The storage world is abuzz this morning over a new paper at this years International Electron Device Meeting talking about recent improvements in phase-change memory. The underlying technology is cool: a bit of germanium-antimony (GeSb) alloy which can be rapidly changed from a low-resistance crystalline state to a high-resistance disordered amorphous state. So, like flash, it is non-volatile storage. Unlike flash, it shouldn’t wear out, which is very cool.
But wait! There’s more!
Touted as 500 times faster than flash, with a feature size one-third that of state-of-the-art semiconductor manufacturing, and requiring half the write power of NAND flash, non-volatile phase-change memory looks very promising – if it can beat the steep learning curve NAND flash is on. Investment trumps invention in today’s storage industry.
Meet the players
Intel has been working on phase-change memory since 2000 and Samsung, world’s largest producer of NAND flash, has also shown a prototype phase-change part and even made vague noises about commercial production in 2008.
Today’s paper, by researchers from IBM, Qimonda and Macronix, ups the ante and the buzz by bringing new entrants to what might be a new race to market.
Today’s leading-edge research against tomorrow’s years of investment
So many promising storage technologies never reach commercial production because the benefits they promise today won’t be delivered for years – while their competitors are investing millions, even billions, of dollars in new technologies, processes, factories and marketing.
For example, Samsung is forecasting cutting the price-per-bit of flash by 50% a year for the next several years. Some flash partisans maintain that flash manufacturing, not CMOS, is now driving the technology investments of chip equipment makers. All that technology and investment delivers a fast-moving target that new technologies find hard to catch.
The StorageMojo take
Nonetheless, phase-change might find a worthwhile market toehold as on-chip NVRAM, replacing and/or augmenting fast static-RAM caches, and then build from there to stay in the race with NAND flash. As promising as the technology is, I expect they’ve got at least 2 years of work to get product to market, and a lot can happen in two years.
Yet if they can pull it off, and I’m rooting for them, they’ll change storage economics in a big way. Imagine 256 GB phase-change solid-state disks that never wear out with one-tenth the power consumption of 2.5″ disk drives offering 100 times the IOPS. Even at several times the price of a rotating disk there are many applications – database and webserving come to mind – where such a device would be economic.
Storage just keeps getting more interesting. Update: The best article on phase-change I’ve seen in the press is here.