The great race: Flash vs. magnetic storage
09/01/2007
Probably one of the most exciting technology races of all time is now underway, and all of us will benefit from the super efforts of both contenders. Unfortunately, most of us know little about the giant steps in technology that are making this race so scintillating. This issue of Data Storage (a supplement to Solid State Technology) helps tell the story.
On one side are integrated circuit designers (chipmonks, as Ted Nelson, the father of hypertext and one-time Autodesk fellow, calls them), who are steadily shrinking flash memory cells while doing multilevel read/writes. They offer a speed as well as a size advantage over disk drives.
On the other side are the much less heralded magnetic storage designers, who exceeded the pace of Moore’s Law for a few years, allowing them to pass 100 Gb/in2 storage densities while heading toward a terabit per square inch on the media surface. They continue to hold a sizable cost advantage with rugged miniature multigigabit hard drives for portable equipment (miracles in themselves), but to push storage densities even further, they face what is called the superparamagnetic limit, where thermal agitation might flip the magnetic bits on a disk.
In the race for the lead, flash designers now plan to move in a new direction: Z. Samsung showed how it can stack flash memory cells on a 3D CMOS chip at the 2006 International Electron Devices Meeting (see “Electron Devices Meeting goes 3D, low-power,” SST, February 2007, p. 24), using common source lines to simultaneously erase a 32-bit string at a time. Presenter Soon-Moon Jung claimed that Samsung sees no reason why they couldn’t stack eight layers of 32-bit flash cells, achieving terabit flash memory chips! Samsung claims that disk drives will no longer be needed for storage in notebook PCs and other portable devices, since so much flash can be economically crammed into them, cutting power drain as well.
Not so fast, say the ever-innovative magnetic storage mavens. They see two routes to pushing disk densities even beyond a terabit/in2. One is to actually pattern the media into isolated patches of a few magnetic grains each, perhaps providing the first large-volume commercial application for nanoimprint lithography. That approach was described by Z.Z. Bandic, et. al., Hitachi San Jose Research Center, in the September 2006 Data Storage supplement.
As is so often the case in frontier technology, another trick might allow even tinier grain patterns to be feasible by turning potential thermal defects into a feature. Using a laser beam and near field rather than more familiar diffraction limited optics, highly localized heating to a few hundred degrees Celsius can be achieved. Sharp local field gradients can be created that weaken or destroy ferromagnetism (above the Curie temperature of the magnetic material) and then restore it with rapid cooling on a subnanosecond time scale.
Field gradient recording in this thermal assist mode is much stronger than can be achieved with conventional magnetic heads, and bit sizes of 10-12 atoms on an edge appear feasible. This approach, heat-assisted magnetic recording (HAMR), is presented in this issue (p. S8) by Terry McDaniel and Michael A. Seigler of Seagate Research.
So who will win this great race? Maybe both teams, suggests Jim Handy of Objective Research (p. S16). Why not use hybrid disk drives incorporating flash chips to gain the tremendous benefits of both rapidly advancing technologies? It appears that this may be an option chosen for many portables in the near future.
Racing fans love a dead heat!
 |
Robert Haavind
Editorial Director