


NCSU's Fingernail-Size Chip Can Hold 1TB 227
CWmike writes "Engineers from North Carolina State University have created a new fingernail-size chip that can hold 1 trillion bytes (a terabyte) of data. They said their nanostructured Ni-MgO system can store up to 20 high-definition DVDs or 250 million pages of text, 'far exceeding the storage capacities of today's computer memory systems.' Using the process of selective doping, in which an impurity is added to a material whose properties consequently change, the engineers worked at nanoscale and added metal nickel to magnesium oxide, a ceramic. The resulting material contained clusters of nickel atoms no bigger than 10 square nanometers — a pinhead has a diameter of 1 million nanometers. The discovery represents a 90% size reduction compared with today's techniques, and an advancement that could boost computer storage capacity. 'Instead of making a chip that stores 20 gigabytes, you have one that can handle one terabyte, or 50 times more data,' said the team's leader, Jagdish 'Jay' Narayan, director of the National Science Foundation Center for Advanced Materials and Smart Structures at the university."
I wonder... (Score:1, Interesting)
nanotech on its way (Score:5, Interesting)
While they are light on details, the article implies this is a long term storage system (IE a flash chip replacement)
One would think creating RAM with a similar density would be possible as well.
I've used a super computer that had 74 TB of main memory, but clearly is something one can not afford nor fit in the home, to put it mildly. In a few years, will we have 1tb dimms at home? That would be sweet.
Even lacking that, a 1tb flash-like chip (not as in technology, but as in purpose/use) is still a huge improvement.
Let's just hope it doesn't go the way of the 100tb optical discs that are 'going into production within a year' for the last 10 years.
On a happier note, just imagine the reactions the RIAA/MPAA lawyers would have to such a thing existing!
"Now all of your 'IP' fits on a nine finger-nail-sized set!"
LoC? (Score:4, Interesting)
The Abstract. (Score:4, Interesting)
Here is the paper's abstract:
Now, my question is, how do you store information in that? If the material is paramagnetic [wikipedia.org], that implies it isn't stored like a disk (read/write using a magnetic field)? How are they planning on storing information in a clump of nickel atoms? (Note: I know absolutely nothing about this stuff)
Re:There is no chip. (Score:4, Interesting)
Years ago I worked a product that had an IC feature that could be manufactured reliably 99.99% of the time. For a real device with millions of such features that averages to almost zero yield, and this problem was not overcome. For some technologies the manufacturing yield hurdle can be overcome, for others it can't be. So although seeing a small number of memory cells work correctly is interesting and worthwhile, by itself that doesn't tell us whether we will ever see this technology in an actual product.
Re:There is no chip. (Score:5, Interesting)
These days the quality of memory is crap though. Just look at NAND flash: there are hundreds of failed blocks on most chips, and these days sectors with a bad bit or two are used and just error corrected. Same with hard disks. You work around this by shoving large amounts of error detection, correction, and relocation logic into the controller.
Let's say each individual bit (!) can be manufactured reliably 99.99% of the time. For a 2048-byte sector (typical for NAND flash), using sector-granularity remapping, there's a ~20% chance of a sector being good. That's not very good, but it still gives you 19% usable capacity. At a terabyte per chip, that's still 190GB of storage. If you add single-bit error correction, you'd get 500GB of storage. At 2-bit correction, 750GB. Current generation Flash memory already uses multiple-bit ECC for MLC level flash memory (where typically 2 bits will fail at once), and sectors with one bad bit(pair) are considered "good enough" and corrected away. If you can manufacture this 1TB storage chip at 99.99% per bit, and especially if most of the failures will happen at manufacture time and not develop later during use, I'll gladly take it given a reasonable amount of error correction wrapping it. It's not like we don't already rely on ECC for our day-to-day storage.
Yield issues affect mainly things like CPUs with no redundancy. With memory, you just lose the damaged parts. Even RAM these days is manufactured with spare blocks that can replace blocks that came out wrong, to increase yield (though it's usually only a few and the remapping is burned in at the factory).