Flash has been successful because it is non volatile, but it consumes a lot of power, has limited life and can't be integrated easily alongside other circuitry. Perhaps the biggest downside for NAND is that you can't erase single bits; it's blocks or nothing. But flash can't replace DRAM; only complement it.
So researchers have been looking to develop a technology would not only replace other memories, but also not use electrons. The reason? As process technologies shrink, there are fewer electrons available to store the charge that indicates a 1 or a 0.
One approach is phase change memory (PCM), where the amorphous and crystalline states of a material represent those 1s and 0s. Other approaches include spin torque transfer and magnetic RAM.
Now we have 3D XPoint, said to combine the performance, density, power, non volatility and cost advantages of all available memory technologies. On the face of it, a lot of boxes are ticked.
But while they tell us what 3D XPoint can do, the partners have yet to tell us how it does it. So speculation, as they say, is rife. Some have concluded that 3D XPoint is PCM. However, if 3D XPoint is based on PCM, that it isn't a breakthrough; the Intel/STMicroelectronics joint venture Numonyx unveiled PCM devices in 2009. And Numonyx claimed its technology could stack. Perhaps what we're seeing with 3D XPoint is the latest version of what Numonyx showed six years ago.
But while 3D XPoint can store 128Gbit, it does so only on two layers. Flash manufacturers – including Micron – are now looking at many more layers than that. Toshiba, for example, is talking about a 256Gbit part with 48 layers. By stacking that number of layers, designers don't have to worry so much about the number of electrons in a cell, but can still take advantage of process technology and, hence, keep the cost down.
Will 3D XPoint succeed? Who knows? But having production devices available will help users to decide.
