GCN LAB IMPRESSIONS
IBM claims breakthrough with 'liquified' memory
IBM already has pretty good memory for a hundred-year-old, but now it has gotten even better. The company’s research division in Zurich, Switzerland, just announced a breakthrough that will enable manufacturers to use a type of memory that was previously deemed too unreliable for widespread usage.
Researchers said their tests of a new approach to phase-change memory showed it performed 100 times faster than the most advanced flash memory, and that was in a worst-case scenario. PCM also can last for at least 10 million write cycles, compared with 30,000 cycles for enterprise-class Flash and 3,000 cycles for consumer flash drives, IBM said.
PCM was initially developed in the 1960s. Developed as solid-state alternative to dynamic random-access memory (DRAM), PCM uses an electrical charge to change a piece of chalcogenide glass — some forms of which exist in CD-RWs — from a crystalline state to a more amorphous one.
Does that seem like sci-fi? Sure, but the difference in the resistances of the two states gives you your ones and zeros. Although its performance was comparable to DRAM, it was considered too much of a power hog and its materials too expensive to be put in mainstream devices. So PCM was essentially the Betamax of computer memory.
Twice Betamax actually, because when flash, another, different type of solid-state memory, was developed, PCM was again overlooked. This was in spite of the vastly superior speed that PCM was capable of over flash. It was simply too expensive to put in a drive that people could wear on their key chains.
But now, as the need for further miniaturization of memory chips continues, we will shortly run up against a wall with both DRAM and flash memory. So researchers have been working for the past several years on making PCM more affordable and stable.
First, they found a way to put two logical bits in each physical cell, thus cutting material costs nearly in half. But this created some problems with the resistance levels shifting more, which would make read errors more likely. The breakthrough that the IBM scientists have announced involves some modulation coding techniques to help allay this problem. So now nothing stands in the way of developing this technology for consumer use.
It’s like some kind of Betamax Phoenix. Hmm, I like that. Hands off, it’s my copyright now!
So when can we expect computers or portable drives to come out with this new memory? Well, IBM expects it to be ready by 2016. So don’t throw your flash drives away yet.