If the products aren't available then the readiness of the process doesn't equate into an advantage.

I think its more like the norm will see after14nm. Intels medfeild kinda revealed this . Us guys here see a chip released and we buy them . Intel is new and as proven by medfield powerful in this areana. The time between when Intel was ready with medfield chips to actual products can be easily tracked. This is the Time gap were all seeing. Time to actual product in the market. Because of RazrI Motorale will be first out of the gate with the next 22nm chip . likely by months

 

[Khato]

Posting quarterly profits that are half of what they posted this time last year when they should be doing better isn't a good start.

What exactly did they post in Q3 2012 that was half of Q3 2011? Revenue of $13.5 billion vs $14.2 billion? Nope. Operating income of $3.8 billion vs $4.8 billion? Nope. Net income of $3.0 billion vs $3.5 billion? Nope.