backtop


Print 29 comment(s) - last by Cullinaire.. on Jan 24 at 1:13 PM

32nm will be critical to Intel's performance and cost advantage

Intel released its results for the fourth quarter of 2008 yesterday - it was only the second time in 20 years that Q4 results were worse than Q3 results.

Q4 revenue was $8.2 billion, down from $10.7 billion last year. Operating income was $1.5 billion, and net income was only $234 million. Earnings per share was 4 cents, down 90% from last year. The results included a billion dollar write-off in Intel's investment in Clearwire.

Over $1.8 billion was spent on CAPEX (capital expenditures) in Q4. CAPEX are used by a company to acquire or upgrade physical assets such as equipment, property, or industrial buildings.

Average Selling Prices (ASP) and gross margins were down due to Fab underutilization, inventory write downs, and the shift from notebooks to Atom and netbooks. Chipset sales by revenue in the Digital Enterprise group were almost half of last year's result.

Intel has only missed earnings estimates once in the last five quarters. The fallout from lower corporate spending is obviously dire, as the most profitable Intel products are in the corporate server space. There was a $250 million charge due to fab underutilization, as Intel now has three Fabs producing at 45nm on 300mm lines. These are Fab 32 in Chandler, Arizona, Fab 28 in Kiryat Gat, Israel, and Fab 11x in Rio Rancho, New Mexico.

For all of 2008, Intel's posted revenue was $37.6 billion, with operating income of $9 billion, and net income of $5.3 billion. Earnings per share was 92 cents.

Intel generated approximately $11 billion in cash from operations while paying cash dividends of $3.1 billion. It used $7.1 billion to repurchase 324 million shares of common stock, which helps to boost its EPS figures.

One major area of concern for technology enthusiasts was that Intel would cut back on CAPEX in 2009, delaying the rollout of its P1268 32nm process and setting back the arrival of Westmere, Nehalem's successor.

However, Paul Otellini, Intel's President and Chief Executive Officer repeated his commitment to 32nm within 2009. In fact, he mentioned 32nm process technology over a dozen times during a conference call.

He stated, "We're going to get there as fast as we possibly can", as 32nm will be critical to Intel's performance and cost advantage. Intel will not delay 32nm, insisting that it is necessary to enable higher production efficiencies.

Intel will be spending $2.5 billion on Research and Development, as well as Mergers and Acquisitions in the first quarter of 2009. Much of that will be on Intel's 32nm process, preparing to launch it in mass production, and ramping it to their other products.

Fab D1D, Intel's 300mm development fab in Hillsboro, Oregon, will be the prototype line, just like for the 45nm process. Over the next six months, this unique factory will turn the 32nm process from the research phase into the production phase. It will be in Hillsboro that Intel will adjust the production line to meet its yield and cost requirements. Once the process has been refined enough, Intel can use the lessons learned to convert other Fabs to the 32nm process.

For all of 2009, Intel plans to spend $5.4 billion on Research and Development. There will be a 2009 introduction for the 32nm process, but it will definitely be a 2010 ramp, just like the 45nm was introduced in 2007 but only readily available for sale to the general public in March 2008. Somewhat troubling was the statement that the "ramp rate will be modulated" based on market demand.

The current plan is to go with mainstream mobile and desktop parts first. Next generation System-on-Chip design, as well as servers will be next. After that, next generation Atom products will be the next priority. No mention was made of Larrabee.

The consensus recommendation of analysts is still to buy INTC. The share price has fallen approximately 20% in the past three months, and it is about 45% lower than it was a year ago.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

By the goat on 1/16/2009 9:14:31 AM , Rating: 2
Did you read the same article? Intel is not going to delay the introduction of 32nm.


By SilentSin on 1/16/2009 11:01:54 AM , Rating: 3
quote:
Somewhat troubling was the statement that the "ramp rate will be modulated" based on market demand.


They will introduce it, but they probably aren't in much of a rush to do so. As much of a geek as I am and love to see as much new tech as possible, I think Intel might be turning into their own worst enemy. Do any customers besides server farms really need a new-and likely premium priced-CPU refresh just 1 year after i7 dropped?

The market will be somewhat flooded with Intel CPUs next year. The Core2 is still a viable CPU thanks to lack of anything better from AMD, and there will soon be the ultra low power Core2 segment to go against AMD's Neo. The i7 can't be touched atm for desktops (servers soon too), the i5 still has yet to be seen, and then the i7+ is scheduled to be out early next year. That's 3 completely different mass market CPUs from one company. How does Intel plan to market that gagglef*ck of options? They might run into some of the issues NVidia has had with its products all over the place from the 8800->9800->GT200 lines, which were all viable competitors to AMD and themselves all in the same timeframe. NVidia seems to have taken the stance of "screw it, let the consumers figure out our mess", but that's not Intel's style. I'm not saying I want to see Intel delay anything without a good reason of course, but I can see where the idea would come from.


By MarchTheMonth on 1/16/2009 12:12:56 PM , Rating: 2
At the end of the day, they are a business, and they exist to make money. I love that Intel has pushed themselves to do the tick-tock we all know about, but I would not hold it against them if they temporarily dropped it in order to save some money. We would be lucky if they were to continue such developments so strong in this economy, but the fact is, when the demand for high performance isn't there (and instead, the demand for conservative money measures is), some things can/should be sacrificed, and in Intel's case, delaying some goals for the sake of better bottom lines in these times isn't unreasonable to me.


By Oregonian2 on 1/16/2009 2:29:40 PM , Rating: 2
Historically, Intel has forged ahead with CAPEX and product development during downturns, even perhaps increasing during those times (rumor is that they'd pick up good people from those other companies).

Intel is under new "administration" so there may be differences, but it's still their culture to use slow periods of their competitors (or potential competitors) to move further ahead.

P.S. - That Hillsboro talked about is the city immediately west of the city where I live (and am now). :-)


By ianweck on 1/17/2009 2:29:06 AM , Rating: 2
quote:
P.S. - That Hillsboro talked about is the city immediately west of the city where I live (and am now). :-)


Me too! Well, northwest anyway.


By kilkennycat on 1/17/2009 8:09:42 PM , Rating: 2
Hillsboro is NW from me. One of my sons summer-interned for a couple of years in DID.


By JKflipflop98 on 1/17/2009 9:29:05 PM , Rating: 2
I watch the OHV's dance to my whim every day. :)


By Cullinaire on 1/24/2009 1:13:43 PM , Rating: 2
OHVs schmOHVs...PGV's are where it's at!


By Jansen (blog) on 1/16/2009 12:16:52 PM , Rating: 2
Paul Otellini's viewpoint is that 32nm will lower production costs, once you have already outfitted the Fabs. The cost per wafer drops since you are geting so many more chips.

Also, the performance per watt of 32nm will enable new applications and markets that Intel wants to get into. I will explain more in a followup article, once I get a few questions answered.

Glad you guys find this as interesting as I do.


By paydirt on 1/16/2009 2:37:50 PM , Rating: 2
also, like the move from 65nm to 45nm allowed more space for either additional processor cores or cache, the move from 45nm to 32nm could allow for more cores/cache.


By Denithor on 1/16/2009 3:29:18 PM , Rating: 3
Well, the cost of individual chips drops because you get more per wafer is true, but it takes a while for this effect to fully develop. Early in the process they have higher rates of rejected chips that are faulty for some reason or other and those prevent them from immediately seeing higher yields from the same size wafers. As they refine the process and the line matures the cost/chip does go down below the previous generation's level.

If you think we're interested here, you should take a look over in the anandtech cpu forum.

;)


By Reclaimer77 on 1/17/2009 8:04:28 AM , Rating: 2
Yup ! Thanks, nice piece.

I wonder if the i5's will be 32nm ?


By Comdrpopnfresh on 1/16/2009 1:17:34 PM , Rating: 3
I think something, Ballmer maybe, said a bit ago may be applicable. It was an example of the importance for better algorithms in software leading to improved performance the user can make use of. What was said was along the lines that no matter how fast the hardware of a speech recognition system gets, if the progression of algorithms slows, speech recognition won't be any more accurate, it'll just return the same wrong interpretation to the user quicker.
I think the same is true on consumer platforms: If new hardware is indeed faster and more efficient, but the consumer's interface doesn't deliver the improvement offered by hardware, then better software is needed. If you could put windows 98 on a computer 100 years from now, it'll still crap its pants when a device is plugged in; just a lot quicker.

If the software isn't there to provide an improvement afforded by new chips to potential PC users, and consumers won't be opening their wallet much in the near future anyway, you're exactly right they'd only harm themselves with inter-competing offerings. Their best bet is to release chips suited to commercial use- ones that use less power, provide more throughput, and are better at visualization. So ++ on the server farm comment


By ekv on 1/16/2009 6:28:30 PM , Rating: 2
I read about it here ...

http://www.eetimes.com/showArticle.jhtml?articleID...

Essentially, it is in the processor vendor's best interest to invest in better algorithms to help differentiate their products. In addition, having a better algorithm than your competitor gives you a technical edge. And it'll keep your hardware design guys busy -- job security -- trying to implement the algorithm.

Speech recognition, noise cancellation, video processing, etc. Slam dunks. No-brainer.

Disclaimer: I have a BS degree in Math (and a minor in numerical analysis, and some crypto). Will work for food ... 8)


By atlmann10 on 1/19/2009 12:29:16 AM , Rating: 2
They have to Intel's largest competition is INTEL right now there latest core 2's are kicking there own a77. I want to build a new system this year myself. However if I do I am waiting for the Phenom2's with the AM3 chip set and DDR3, or I will build one on the new INTEL when they go to the dropped nm processors. I figure DDR3 and the X58 MB's should have dropped pretty well by then to. I imagine June to August when the price drop hits. I am also sure that sales in general are going to be slower than the norm because of the recession. Hopefully it means a double price drop for newer technology.


By thepalinator on 1/16/2009 11:41:55 AM , Rating: 2
They'll introduce SOMETHING at 32nm, but they're going to "ramp" production more slowly. Meaning some parts are going to stay at 45nm longer.


By masher2 (blog) on 1/16/2009 12:16:43 PM , Rating: 3
The money to implement 32nm is, by now, almost entirely sunk costs. Intel would be crazy to delay it substantially.

If anything, this means a delay in 22nm and future nodes, ones for which Intel hasn't already spent the majority of R&D costs.


By ianweck on 1/17/2009 2:34:36 AM , Rating: 2
quote:
The money to implement 32nm is, by now, almost entirely sunk costs. Intel would be crazy to delay it substantially. If anything, this means a delay in 22nm and future nodes, ones for which Intel hasn't already spent the majority of R&D costs.


Yes, just what I was going to say. 32nm is paid for. Well, development is anyway.


Clearwire
By therealnickdanger on 1/16/2009 8:38:01 AM , Rating: 2
ZOHM, WiMax, whatever they call it, the concept kicks some major arse. All I care about is getting my parents off their crappy* satellite connection and on to some real broadband. The idea of Clearwire is great, like I said, but my parents had a wireless broadband system through a local, small co-op that they would sometimes lose reception for days at a time. When it worked, it had low latency (24ping in CS:S) and a solid 1Mb/2Mb up/down.

I know some people who live in St. Cloud (Minnesota) that had Clearwire years ago... opinions are mixed. I would assume/hope that the new Clearwire service being rolled out would be reliable, fast, and affordable.

*way faster than 56K and very reliable and consistent, BUT it's too slow, too expensive, high latency, and the bandwidth cap is a rolling 17GB/mo.




RE: Clearwire
By TomZ on 1/16/2009 8:47:03 AM , Rating: 2
quote:
ZOHM, WiMax, whatever they call it, the concept kicks some major arse.
WiMax also seems to be dead-on-arrival. It doesn't seem to be catching on really.


RE: Clearwire
By therealnickdanger on 1/16/2009 9:09:37 AM , Rating: 2
Clear, ZOHM, and WiMax are all the same thing.


RE: Clearwire
By Oregonian2 on 1/16/2009 2:33:48 PM , Rating: 2
AFAIK the Intel driven Wimax has been rolled out in only two cities so far (my metro area being the second, and it's still not "up" yet, it's behind schedule) so I think the jury is out. At least locally, AFAIK it's only been used by "test folk" at Intel in Hillsboro (just west of Beaverton that's just west of Portland).


RE: Clearwire
By ironargonaut on 1/19/2009 2:28:18 AM , Rating: 2
Pure FUD. I have the service and it works just fine in Forest Grove.


very difficult to convert to 32 nm
By flipsu5 on 1/16/2009 10:08:50 PM , Rating: 2
I doubt they will take out a 45 nm fab to convert to 32 nm because there will still be product. Much more likely is the conversion of a 65 nm fab. But this will not be easy due to: 1) introduction of high-k/metal gate modules (very critical), and 2) introduction of immersion lithography (also key). This is a big jump, which AMD will luckily not have to face.




RE: very difficult to convert to 32 nm
By ianweck on 1/17/2009 2:16:57 AM , Rating: 2
quote:
and 2) introduction of immersion lithography


Introducing immersion lithography into a 65nm fab isn't a big deal (far as I can tell, the distinction in fabs is 200mm vs 300mm, not the process node). WRT the facilities side, the only difference between a dry tool and a wet one is the amount of ultra pure water supplied to the immersion machine. Adding more UPW capacity in a 65nm fab is just a matter of running the water lines, and possibly adding more UPW storage if needed. As long as the fab is 300mm, adding an immersion machine isn't much harder than adding another dry machine. Their footprint is identical in the fab, and pretty close in the subfab.


By flipsu5 on 1/17/2009 3:01:08 AM , Rating: 2
It would be ideal if it were that straightforward, but the track systems are also impacted.


They should slow 32nm
By androticus on 1/17/2009 1:57:48 AM , Rating: 2
Speeding forward with 32nm is ridiculous given the recession. No one is going to be competing seriously with them, and their current stuff on 45nm is awesome. I think this is a (bad) emotional decision based on inertia and blind adherence to continuing the pace of development -- hell, we can barely make effective use of the existing cores.




RE: They should slow 32nm
By ianweck on 1/17/2009 2:25:21 AM , Rating: 2
All things being equal, producing at a smaller node is cheaper for chipmakers; more die per wafer means more money per wafer. In this case it has everything to do with cost and very little to do with blind adherence, imho.


"Game reviewers fought each other to write the most glowing coverage possible for the powerhouse Sony, MS systems. Reviewers flipped coins to see who would review the Nintendo Wii. The losers got stuck with the job." -- Andy Marken














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki