Phase change memory will probably just miss the Windows 8 schedule but it's been looking the most promising advance on flash storage - which is going to hit a problem with density and storage life in the not too distant future. The larger your flash memory, the shorter the storage life and some of the really high density flash memory that's been talked about could store information for as little as an hour.
The memristor - the long-predicted, once mythical fourth electronic component that combines memory and a resistor - made it from vision to lab technology a couple of years ago but originally HP said it didn't want to manufacture them itself. Memristors are simpler than transistors, can be stacked three dimensionally and - like flash but unlike RAM - store data without power (ideal for mobile storage). We heard reports that HP had been shopping the technology around possible partners but Intel is already working on phase change memory (storing information on a material that changes to and from a crystalline state) and no-one had been picking it up.
Meanwhile, the HP Labs team has carried on working on the fundamentals. HP researcher Stan Williams recently announced in Nature that memristors can act as logic gates as well as memory, using resistance instead of charge. What could you do with that, I wondered? As Simon is the electrical engineer, I turned to Our Esteemed Editor for his view. Just about anything you can do now, but much more cheaply, for a start, he replied. "That's the fundamental building block of logic - an inverter and an or gate - so you can build any logic from that. Now we know we can make memristor memory circuits with integrated logic in the same technology, it should be much cheaper to do that than to do it with one memristor component and one CMOS component (where you have double the production costs)".
Williams told the NYT that the team has the switching speed up to the same as silicon transistors and the reliability up to hundreds of thousands of reads and writes. And they're a lot smaller; instead of state-of-the-art 32 nanometre silicon, HP has a 3 nanometre memristor with a nanosecond switching time. There isn't the same problem of trying to force electrons into smaller and smaller gates that's facing silicon manufacturers, because memristors store information by pushing atoms around in a film of titanium dioxide and then reading back the position of the atom. HP has already turned its printing expertise to printing photovoltaic cells and displays on plastic sheeting (and it's developed and licensed out a way for ink-jet printing drugs onto patches you can wear on the skin). Memristors come from the same ability to manufacture at the nanometre scale that HP uses to make all those inkjet cartridges and printheads that deliver 18,000 4-picolitre drops of ink a second (at a temperature five times as hot as the sun) manufactured with subs-nanometre micromachine processes; it isn't the technical issues that would stop HP manufacturing them (it might be the margins on components).
It's still a question of getting memristors out of the lab and into production (and then into products), but while I haven’t seen a new battery technology make it to the market in years, I think we might have two new memory technologies on the way within the next three years - and at least one of them might use less power.