The search for an alternative non-volatile memory technology to traditional floating gate flash has been boosted by an attempt to raise MRAM to the Gbit level. But the timescales for achieving it are tight.

The urgency for finding a replacement for conventional floating gate flash based on Fowler-Nordheim tunnelling and hot-electron injection is because these techniques are thought to be unusable at smaller geometry processes due to the thinness of the oxides.

One way round this barrier is the NROM, MNOS-based technology, pioneered by Saifun of Israel, and licensed to Spansion and Macronix. NROM has not made it to the leading density level of traditional floating gate flash memory, although its new Quad-bit technology, offering four bits to the gate, may do so.

Saifun said it has made parts with an oxide thickness of 3nm, about a third the oxide thickness of leading edge conventional floating gate flash processes.

The IBM-TDK joint R&D initiative to develop multi-Gbit density MRAM could provide an alternative to traditional floating gate flash but, with flash densities already at 16Gbit, and IBM suggesting the programme may take four years, it is doubtful whether it can catch up with flash, which is doubling in density every couple of years.

The highest density commercial MRAM is the 4Mbit MRAM being sold by Freescale for about $25 which is hardly a mass market proposition when 8Gbit NAND flash chips go for around $10.

Another technology being developed to replace traditional floating gate flash, is the chalcogenide-based, phase-change approach called Ovonic Unified Memory (OUM) pursued by Numonyx, the Intel-STMicroelectronics joint venture, among others.

Samsung is said to be putting a 512Mbit OUM memory on the market next year.

Meanwhile, Samsung has said that traditional floating processes will be possible at 30nm, making the search for a replacement technology less urgent.