Transition to DDR3 is Upon Us

DDR2 will be with us for a while, but DDR3 is set to seize the throne as the king of DRAM.

As 2009 draws to a close, so, too does the reign of Double Data Rate 2 (DDR2) as the king of DRAM. DDR3—a memory that is at once faster and consumes less power, and that has been around for a few years—now is poised to seize the mantle of the world’s most prevalent memory technology, according to iSuppli Corp.

While DDR2 is hardly a legacy technology and won’t become one for some time to come, its pricing has increased significantly during the past few months. Nonetheless, it is DDR3 that will be the big money maker in 2010. Two main reasons account for the transition to DDR3: new Intel CPUs and manufacturing maturity.

New Intel CPUs
The transition to DDR3 is being helped by the fact that the new Intel microprocessors will require DDR3. While many of Intel’s recent chips have been capable of handling either DDR2 or DDR3, the new Nehalem-based CPUs have an integrated memory controller that can support multiple channels of DDR3 alone. For 2010, Intel’s roadmap is dominated by Nehalem-based microprocessors.

Manufacturing Maturity
Second, the technology for a crossover is ready from a manufacturing maturity perspective. There are a few components to this.

To start with, manufacturers are up the yield and learning curves and now can produce DDR3 at a competitive cost on leading-edge lithographies.

Furthermore, product manufacturing maturity is such that DDR3 is now available from multiple sources. OEMs, generally loathe to transition to a technology that is not supported by numerous manufacturers, surely will not want to sole-source a component as vital as DRAM. Thankfully, DDR3 now is available from all DRAM developers.

DDR2 Sticking Around
And yet, DDR2—with more than 3.5 billion gigabits forecasted to ship in 2010—remains a long way from being relegated to history.

Not only will DDR2 ship in significant volume through next year, it will also ship about as many bits in the third quarter of 2010 as DDR3 did in the third quarter of 2009. So, while the transition is upon us, it will not be instantaneous.

The Future of ASPs—DDR2 vs. DDR3
One key question regarding this transition is what will happen with relative Average Selling Prices (ASPs). If past transitions from Single Data Rate (SDR) to DDR, and DDR to DDR2 are considered, they would indicate that DDR3 ASPs will approximate those for DDR2 over the coming year.

Another element to consider regarding ASPs is cost. All things being equal, DDR2 yields more gigabits per wafer than DDR3 and can, therefore, be made more cheaply than DDR3. At times, however, it can seem that cost has little to do with DRAM ASPs, particularly when manufacturers sell below costs, like some did in early 2009. However, the DRAM industry is entering what promises to be a bountiful period, and DRAM makers are eager to improve their balance sheets and return to profitability—both of which can be accomplished if makers pay close attention to product mix.

A final consideration regarding DDR2 and DDR3 ASPs is product availability. The majority of DRAM makers have DDR3 products, but several suppliers are not participating extensively in the DDR3 market. All DRAM manufacturers have access to DDR2 technology, but some face challenges with DDR3 technology, particularly at leading-edge lithographies. This could lead to a relative abundance of DDR2 and thereby lower DDR2 ASPs.

iSuppli believes these supply and demand dynamics bode well for DDR3 pricing relative to that of DDR2. With strong pricing and increasing volumes, 2010 will be the year, indeed, when DDR3 takes the crown from DDR2. The king is dead, long live the king.