Skip to main content

Samsung pushes for spin-transfer RAM with Grandis

Samsung has announced that it has acquired Grandis for an undisclosed sum, telegraphing the its intentions to commercialise the spin-transfer torque technology in which the company specialised.

Grandis first hit the headlines in April 2008 when it entered into a partnership with memory maker Hynix Semiconductor to investigate the commercial exploitation of a relatively new form of memory known as spin-transfer torque RAM, or STT-RAM.

Unlike traditional DRAM, which uses tiny capacitors to store a charge that leaks over time, STT-RAM uses a spin-polarised current to modify the orientation of a small magnetic element. As a result, it's non-volatile - meaning that the data remains even when no power is applied - and extremely efficient.

While research into STT-RAM - which is a variant of magnetoresistive RAM, or M-RAM, offering a simpler design and improved scalability - has been ongoing since the 90s, a commercial implementation has yet to appear. A working 32Mb STT-RAM module was created and demonstrated by Hitachi in partnership with Tohoku University in 2009, but never made it to market.

As companies invest in improving the densitites and performance of conventional volatile DRAM and NAND flash, interest in STT-RAM was thought to be waning. The promise of 'universal memory' - a single storage medium that can act as high-speed RAM or non-volatile permanent storage depending on need - appears to be causing a renewed interest in the technology, however.

The deal will see Samsung fold Grandis into its research and development division, which makes it unlikely that we'll be seeing a commercial product any time soon. With Samsung being a major player in both the DRAM and NAND flash markets, it's a clear indication that the company believes there's a future in STT-RAM technologies.

Sadly, there are some hurdles to overcome first: at present, the current required to flip the orientation of a bit in an STT-RAM module is too high to make the technology worth implementing. Samsung's first task, therefore, is to find a way to drop that current down to a more manageable level.

The company will have to be quick, though: rivals including IBM and HP are working on a competing 'universal memory' technology which uses a passive two-terminal circuit element known as a 'memristor.' While that, too, is some years away from a commercial implementation, whichever makes it to market first is likely to become the de facto standard.