I suspect some of the delay around DDR4 finalization also relates to their eventual need to license NLST IP.
Netlist says they will be the first proprietary standard which will be adopted by the industry – for DDR4.
Perhaps JEDEC were waiting to see how Inphi (makes LRDIMMs by copying NLST IP) will fare against NLST HyperCloud - turns out Inphi has failed to challenge NLST IP - as USPTO has reaffirmed '537 and '274 patent with all claims intact - which is going to make NLST vs. Inphi a slam-dunk.
LRDIMMs are copying NLST IP.
DDR4 is even closer to NLST IP - as they have dispensed with the asymmetrical design in the LRDIMMs - and gone further and adopted the symmetrical design of the NLST IP.
The debt/loss cycle for DRAM mfgs is well known and has always been an issue. In the market where only the biggest can eventually survive (due to economies of scale), Samsung and Micron look like an interesting match. It's not clear where Hynix fits in at this point. I think Samsung strategy in DRAM may have to change further in light of this acquisition. The NAND flash/DRAM mix has already been an interesting wrinkle and new memory technologies are coming over the next five years.
Elpida did not have self-sustaining operations, financially, and it's going to be that way for a while even under Micron, who had negative earnings the last few quarters. The internal friction that is inevitable means the whole is less than the sum of its parts, initially. But that period of initial chaos is enough for Samsung and Hynix to take almost 100%.
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...