Computer memory advances – Moore’s Law safe for at least another 2 decades
ON the heels of last months Racetrack memory development, IBM announced that it achieved a proof of principle test of anti-ferromagnetic theory that allowed the reliable encoding of a bit of information using 12 atoms. Typical magnetoelectronic drives require on the order of 1 million atoms per bit encoded.
2008 saw graphene matrix technology achieve milestones that could see 3D data arrays with extremely low “leakage” rates (‘offs’ turning ‘on’ by proximity) and very low maintenance power.
2007 saw phase change memory take off as the follow-on memory to NAND (Flash-drive) and possibly DRAM (the current memory tech in PC RAM chips).
The rapid advances on these many fronts has disrupted some plans for industry wide memory standards, with companies hedging investments in near term advances like PRAM, with mid term investment in graphene and now long term R&D into anti-ferromagnetics. This could lead back to ‘bad old days” when companies had individual, proprietary memory solutions. This will likely shake out into a well-ordered progression over the next 15-20 years, but as long term alternatives present themselves, companies will retreat from going “all-in” on near term solutions, likely delaying their introduction.
The good news is that reports of 10nm being the limit of Moore’s law are being smashed, with a tech pathway for short term “universal memory” – technology that eliminates the need for separate “active memory” (ie RAM) and “storage memory” (ie hard drives) and long term “distributed computing” where CPUs and memory on a fast network are all available as needed to support processor tasks and memory allocation tasks.