MileHigh, and the FUD as predicted continues, while painted to be "an objective" view of the memory market, here is what Appleton and Co have to say through -Jeff Mailloux, DRAM marketing manager for Micron Technology Inc., Boise, Idaho. (lifted from the friendly MU thread)
News article received, Saturday, May 29, 1999 3:07:07 PM EST
DRAM changes challenge suppliers
May. 28, 1999 (Electronic Buyers News - CMP via COMTEX) -- The DRAM
landscape is changing in two key ways. First, the extended downturn and recent
consolidation in the industry appears to be reducing the number and increasing the size
of viable DRAM players. Second, market fragmentation for next-generation DRAM
technology is forcing suppliers and PC makers to examine the price/performance issues
of competing DRAM technologies.
Despite healthy demand, continued increases in supply have made the market
unpredictable and difficult. The rapid price declines of the last three years have been
unprecedented. In 1995, one megabit of DRAM was selling for around $3.75; today
that same megabit sells for less than 20 cents.
The consolidation that has resulted from this prolonged financial challenge could limit the
bulk of the DRAM market to three or four players, as smaller producers struggle to
maintain their capital investments.
A new landscape is also emerging with regard to interfaces and operational options.
Beyond PC100 SDRAM, the industry is facing a fragmented market. The next step in
PC performance will include an increase in the processor's frontside bus clock speed
from 100 to 133 MHz. In addition, AGP clock speeds will increase from 2x to 4x. Both
of these buses will hit the 1-Gbyte/s transfer-rate level.
Since PC100 SDRAM is limited to an 800-Mbyte/s transfer rate, a performance bump
on the DRAM bus is needed to keep pace with system buses. To this end, three
technologies are being developed-133-MHz SDRAM, double-data-rate (DDR)
SDRAM, and Direct Rambus DRAM (Direct RDRAM).
The 133-MHz SDRAM leverages existing SDRAM technology and is gaining broad
support in platforms targeted for the second half of the year due to its cost and
time-to-market advantages. There are two segments driving the adoption of PC133
SDRAM-NT servers/workstations and low-end to midrange desktop PCs.
To further speed things along, DRAM suppliers have developed DDR SDRAM, which
increases bandwidth by clocking data on both edges of the DRAM clock. This
technology also preserves the latency of standard SDRAM technology. Chipset
development is beginning for servers, workstations, and desktops, and DDR will be
widely adopted by networking-equipment suppliers.
The third option is a new memory interface known as Direct RDRAM. Later this year,
Intel plans to introduce a high-performance processor and chipset optimized for the high
bandwidth offered by the Rambus interface. Whereas past DRAM technology
transitions have been evolutionary, the move from SDRAM to Rambus is revolutionary,
particularly with respect to chip and module assembly and electrical test operations.
As a result, there is industrywide speculation that the push to Direct RDRAM will cause
the market to fragment into multiple high-volume main-memory solutions. By the end of
the year, we are likely to see concurrent high-volume implementations of PC100
SDRAM, PC133 SDRAM, DDR SDRAM, and Direct RDRAM.
With the proliferation of divergent DRAM technologies, it is paramount that component
suppliers match device capabilities with the needs of the customer. Industry experts
believe that initial price premiums for Direct RDRAM will range from 30% to 100% of
the price of an equivalent SDRAM. Value delivered for this premium must be valid and
well communicated.
For DRAM suppliers, segmentation will affect cost, price, revenue, and profitability. As
a result, traditional core competencies of manufacturing and process development must
be matched by agile design and marketing departments.
financialweb.com |
