Previously, DailyTech reported on how DDR3 DRAM produced on 50nm production lines would be much cheaper and faster than those being produced on current production lines.
Elpida, Samsung, and even beleaguered Qimonda are either in the process of or planning to transition to these new geometries in order to lower costs and increase profits. DDR2 DRAM is currently selling at below the breakeven point for most DRAM manufacturers.
Samsung, the world's top DRAM producer, is announcing that it has developed a 1GB DDR2 SODIMM using 1Gb DDR2 DRAM on a 40nm process. The process size refers to the average half-pitch of a memory cell. A smaller die size means that more dies can fit on a silicon wafer, reducing production costs. Samsung expects power savings of 30 percent if lower voltages are used, made possible by the smaller size.
A smaller size also means that higher memory densities can be introduced, such as Samsung's 4Gb DDR3 chips. These can be used to produce 8GB DIMMs and SODIMMs. Samsung said it plans to apply its 40nm technology to develop a 2Gb DDR3 device for mass production by the end of 2009, lowering DDR3 prices even further.
Intel will be launching its mainstream Nehalem products in the third quarter of this year. Intel will be also be using DDR3 exclusively on its 32nm Westmere CPUs while AMD's Socket AM3, which uses DDR3, is coming out soon.
With DDR2 prices depressed, it makes economic sense for Samsung to concentrate on ramping DDR3 production on 40nm. Applying the economies of scale, the price premium of DDR3 could drop from 100% to 10% by the time Lynnfield and Windows 7 launch together in Q3.
The 40nm process can also be adapted for a new generation of GDDR5 chips, which will be used with AMD and NVIDIA's DirectX 11 GPU parts to produce cheaper and faster video cards.
Samsung is also currently researching 32nm process technology for future DRAM production.