Powerchip plans DRAM production cutbacks; The 50 nm challenge for the DRAM industry- Part II; From new SATA and USB standards to SSD and USB drive applications
Published Sep.9, 2008
Powerchip plans to cut its DRAM production--a prelude to more?
Powerchip announced during the late afternoon on Sep.8 that it plans to cut its DRAM capacity 10%-15% in Q408. Currently, Powerchip’s fab output ( P1+P2+P3, including the portion allocated to Elpida ) accounts for nearly 10% of the total worldwide DRAM output. The 10%-15% capacity cut is expected to reduce the worldwide output by 1%-2%.
Both the DRAM spot and contract prices have dropped sharply in early Sep. 08. The DDR2 1Gb spot prices have dropped near the cash cost.($1.3-$1.5, DRAMeXchange estimates), while the DDR2 512Mb has dropped to the cash cost level in Q407, thus pushing DRAM makers to shift to the 1Gb in order to further lower the cost. However, we expect to see the DDR2 1Gb to reach the cash cost soon.
Though Powerchip's planned capacity cut won't ease the oversupply too much, Powerchip hopes its action will prompt others to follow suit and slow down the migration in helping the DRAM industry to recover sooner."It costs $100 million USD to upgrade equipments and process for every 10,000 wafers to run 5xnm technology node and 5xnm won't promise profitability in the future."
In the past year, DRAM makers sped up the technology migration and reduced costs by 20%-30%. However, the prices dropped faster than the cost down. Though DRAM makers are making progress on 5xnm now, DRAM spot and contract prices have already dropped to $1.5-$1.6 -- lower than the 5xnm mature cost of $1.8. "It makes no sense to migrate to 5xnm if huge investment won't make profit. DRAM makers should cut capacity to help the DRAM industry recover ASAP" Frank Huang, PSC chairman commented.
In 2008, DRAM makers have cut capital expending up to 40%, due to huge losses in 2007. Qimonda has reduced its DRAM bit growth to 20% only in 2008 and caused its market share by revenue in Q208 dropped to 9.2% compared to its peak market share of 16.9% in Q306.
However, Samsung still guided 100% DRAM supply bit growth this year. Hynix has increased its DRAM bit growth to near 60% after transferring NAND Flash capacity to DRAM in early 2008. Elpida/Powerchip group also ramp up the JV- Rexchip to 80K per month. DRAMeXchange estimated DRAM industry to reach 65% - 70% in 2008 and oversupply remains.
For 2009, some DRAM makers still have plans to ramp up new capacity and technology migration to 5xnm. However, DRAM makers face the difficulty of fund raising amid a deteriorating economy, high debt to equity and constrained cash flow. According to DRAM makers announced plans, the DRAM industry's bit growth is estimated to be 40%+ in 2009, however, if DRAM makers cut capacity and stop the capacity expansion plans, the DRAM industry bit growth can drop to 30% , thereby helping DRAM prices to recover.
Special report
The 50 nm challenge for the DRAM industry- Part II
The key differences among all major immersion tools are the resolution and throughput. Among the leading vendors, ASML stands the most competitive position upon its technology edge in dry tool and a relatively high throughput performance. The company current occupies more than 70% market share. Rival Nikon has its immersion tool sales focused at Japanese customers with only limited shipments to US-based customers. According to buyers’ comments, Nikon’s immersion tools deliver a better alignment performance but a relatively low throughput.
Reviewing pricing among the leading immersion tool suppliers, ASML and Nikon has their difference built on procurement amount, customer relationship, after-sale service and component maintenance cost. Currently one set of immersion lithography tool for 12-inch wafer fabrication from ASML, the price is over 20mn Euros. Newer models price even surpasses 30mn Euros. Whereas for Nikon, its tool also prices above US$20mn. Regardless the deal amount nor prices, immersion lithography tool investment is undoubtedly implying a heavy burden on CAPEX.
A minimum of 5-6 processes for a single piece of DRAM wafer have to be gone through in yellow room, with some even have to undergo 8-9 processes if circuitry design is more complicated. An ideal solution to boost performance with minimum parameters adjustment and pollution is to assign a dedicated lithography tool for every single step of process. Assuming a fab houses a monthly capacity of 70,000 12" wafers with 8 lithography processes are required, an investment budget of US$140mn is needed to update existing tools because a total of 7 sets of immersion tools will be required to deliver a 140 pieces of wafer output per hour and a 22-hour daily operation.
When semiconductor geometry design migrates to sub 65nm era, replacing dry tools with immersion tools is the mainstream solution. Yet, the vast CAPEX required has prompted some players to seek for other alternatives, with double patterning being the most matured solution because production down to 32nm could be made on dry tools theoretically. Therefore, migration time to immersion tools could be postponed.
Prolonging life cycle of present dry tools come with a cost as required lithography tools amount will be double than present’s. This means that industry players still have to reserve budget for additional lithography procurement. Assuming a 70nm production line utilizes 7 lithography tool sets, another 7 sets of equipments are required when migrating the geometry to 35nm. A space issue at clean room thus surfaces when more equipments have to be installed. Nevertheless, industry players have a relatively high flexibility over dry tools procurement as they can source the equipment from second-hand market and a stronger price bargaining power is acceptable. Currently Intel, Samsung and IM Flash have deployments in double patterning, with most applied the equipment on NAND Flash production.
Memory makers have no option but to make their steps forward in terms of technology migration in order to maintain their foothold, regardless a straining CAPEX. Given that cautious considerations have to be made for equipment vendor's background, relationship with vendors, equipment performance, technology licenser's equipment preference and pricing, the considerable ASP of an immersion tool implies that equipment procurement is no longer a simple equation of price and throughput only. The decision making process is deeply involved with business thinking.
From new SATA and USB standards to SSD and USB drive applications
SATA and USB are the two most popular and important I/O interface standards used by the PC, with the former regulating data transfer speed among motherboards, HDD and optical disk, and the later regulating data transfer between internal PC component and relevant peripherals (e.g. keyboard, mouse, USB drive, etc.). A growing demand for high-definition video (in 720p or 1080p) transmission between PC and other consumer electronics has motivated the launch of new standards, which will further enhance the competitive data transmission performance in corresponding devices such as SSD and USB drive. Although the launch of new standards will have limited impacts over NAND Flash applications in the early stage, the rising penetration of such standards are expected to be seen in SSD and USB drive in the future.
Soon after the Series ATA International Organization (SATA-IO) approved the SATA 6.0Gb/s standard in mid August, Intel also introduced its USB 3.0 specification, bringing new page of the two key interfaces migration. The current mainstream PC interface standards are SATA 3.0Gb/s and USB 2.0. A maximum of 3Gb, or 375MB of data, could be transferred under a SATA 3.0Gb/s specification. Since SATA adopts algorithm that calls 8B/10B to encode and decode data for transmission over high-speed serial buses, the 1500MHz data transmission frequency is reduced by 20% to 1200MHz, meaning only 80% of the promised data transmission (300MB/s) could be kept. Whereas for USB 2.0, the suggested data transmission performance is 480Mb/s or a 60MB amount of data. But the actual transmission speed averages at 3MB/s (write) and 10-20MB/s (read) only, due to the physical specification limits of USB drives that stem from adoption of MLC Flash.
The new SATA 6G/s protocol is expected to complete standardization by the end of 2008. When the new standard is officially launched, it means serial buses' data transmission speed will be doubled to an amount of 600MB/s. The latest USB 3.0 standard as suggested by Intel is also going to bring a leap in data transmission as the present speed of 480Mb/s will be boosted to 4.8Gb/s, implying a 10 fold jump in the carried data amount.
Most SSDs now support SATA 3.0Gb/s standard. Amid the standardization of the new format, it is expected that upcoming SSDs, HDDs and optical drives in 2009 will all support the latest standard. The launch of new SATA and USB standards will have minimal impact on NAND Flash end applications, such as SSD and USB drive during the early stage because these products are more price-, rather than performance sensitive. But when vendors are able to reduce the prices of such devices, the new SATA and USB standards are expected to gain a rapid penetration as more consumers will be encouraged to save and transfer their huge AV files in such devices.
Optimistic NB shipments guidance suggests hidden risks
PC makers have posted mixed 2Q08 results. Of which, HP has beat expectations but some others have failed guidance. While some players such as Quanta remain confidence about shipments guidance for 2008,but some players have cut their guidance. One thing that is for sure is, PC demand should fail historical pattern in 2008 with some gain more market share in expense of loss from others.
Amid an economy recession in mature markets such as Europe, Japan and the US, enterprises have trimmed their IT expenditure, resulting in a weaker-than-expected NB shipments trend at a QoQ growth of 7% in 2Q08. Upcoming back-to-school is expected to fail expectation. After experiencing a rapid surge in July shipments, the growth has already slowed down in August, with uncertainties looming in September. In contrast to a minimum QoQ shipments growth of 20%, the historical pattern will be different in 2008. Leading players such as Quanta and Compal have already revised down their QoQ NB shipments growth to 15-20%.
Projecting 4Q08 outlook, NB OEMs are relatively optimistic with most expect a QoQ growth of over 20%. Therefore, global NB shipments are likely to hit 124mn units in 2008. Key factors for the growth are: 1, some shipments of new NB models will be postponed to 4Q08 as Intel has postponed launch of its new Centrino 2 platform until early July ; 2, some NB OEMs will include shipments of netbooks to total NB shipments in 4Q; 3, an easing battery supply with LG supply smoothen from September and an overall equilibrium expected in October and; 4, a relatively low shipments base in 3Q08.
A smoothen supply status and a relatively low shipments base in 3Q08 have foretold a relatively optimistic shipments trend in 4Q08. But a soft economy trend worldwide should still weigh on future demand trend. Since demand trends during Thanksgiving Day and Christmas still remain unknown, industry players are still exposing to possible risk to trim shipments. If NB vendors ship NBs without judging market status, it implies that a possible inventory shadow will be built. |
