Slashdot Mirror
Scientists Unveil Most Dense Memory Circuit Ever Made
adamlazz writes "The most dense computer memory circuit ever fabricated, capable of storing around 2,000 words in a unit the size of a white blood cell, was unveiled by scientists in California. The team of experts at the California Institute of Technology (Caltech) and the University of California, Los Angeles (UCLA) who developed the 160-kilobit memory cell say it has a bit density of 100 gigabits per square centimeter, a new record. The cell is capable of storing a file the size of the United States' Declaration of Independence with room left over."
Rough comparison here. Short answer: DNA is far more dense information storage than this technology. Never mind that human white blood cells also contain the machinery to both compute and replicate data stored within DNA (as well as replicating the computation machinery).
Biology still wins. But nanotechnology creeps ever closer year by year...
DNA has built-in fault tolerance. By contrast, this new memory circuit by Caltech would vaporize once an alpha particle hits it.
A word is 16 bits or so I think.
I blame geof's speakers.
The Yahoo! News article got the figures wrong. To get only 2,000 words (a computer term, not a linguistic one) out of 160-kbits they'd have to be 80-bit words. The article at Technology Review has better maths and more information to boot.
In the x86 world it is.. A word can be the width of the bus on other architectures.
The piece on Yahoo! News was pretty low on details, so here's the abstract from the Nature paper:
A 160-kilobit molecular electronic memory patterned at 1011 bits per square centimetre
Jonathan E. Green1,4, Jang Wook Choi1,4, Akram Boukai1, Yuri Bunimovich1, Ezekiel Johnston-Halperin1,3, Erica DeIonno1, Yi Luo1,3, Bonnie A. Sheriff1, Ke Xu1, Young Shik Shin1, Hsian-Rong Tseng2,3, J. Fraser Stoddart2 and James R. Heath1
The primary metric for gauging progress in the various semiconductor integrated circuit technologies is the spacing, or pitch, between the most closely spaced wires within a dynamic random access memory (DRAM) circuit1. Modern DRAM circuits have 140 nm pitch wires and a memory cell size of 0.0408 mum2. Improving integrated circuit technology will require that these dimensions decrease over time. However, at present a large fraction of the patterning and materials requirements that we expect to need for the construction of new integrated circuit technologies in 2013 have 'no known solution'1. Promising ingredients for advances in integrated circuit technology are nanowires2, molecular electronics3 and defect-tolerant architectures4, as demonstrated by reports of single devices5, 6, 7 and small circuits8, 9. Methods of extending these approaches to large-scale, high-density circuitry are largely undeveloped. Here we describe a 160,000-bit molecular electronic memory circuit, fabricated at a density of 1011 bits cm-2 (pitch 33 nm; memory cell size 0.0011 mum2), that is, roughly analogous to the dimensions of a DRAM circuit1 projected to be available by 2020. A monolayer of bistable, [2]rotaxane molecules10 served as the data storage elements. Although the circuit has large numbers of defects, those defects could be readily identified through electronic testing and isolated using software coding. The working bits were then configured to form a fully functional random access memory circuit for storing and retrieving information.
Also, an interesting bit from the very end of the paper:
Many scientific and engineering challenges, such as device robustness, improved etching tools and improved switching speed, remain to be addressed before the type of crossbar memory described here can be practical. Nevertheless, this 160,000-bit molecular memory does indicate that at least some of the most challenging scientific issues associated with integrating nanowires, molecular materials, and defect-tolerant circuit architectures at extreme dimensions are solvable. Although it is unlikely that these digital circuits will scale to a density that is only limited by the size of the molecular switches, it should be possible to increase the bit density considerably over what is described here. Recent nano-imprinting results suggest that high-throughput manufacturing of these types of circuits may be possible29. Finally, these results provide a compelling demonstration of many of the nanotechnology concepts that were introduced by the Teramac supercomputer several years ago, albeit using a circuit that contained a significantly higher fraction of defective components than did the Teramac machine4.
This one is a bit better, but apparently the Nature article will be released tomorrow, which I assume would have the sort of detail you're asking for.
https://www.eff.org/https-everywhere
Original text from NARA
Wikipedia
Microsoft Word say:
3 pages
8 paragraphs
111 lines
1338 words
6782 characters
8114 characters (with spaces)
HBI's Law: Frequency of calling others Nazis is directly correlated with the likelihood of the accuser being Communist.
like "size of a red blood cell"
The "size" of a red blood cell is around 7 micrometers thick, and around 30 micrometers in diameter IIRC... can't remember white blood cells but they're quite a bit bigger.
Seven puppies were harmed during the making of this post.
Although the news writeup doesn't make this clear, the original scientific paper is comparing the data density achieved with RAM, not with hard disks. What they are doing is minituarizing high-speed volatile random-access memory. Although a 3 terabyte hard disk may not excite you, would you be impressed with 3 terabytes of RAM?
If we could increase the data-density of RAM by a few orders of magnitude (without sacrificing access times, of course), we could avoid one of the main bottlenecks in modern computers.
Stop opressing me, I can post where ever I wanna!
But seriously, using the estimate from wikipedia: "It is estimated that the print holdings of the Library of Congress would, if digitized and stored as plain text, constitute 17 to 20 terabytes of information", we can use google to calculate how many such chips would be required to store the US Library of Congress:
Enter into google: (20 terabytes) / (160 kilobytes) = 134 217 728
Now, with some reasearch into White Blood Cells, we learn that a normal human has between 7000 and 25,000 white blood cells in a drop of blood. So going with a conservative estimate of 10,000 white blood cells per a drop of blood, we could store the Library of Congress in
134 217 728 / 10 000 = 13 421.7728 drops of blood.
That's not very accurate, let's try to get a better estimate. Wikipedia to the resque:
Again, with a conservative estimate of 7 x 10^9 white blood cells per liter, we get
134 217 728 / (7 * (10^9)) = 0.0191739611
Entering into google 0.0191739611 liter to centiliter, we get
0.0191739611 liter = 1.91739611 centiliter
In other words, storing the whole Library of Congress using these chips would take about half a shotglass of blood.
- The memory is nonvolatile (technically speaking), but bits decay in about an hour.
- The technology is not at the point were word length is a concern. The researchers were reading and writing individual bits. About half of all bits were deemed defective (having an on/off ratio of less that 1.5)
- As said above, bits last for about an hour. I'm sure they intend to improve this.
- Bits were written using
.2 second pulses of 1.5 volts and read using .2 volts. I doubt any attempt was made to optimize these quantities. Remember, voltages are not being applied directly to the molecules storing information, they are applied to sets of perpendicular nanowires. The electrodes controlling these nanowires might be the cause of the slow switching time.
- There is no mention of power requirements in the article
As you can see, this technology is not coming to market any time soon. The research demonstrates that a grid of nanowires can control a layer of programmable molecules, yielding a very high storage densities. High defect rates, however, may require a large overhead (think of the size of the defect map required). More importantly, the nanowires that form the grid cannot yet be individually controlled. In this experiment, electrodes were used to turn on 2 to 4 nanowires at a time. To date, no reliable way of controlling many individual nanowires has been experimentally demonstrated.