Thursday, April 26, 2012

Intel: "Fabless model collapsing". Is it correct?

Intel’s top process technology exec says the foundry model is collapsing.  A good manufacturing process is only one of the factors that leads to success in product and circuit strategies. Intel had been leading in manufacturing with the HKMG process since the 45nm process.

However, they were not able to use their manufacturing clout to lead in new areas such as the mobile phone microprocessor market.

TSMC, GlobalFoundries, and UMC will have to work closer with the fabless companies sharing resources and cost to address these challenges.

Ron Maltiel

Friday, April 20, 2012

Qualcomm and Nvidia 28 nm Wafers? Shortage...

Last quarter results from Qualcomm show again the impact of Moore's law breaking down. As detailed below, Qualcomm growth this quarter was limited by the supply of 28 nm wafers. Nvidia has been complaining about similar type of issues in the last few months.

It all is really tied to the bigger issues of increasing cost of developing and producing new process technologies. This fact is reducing the number of leading edge fabs.

Fabless companies to ensure their supply will need to share more of the development costs in order to gain higher priorities.

More information on foundries, fabs, and wafer supply is available at Forecasting Wafer Demand: Technology Migration, Bottlenecks (Link includes a chart of wafer demand by IC product type (i.e. DRAM, NAND, NOR, MPU, PLD, etc))

Ron Maltiel

Monday, April 16, 2012

Intel, AMD, or ARM Servers?

The article below about low-power servers doesn't mention Intel's large growth in database servers.  Intel had more than $10 billion in revenues from ICs sold into data centers, including servers, storage products, and networking.

Considering the high power usage by databases, it is not surprising that Intel wants to dominate the low-power servers market.  It is likely that databases and cloud servers  (20% of sales, 3x PC segment growth) added to Intel's growth in 2011.

Ron Maltiel



Intel to Face Off Against AMD, ARM with New Low-Power Server

Intel may think extremely low-power servers are only a small part of the server market, but that isn't stopping the company from competing in this segment.

At its Intel Developer Forum in Beijing, the company said it will release a six-watt, dual-core processor known as Centerton, based on a dual-core Atom design, in the second half of this year.
Atom is Intel's low-power core design and the company has been pushing variants of it for everything from smartphones and tablets to set-top boxes and netbooks. Centerton will be a 32nm, dual-core system-on-chip (SoC) that will draw six watts of power. This will be a 64-bit chip with support for the large amounts of memory that many server applications require.

The microserver market is particularly interesting because of a theory that massive "scale out" deployments—typically large Web farms—would do better with a larger number of physical cores if those processors used much less power than traditional servers. SeaMicro was perhaps the biggest early advocate of this process, coming out with its SM10000 server using up to 512 Atom cores and its proprietary fabric for connecting multiple processors, along with the associated memory and input/output. Last year, it upped the ante by switching to 64-bit Atom N570s.

SeaMicro was acquired by AMD earlier this year, and the company is widely expected to start using that fabric with its own processors and offer the technology to its OEMs.

Meanwhile, a number of companies—notably Calxeda, with its EnergyCore ARM processor— have been talking about using multiple ARM-based cores to compete in the microserver market. The concept is good, and Calxeda says its processor can draw as little as 1.5 watts for a dual-core server, although it is 32-bit only. This, too, is expected to go into real production in the second half of this year.

ARM has announced a 64-bit architecture that many vendors are expected to embrace, but 64-bit ARM cores aren't expected to be available for volume production until 2014.

Intel also said it will be shipping a new version of its Xeon E3 processor based on the Ivy Bridge architecture, manufactured on a 22nm process using "tri-gate" transistors. These chips are aimed at slightly larger servers, usually a single socket 1 U rack or blade server that effectively is the equivalent of a high-end desktop. Currently, Intel offers 45- and 25-watt versions of the 32nm Sandy Bridge-based E3s, along with a 15-watt Pentium 350. It hasn't yet listed power requirements for the next generation E3s, but suggests the 22nm process will be more power efficient.

These should be competing more with AMD's recently announced Opteron 3200, which is meant to be a lower-power, low-price variant on the company's Opteron 4200 chip, based on the Bulldozer architecture. The four-core versions of these chips are rated at 45 watts.

Whether from Intel, AMD, or one of a variety of ARM providers, we're seeing more competition in the server market, particularly in the low-power segment. That's leading to lots of innovation and the potential for companies to save a lot on power bills.

Wednesday, April 11, 2012

Top 25 2011 Semiconductor Sales Ranking

IC Insights'  report of the top 25 semiconductor companies (see below) states that Intel grew 24% in 2011 due to the purchase of Infineon's wireless segment. It is likely that the growth of databases and cloud servers (20% of sales, 3x PC segment growth) also added to Intel's growth.

It is important to note that the semiconductor sales and growth rankings do not include SanDisk's $5.66B sales in 2011. See more details about SanDisk's results from 2008 to 2011 at Is Apple Squeezing Suppliers, Or Is SanDisk Simply Missing Out?

Ron Maltiel

Extreme Results in Top 25 2011 Semiconductor Sales Ranking!

ON Semiconductor and Elpida represented an 89-point swing in 2011 growth rates.
Although only 2% growth was registered in the worldwide semiconductor market (including ICs and Optoelectronics, Sensors, and Discretes; O-S-Ds) in 2011, several companies posted results that were far different (Figure 1). Boosted by its acquisition of Sanyo’s semiconductor business, ON Semiconductor registered a 49% sales increase last year and moved into the top 25 ranking for the first time! Qualcomm, spurred by a 73% increase in smartphone unit shipments last year, logged a strong 38% increase in semiconductor sales in 2011. At the other end of the spectrum was Elpida, which registered a steep sales drop of 40% (45% when expressed in yen)!

Although 15 of the top 25 semiconductor sales leaders posted negative results last year, in total, the top 10 and top 25 semiconductor suppliers grew 7% and 4%, respectively, in 2011 as compared to 2010. These growth rates were much better than the 2% increase shown in the worldwide 2011/2010 semiconductor market. A listing of the top 50 semiconductor suppliers of 2011 will be included in IC Insights' April Update, which is part of the subscription to The McClean Report.

As shown, Intel remained firmly in control of the number one spot in the ranking last year. In fact, Intel, helped by its acquisition of Infineon’s wireless IC business, extended its lead over second-ranked Samsung by registering 48% more in semiconductor sales than Samsung in 2011 as compared to a 24% margin in 2010.

Healthy growth in its graphics and communications processor business helped Nvidia jump five positions and move the company to 18th place in the top 25 ranking. In contrast, Elpida fell six spots in the ranking last year (from 13th to 19th) as the collapse in the DRAM market had a disastrous effect on the company. In fact, Elpida lost almost $1.2 billion in the second half of calendar 2011 and filed for bankruptcy in 1Q12.
In contrast to 2010, memory companies did not secure the top growth rate positions in the ranking in 2011 (Figure 2). In fact, as shown below, non-memory suppliers logged the top six growth rate increases last year. Moreover, in the case of ON, Qualcomm, and Intel, 2011 semiconductor sales growth rates were significantly boosted by company and/or business segment acquisitions.

1) ON Semi; 49% jump primarily due to the acquisition of Sanyo’s semiconductor business
2) Qualcomm; 38% jump due to surging smartphone IC sales and its acquisition of Atheros
3) Infineon; 29% growth from continuing operations (23% using 2010 exchange rates)
4) Intel; 24% increase spurred in part by its acquisition of Infineon’s wireless business
5) Nvidia; 10% growth due to healthy sales of graphics and communications processors
6) TSMC; 10% (2% using 2010 exchange rates)

Of the big five memory suppliers in the top 25 ranking (i.e., Samsung, Toshiba, Hynix, Micron, and Elpida), only Samsung registered 2011/2010 growth. In total, only 9 of the top 25 suppliers (including Infineon’s 29% sales growth from continuing operations) outperformed the total worldwide semiconductor industry 2011/2010 growth rate of 2%.

Figure 1

Figure 2

Friday, April 6, 2012

TSMC 28nm Capacity Large Shortage

Nvidia has been concerned about their relationship with TSMC for awhile.

It seems that their yield for the 28 mm process is too low. There is a reasonable limit of how many additional wafers they should manufacture to cover shortfalls of good dies coming out of the fab.

Ron Maltiel

TSMC 28nm capacity in large shortage
Monica Chen, Taipei; Adam Hwang, DIGITIMES [Friday 6 April 2012]

Taiwan Semiconductor Manufacturing Company's (TSMC) 28nm foundry capacity has been drastically short of demand from Qualcomm, AMD and Nvidia mainly, but the shortage is expected to relax at the end of the third quarter of 2012, according to industry sources. Qualcomm, in view of the shortage, has shifted some orders to United Microelectronics, but has been unable to meet its clients' demand for processors for smartphones and tablet PCs, the sources indicated. AMD launched the 28nm-based Radeon HD 7970 in the first quarter of 2012, but has actually shipped a relatively small volume of the GPU due to TSMC's short 28nm capacity, the sources noted. Nvidia launched only one 28nm-based GPU, GeForce GTX 680, in late March and has had to delay the launch of Kepler series GPU models due to the shortage, the sources said.

While yield rates of its 28nm process are slowly improving, TSMC is conservative about expanding 28nm foundry capacity in order to maintain gross margins, partly accounting for the capacity shortage, the sources said. In related news, TSMC will start construction of the 5th-phase expansion of Fab 14, its 12-inch fab located at the Southern Taiwan Science Park, on April 9. TSMC's total foundry capacity in 2012 is expected to increase by 10% from 2011

Thursday, April 5, 2012

Flash is not just about storage

Adding NAND Flash as a new cache memory between the microprocessor and the hard drive can also increase computer system speed for either a laptop or a PC. There have been rumors that Intel was developing such a system though it likely was been postponed due to software development difficulties.

Ron Maltiel

Fusion-io CEO: Flash is not just about storage
Companies like EMC, that build Flash into their storage arrays, are
missing a trick, says David Flynn
By Sophie Curtis | Techworld | Published: 11:57 GMT, 04 April 12

The real advantage of high density NAND Flash is more its ability to
carry out application workload acceleration than its capacity for
storage, according to Flash storage vendor Fusion-io.
Speaking to Techworld last week, the company's chief executive David
Flynn said that companies like EMC are approaching the problem from
the wrong direction, integrating Flash into their storage arrays to
speed up the read/write process, rather than harnessing its memory
Fusion-io has built a storage memory platform that improves the
processing capabilities within a data centre by relocating
process-critical, or "active" data from centralised storage arrays to
the server where it is being processed – a method known as data
"The thing that has been the choke point or the constraint on how much
useful work you can get from a server hasn't been the amount of
processing, but the amount of data that you can feed it," said Flynn.
"RAM is a way to feed it data, and disks are a way to feed it data,
but RAM is too small and disks are too slow, so either way you can't
get large quantities of data into the processor fast."
Flynn said that the company's ioDrives have a hundred times the
capacity of a memory module and thousands of times the performance of
a disk drive. The devices plug directly into the server, meaning that
the server can act as primary storage and access applications faster.
"I'm not saying it gets rid of your disk drives or memory, it allows
those to be optimised – one for lots of capacity and the other for
performance," said Flynn. "So you still have RAM, you just don't need
as much as it. You still have disks, you just don't need as many
Flynn said the power is in being able to purchase performance
separately from capacity, and scale the two independently. With
today's storage arrays, improving performance means buying lots more
mechanical disks, and you get capacity even if you didn't ask for it,
he said. Satisfying those needs separately is cheaper, because you
don't end up overbuying capacity in order to supply performance.
This is particularly significant in a virtualised environment, where
the savings made on server consolidation are often dwarfed by the
additional spend on the storage arrays. By plugging a Flash drive into
a server, the data from the storage area network (SAN) is cached, so
it is no longer necessary to keep adding disks to get performance.
"The biggest limiting factor on how many VMs you can put on a server
is how much memory you put in it. If you had a huge amount of memory,
you wouldn't care about your storage performance," said Flynn. "This
gives you that memory capacity, so you don't have to worry about the
performance in your storage and can size it solely for capacity."
By giving companies the ability to add performance from outside of the
SAN, where vendors control the markup, they are able to scale things
more cost effectively and benefit from buying more software licenses,
added Flynn.
"This isn't just about the fundamental change of the medium, it's
about the change of the market to an open market," he said. "The open
models win. This is how we will compete with the likes of EMC."

Wednesday, April 4, 2012

Six Technologies that Save Smartphone Batteries

These six new technologies are interesting approaches for cutting down the power used during battery operation.
Some of the technologies, such as operating near the threshold voltage, would be difficult to manufacture consistently in the fab with reasonable yield.

Ron Maltiel

Six Paths to Six Technologies that Save Smartphone Batteries
By Rachel Courtland / April 2012

The boom in mobile devices and data centers has circuit designers
racing to find new ways to slash power consumption. At this year's
International Solid-State Circuits Conference, in San Francisco, six
power-saving technologies took center stage. Some will emerge in
products this year, while others are just beginning to catch the
interest of major chipmakers.rigueur for today's processors—all the
way down to 280 millivolts. The chip's sweet spot for energy
efficiency was 450 mV, just above the threshold voltage. At that
level, Intel's chip ran slowly, at less than 100 megahertz, but it
also consumed just about a fifth of the energy it did at 1.2 V.
Parallel processing could be used to pick up some of the slack in
performance.Mudd College, in Claremont, Calif., showed that the
approach works on an ARM Cortex-M3 processor, boosting energy
efficiency by 60 percent. The team says it's the first implementation
of a Razor-style scheme on a complete commercial processor. nm Ivy
Bridge chips can be run just as fast as the company's previous chips,
but with an operating voltage that's 200 mV lower. Intel has also
incorporated designs at the circuit and core level to improve the
chip's power management. A separate system-on-a-chip code-named
Silvermont, based on the same transistor-making process, will be
geared for mobile handsets. Samsung presented a new improvement on the
all-digital phase-locked loop, a 0.012-square-millimeter circuit that
consumes just 2.5 mW. Intel showed off a version of the circuit, built
with the company's 22-nm technology, that consumes as little as 0.7
mW.Chiao Tung University, in Hsinchu, Taiwan, found a way to save
power by working on signals in two separate stages—one for crude
processing and the other for fine-tuning—that can each be
optimized.Samsung and Hynix Semiconductor both unveiled details on the
next incarnation of synchronous DRAM, the memory that drives today's
processors. The new generation, which goes by the name DDR4, boasts
circuit tricks that let Samsung drop the supply voltage to its memory
modules from 1.5 V to 1.2 V. The modules also include better clocking
and faster algorithms for encoding data to be sent to and fetched from
memory. DDR4 may make its commercial debut as early as 2013.

Monday, April 2, 2012

The Real Growth Paths for Flash Memory?

The article below addresses flash memory growth in SSD as a replacement for hard drive, which is one of the main growth areas for flash.

However, some other very large growth areas for flash memory are cell phones and consumer products.
For example, in the new iPad from Apple, the cost of  NAND Flash is between 5-18% .

April 24 Update: Apple's quarterly results highlight the impact that cell phones are having on the growth of flash.   "Surging iPhone Sales Propel Apple's latest quarter results". Total revenue of $39 billion...nearly 75% of revenue derived from the iPhone and iPad

Ron Maltiel

What are the Real Growth Paths for Flash Memory?
Tom Coughlin, 3/29/2012

Many pundits predicted that the shortage of hard disk drives would cause a significant shift to SSDs last quarter and this quarter. But as indicated in Micron’s report on its F2Q results, many OEMs ordered what became excess inventory of SSDs that will take another quarter to use up. The basic reason is economic, even with HDD supply much less than demand and higher HDD prices the price of a given amount of flash memory capacity is still much higher than that of HDDs.
Currently even the lowest price SSDs sell for about $0.70/GB while the most price impacted HDDs sell for $0.14/GB (a difference of 5:1). As the available production volume for HDDs recovers the price of HDDs in real dollars will continue to fall. In addition if higher areal density HDDs are introduced later this year, capacity prices will drop to pennies per GB while the least expensive SSDs will likely drop to $0.50/GB by the end of 2012. Economics is an important consideration for consumer and business purchases, and price does matter.

Over the next year or two there will be additional consumer and business products that shift from using HDDs to only using SSDs or flash memory. In mobile devices and automobile applications the ruggedness of flash memory and its ability to be put into smaller volumes than HDDs provide other advantages to users even if the purchase price is greater. In addition, many mobile applications have limited local storage (to control the product price) and depend upon storage and other resources from “the cloud.” Mobile consumer electronics will be a major growth area for flash memory, but likely not as much for traditional SSDs.

SSDs are storage devices that contain flash memory chips but also a separate storage controller that manages wear-leveling, memory management and interface control. It is likely that many future flash memory implementations in mobile devices will incorporate the controller into the overall system electronics and the storage device will be a collection of flash memory chips with no dedicated controller. This helps to reduce the overall system cost as the controller is part of the system electronics but also allows tighter integration and proprietary control of the controller functions by the mobile device designer. These considerations are likely a major reason for the purchase of Anobit by Apple.

The growth areas for SSDs will primarily be in two areas. In client computing applications, such as Ultrabooks, SSDs will be used alone in the most expensive machines while less expensive Ultrabooks will use a combination of a HDD and an SSD, or else flash memory cache in the HDD (a hybrid HDD), or possibly on the motherboard (although this seems less likely). These combinations of flash memory with a HDD provide performance boosts like those with pure SSD computers but offer the lower cost capacity of HDDs and will allow sub-$600 Ultrabooks, which are likely to be the most popular price point.

The other growth area for SSDs is in enterprise applications where SSDs can provide fast transaction processing, partly to support cloud services and storage in “the cloud.” Although HDDs (and even magnetic tape) will continue to supply inexpensive mass storage, intelligent storage tiering using SSDs allows must faster access to content and at lower expense than a DRAM-based solution. There are currently many storage systems available by every major enterprise storage vendor offering SSDs as part of their storage tiers and several companies offering pure flash-based appliances. These companies include EMC, HDS, HP, Oracle, NetApp, Nimble Storage, Texas Memories, X-IO and many others.

The combination of SSDs and flash memory in mobile devices and in remote data centers supporting cloud-based storage and services is important enabler of today’s information economy. However these performance and mobile applications would be very limited without access to low cost content and information storage on magnetic storage devices. Thus in a real way flash memory and HDDs are much more symbiotic than might initially appear to be the case . In a real way the growth of flash memory and SSDs is dependent upon the growth of HDD storage and likewise the growth of HDDs is enabled by the faster data access enabled by flash memory.

New Foundry Ranking (By IC Insights)

IC Insights data (in the link below), has details of the additional $1.975 billion that Samsung manufacture for Apple.
Foundry Rankings (Including Samsungs' iPad, iPhone Breakdown)

Samsung's foundry breakdown shows how much of its business is tied to the iPad, iPhone, and iTouch products. For example, in 2011, Samsung's foundry produced $1.975 billion worth of chips. Apple products were $1.5 billion of the total chips manufactured by Samsung Fabs.

Ron Maltiel

IC foundry market grows 5.1% in 2011, says Gartner
Jessie Shen, DIGITIMES [Friday 30 March 2012]

The worldwide semiconductor foundry market totaled US$29.8 billion in 2011, a 5.1% increase from 2010, according to Gartner. The semiconductor supply chain experienced some impact from the Japanese disasters and Thailand flooding. However, without the steep depreciation of US currency, foundry growth in 2011 would have been just 0.7%, Gartner said.

"Thanks to stable media tablet and mobile phone sales, a slide of the semiconductor and foundry revenue in 2011 was prevented," said Samuel Tuan Wang, research director at Gartner. "After 40.5% growth from 2009 to 2010, the foundry market maintained relatively flat business in 2011 due to the weakness in PC production and an overall consumer demand hit, as well as a leaner inventory practice by customers that started in mid-2011."

Consolidation and domination of business continued. The top five foundry players accounted for almost 80% of the foundry market share, with the top player, TSMC, expanding its revenues over 2010 and reaching 48.8% share in 2011, Gartner indicated.

Samsung's foundry, with US$470 million in revenues, ranked No. 9. However, Samsung Electronics had been very aggressively expanding its LSI business in 2011. Had the estimated US$1 billion Apple wafer business been included in its foundry revenue, Samsung would rank as high as No. 4 in the foundry ranking, Gartner observed.
Powerchip Technology had a nearly threefold increase in foundry revenues in one year due to the strategic decision to shift from the commodity DRAM business to foundry in early 2011, Gartner disclosed.

Communications, consumer and data processing continued to be the three key applications driving the foundry business; they accounted for 42.7%, 20.9% and 20.3% of the foundry revenues, respectively, in 2011, according to Gartner.

Fabless customers contributed 77.8% of the foundry business, IDMs contributed 20.2%, and the remaining came from system companies. By region, America's customers generated 62.8% of the foundry revenues, Asia/Pacific 22.2%, Europe 10% and Japan 4.9%.

"Given the aggressive capital spending by large foundries during 2010 and 2011, the oversupply of foundry capacity was inevitable," Wang continued. "The utilization rate for foundries continued to decline quarter to quarter in 2011, causing the annual average utilization rate to drop to 81% from 91% in 2010. Advanced technology for mobile applications was the driver for the growth of foundry business in 2011, and the demand is expected to remain high during the next few years."

Gartner: Top 10 foundry firms revenues and market share, 2010-11 (US$m)
2011 Rank
2010 Rank
2010 sales
2010 market share
2011 sales
2011 market share
IBM Microelectronics
Vanguard International
Dongbu HiTek
Powerchip Technology
Top 10 for 2011
Total Market
*Note: Samsung revenues do not include ASIC business from Apple
Source: Gartner, compiled by Digitimes, March 2012