New research: Flash is DEAD. Yet resistance isn't futile - it's key
Electro-boffin may have SAVED the storage WORLD
Flash is at a crossroads. It cannot keep shrinking dies because, beyond a certain point, NAND cells produce too many errors, are slower to respond and have a shortened working life. Yet a University of Michigan professor's research may hold the key to the future of flash storage technology.
Flash foundry suppliers are working on vertical layer stacking to cram more cells at current geometries onto the same die footprint. This is exemplified by Samsung's V-NAND.
But the suppliers know that vertical layer stacking will reach its limit in turn. Once they reach that point where will they go to keep increasing density and so growing sales?
Several post-NAND technologies are in development. One uses Phase Change Memory (PCM), which features an electrical current changing the nature of a Chalcogenide substance from an amorphous state to a crystalline one and back again. Each state has a different resistance level, meaning these levels can be detected and signal binary one and zero.
Another technology is called Resistive RAM (ReRAM or RRAM) and uses the same ultimate characteristic, different resistance levels, to indicate binary values.
Such technologies promise data access speeds nearer to that of DRAM than NAND, byte instead of block addressability, and a longer working life: three much desired benefits.
A startup called Crossbar has demonstrated a 1MB die – although it hinted at 1 terabyte one to come - using ReRAM technology. Current NAND dies are at around the 128Gbit level so this is a huge advance. Samsung foresees a 1Tb - terabit - V-NAND die in 2017 and that's with a 96-layer die.
The die, a test die, embodies a 1TnR:1 design in which 1 Transistor driving n Resistive memory cells, more than 2,000 in the current design which, ironically, uses 3D stacking of cell layers, similar to, we understand, V-NAND's layering technology. Crossbar says this opens the road towards "commercialising terabyte scale memory arrays on a postage stamp size chip."
It claims that PCM and other RRAM suppliers' technologies suffer from a "sneak path current" phenomenon when operating in high density configurations. This hinders data access and entails using more power, Crossbar says excessively more power, than Crossbar's tech. This is based on the formation of a filament in a silicon-based switching material when a voltage is applied between top and bottom electrodes either side of the switching material.
How electro-boffins SAVED STORAGE
All of this is connected with a research paper published by Crossbar co-founder and boffin, Wei Liu, and colleagues, entitled "Electrochemical dynamics of nanoscale metallic inclusions in dielectrics" and published in Nature Communications.
An IMFT 25nm Flash die. Will it be obsolete soon?
Wei is a proper boffin, being associate professor of electrical and computer engineering at the University of Michigan.
His research paper shows that, when a voltage is applied, metal atoms in the switching medium become charged ions and cluster together, forming metal nano-particles. These move to form a bridge, or a conducting filament, between the two electrodes. In a demonstration using silver and platinum, the bridge stayed in place when the power was switched off.
The University of Michigan announcement says: "The electric field can be used to change the shape and size of the filament, or break the filament altogether, which in turn regulates the resistance of the device, or how easy current can flow through it."
The research work "unmasks some of the magic behind memristors and "resistive random access memory," indicating that there is a logical connection to HP's memristor efforts.
Crossbar CEO George Minassian has a canned quote enthusing about the test dies.
"With 1TnR, companies will realise the dream of extremely dense, highly reliable, and high performance solid state storage. It’s truly ground breaking and has the potential to redefine what’s possible in enterprise storage and high-capacity non-volatile SoC memories."
The company identifies two attributes of its technology:
- Cell physics –- RRAM chips do not suffer from NAND wear-out issues due to the technology’s fundamental memory cell structure, based on metallic nano-filament in a non-conductive layer.
- Economics – semiconductor manufacturing facilities for advanced NAND Flash require multi-billion dollar investments. Crossbar’s RRAM can be stacked in 3D directly on top of standard CMOS wafers, making it cost effective to manufacture.
It says super-dense memory array architectures will be able to be constructed when its technology is commercialised. In the company's words: "CMOS compatibility enable logic and memory to be easily and cost-effectively integrated onto a single chip, at the latest technology node, on standard manufacturing processes."
There is also a 1T1R technology for embedded code applications.
Get a useful Crossbar white paper here (pdf).
Crossbar says it's currently finalising agreements with several leading global semiconductor companies and plans to announce its first licensing agreements shortly. ®