Apple turned to Intel chips, the same ones used to power most PCs using Windows, after saying its previous suppliers, IBM Corp. and Motorola Corp.'s spinoff Freescale Semiconductor Inc., couldn't meet Apple's needs for faster, more energy-efficient chips.
I wonder why AAPL is so unhappy with IBM;
IBM seems to have its finger in every pie but still IBM's semiconductor business appears dismal
I'm not familiar with ISTA but I doubt toolsets are matching though IBM makes probably recommendations but there's quite a bit autonomie when it comes to selection.
APM sounds great but what about multiple manufacturing sites? if "a problem" shows up at one site but not at the other site?
In one instance last decade, AMD discovered that circuits were peeling on its chips. An investigation uncovered that a person in purchasing switched suppliers of a particular chemical. Though technically the same chemical, there was an extra atom in the new stuff, causing problems.
I don't follow this at all (in particular making it a case for APM) because this chemical problem is a prime example for "copy exactly" which would dictate to use the same vendor for this particular chemical at all manufacturing sites.
I think in case of (yield) problem arises AMD would be more open to make a change to fix the problem while Intel would do everything to figure out the rootcause for the problem and figuring out the rootcause makes more sense to me when you have several manufacturing sites because this particular problem might only show up at one site but not at the other.
So in case every manufacturing site would select its own fix to the problem the process overall would drift away.
Under APM, AMD can tweak the manufacturing recipe of a single wafer as it winds through the entire production process, which takes weeks. In the past, semiconductor makers had to run several wafers, look at the results, and then adjust the formula. Imagine every time the (post etch) CDs would change adjusting the etch process to compensate - and there are hundreds of processing steps.
I suppose the future will tell
New AMD factory ups the ante against Intel
Michael Kanellos
CNET News.com
October 14, 2005, 11:45 BST
Talkback
Tell us your opinion
A massive fabrication facility opening this week could help AMD catch its white whale - 30 percent of the PC chip market
<IMG SRC="http://www.silicon.com/i/s/ads/ha/etf/mpu.gif" WIDTH=336 HEIGHT=280 BORDER=0 ALT="Click here!" USEMAP="#ad_etf_govukitsummit_mpu">
For years, AMD has dreamed of nabbing 30 percent of the market share for PC microprocessors. Though achieving that goal won't be easy, a new fabrication facility opened in Germany on Friday gives the company the real estate to try.
Fab 36, located next door to an existing fab outside the city of Dresden, adds 13,400 square metres of clean-room space to AMD's manufacturing arsenal. Combined with factory capacity at Chartered Semiconductor that's available to AMD under an existing alliance, AMD will have a substantial part of the infrastructure needed to churn out 100 million processors a year by 2008, said Daryl Ostrander, senior vice president for logic, manufacturing and technology.
Running full tilt, that should be enough to hit the elusive number, or at least help AMD participate in all segments of the PC market. "We can get to 30 percent market share from Chartered and Fab 36," Ostrander said. "We have become a manufacturing powerhouse."
Roughly 1,000 people will work in Fab 36.
Though chip designers often get most of the attention and glory in the industry, semiconductors rise and fall through manufacturing. Efficient manufacturing techniques and its meticulous "copy exactly" philosophy for building fabs have been significant pillars in Intel's rise, according to, among others, chairman Craig Barrett and "employee No. 4" Les Vadasz.
Conversely, AMD often stubbed its toe in the past in getting chips out of the factory, resulting in product delays, chip shortages and huge financial losses. In terms of manufacturing, AMD has typically done better than most other chipmakers, according to Dan Hutcheson, chief executive of VLSI Research, one of the chief research houses on semiconductor manufacturing. It was simply pushing too hard to stay up against the mass-production monster that is Intel.
In one instance last decade, AMD discovered that circuits were peeling on its chips. An investigation uncovered that a person in purchasing switched suppliers of a particular chemical. Though technically the same chemical, there was an extra atom in the new stuff, causing problems.
"Microprocessors are one of the most difficult things to manufacture. It is easy to fall out of bed," Hutcheson said.
The situation began to change in the late '90s under executives such as Bill Siegel. Now the company wins citations from, among others, Sematech, an industry trade group.
The secret sauce in AMD's production is a methodology the company calls Automated Precision Manufacturing (APM). Under APM, AMD can tweak the manufacturing recipe of a single wafer as it winds through the entire production process, which takes weeks. In the past, semiconductor makers had to run several wafers, look at the results, and then adjust the formula.
As a result, AMD can get to what the industry calls "mature yields", or the situation where the majority of chips in a given wafer work. APM also lets the company ramp up or decrease output of specific chips while in the middle of a production run to better suit what's happening on store shelves.
"APM is allowing them to identify issues with yields much quicker, and once you fix a problem, it stays fixed," Hutcheson said.
Success, however, is not foreordained. Fabs cost billions to erect and fill with equipment. Roughly $2.5bn (£1.4bn) will get invested in outfitting Fab 36 through 2007, according to the company. An industry downturn or product delays can turn large factories into liabilities.
Intel will also likely continue to erect fabs. Another risk: Chartered. Turning a profit on PC processors through foundry arrangements has not been easy to date.
To ameliorate some of the financial risks, AMD will not completely build out the facility just yet. As it stands now, AMD can start production on 13,000 silicon wafers a month. Enough empty space, however, exists to crank that up to 20,000 wafer starts a month. Getting to the 100 million mark will involve populating the current empty space in the 13,400 square metre plant with equipment.
The wafers from Fab 36 will have 300-millimetre diameters. Right now AMD produces chips on 200-millimetre wafers. A wider wafer means more chips without a huge increase in costs. Intel, IBM and others have already graduated to 300-millimetre production.
The facility will produce chips made on the 65-nanometre process, which will hit store shelves next year. Later, AMD will churn 45-nanometre and 32-nanometre chips out of the facility.
Ostrander said AMD has not fully formalised plans on what to do with Fab 30 next door. It may retrofit it for 300-millimetre manufacturing, or it may use it for another purpose.
The 30 percent mark has been a white whale for the company for years. Founder Jerry Sanders often set it as a goal, and always fell short. AMD currently has about 17 percent of the market, though at one point a few years ago it commanded a little more than 21 percent.
One of the changes current chief executive Hector Ruiz brought to AMD was not to mention the figure early in his tenure. The number, however, has been cropping up more recently in speeches from AMD execs.
Why did AMD build in Germany, a company known for tight labour laws and high-priced employees? Tax breaks and subsidies. The vast majority of the costs go into semiconductor manufacturing equipment, which roughly costs the same all over the world. Labour costs are a minor factor in the multibillion-dollar budget.
Thus, the most significant variable is the amount local governments are willing to pay to bring in a what will become a major economic engine for them. By 2007, AMD and the local government estimate that the two fabs will have created 7,500 new jobs, directly and indirectly, in the former East German city.
Government loans and subsidies come to about $1.5bn, according to AMD.
IBM, Samsung, Infineon, Chartered describe 65-nm process
IBM, Samsung, Infineon, Chartered describe 65-nm process IBM, Samsung, Infineon, Chartered describe 65-nm process.
Infineon begins 'fab-lite' era with 65-nm samples
Christoph Hammerschmidt
EE Times
(01/31/2006 7:09 AM EST)
MUNICH, Germany — As a result of working with R&D alliance partners IBM, Chartered Semiconductors and Samsung, Infineon was able Tuesday (Jan. 31) to demonstrate its first samples of logic ICs made using a 65-nanometer manufacturing process.
At the same time Infineon made its first steps towards operating as a “fab-lite” chip company, using 65-nm technology on a foundry basis. Wolfgang Ziebart, Infineon CEO, had said late in 2005 that the German chipmaker it did not plan to manufacture logic chips for itself below the 90-nm manufacturing node (see Dec. 1, 2005, story). The logic chip samples were designed as a test of Infineon's 65-nm production process and comprise, among others, an ARM9-based microprocessor and a DSP core as material components of mobile handset baseband chips. Beyond this, the chip contains several digital cell libraries and macros as well as SRAM, ROM, RF and mixed-signal circuitry.
Initially, Infineon plans to introduce logic elements in 65-nm technology for its mobile phone sector. “In this segment we have the best chances to get significant production volumes quickly,” said Heinz Schuetzeneder, vice president and overall project leader for Infineon's 65-nm platform.
The first products are scheduled for delivery as samples in the current quarter with volume production ramp-up due to start in Q4 of 2006. The 65-nm technology is going to be used for the production of ultra low cost handset chipsets, with RF and digital circuitry integrated on the same die, Schuetzendeder said.
The test chips were produced first at IBM’s facilities in East Fishkill, New York, and then the same tape-out was used to reproduce the chips at a Chartered wafer fab in Singapore. “While the fabs use slightly different equipment sets, the samples showed comparable results at the first go,” Schuetzeneder said. “Thus, we can do mass production in both fabs.”
In contrast to the forecast of many semiconductor experts, the ratio of leakage current to speed did not worsen compared to 90-nm technology, Schuetzeneder stressed.
According to strategy outlined last year by Infineon CEO Ziebart, the company does not intend to manufacture the integrated circuits itself. Instead, the company wants to have them manufactured by Chartered Semiconductor. In this way Infineon can get access to modern production technologies without having to make large capital investments in a dedicated 65-nm fab, the Infineon manager explained. “This brings us a lot more flexibility at lower cost,” Schuetzender commented.
Nevertheless, for its DRAM fabs in Dresden, Germany and Richmond, Virginia, the company has a different blueprint. In these wafer fabs, conversion to 90-nm manufacturing processes is under way. Schuetzender said that more than 50 percent of Infineon's 300-mm fab capacity in Dresden is already running 90-nm processes.
Meanwhile Infineon is cooperating with Nanya Technology Corp. (Taoyuan, Taiwan) on the development of a 70-nm DRAM process. Schuetzeneder said that 70-nm DRAM samples had been produced several months ago and that the introduction of this DRAM technology node is on schedule. |
