Slashdot Mirror
Toshiba Developing High-Density 1TB SSD
MojoKid writes "A new partnership between Toshiba and Tokyo's Keio University has led to the creation of a new technology that could allow SSDs up to 1TB in size to be made 'with a footprint no larger than a postage stamp.' The report states that the two have been able to integrate 128GB NAND Flash chips and a single controller into a stamp-sized form factor. They've even made it operational with a transfer rates of 2Gbps (or about 250MB/sec) with data transfer that relies on radio communication."
Probably more than the naive observer would expect; but less than you would think.
My understanding, from TFA, is that the radio communication being used is very short range, a substitute for the usual maze of tiny and hard to fabricate gold wire interconnects that go between stacked dice. Die stacking itself isn't new; but the real-world manufacturability drops off unpleasantly as you stack higher, because of all the little wires. If you can use very short range RF instead, your life becomes rather less painful.
Assuming a suitable faraday cage layer isn't baked in, somebody with a nice antenna and some serious DSP could probably capture some of the traffic from the chip if they could get within a few cm of it. I'd hesitate to base the next generation of smart cards on such a thing; but it isn't as though it would necessarily be a radical advance over what you can do today with a few needles and a logic probe.
256 TB of flash... The storage device will be delivered in a standard fedex mailer, the payment will have to go by pallet....
Whereas mine ran for 3 years until I replaced the whole device.
Aren't anecdotes great!
Was your SSD from the cheap seats, or one of the decent ones? People were doing substantially better than that, in terms of lifespan, back when "SSD" meant "CF card in an IDE adapter"... With an N of 1, I suspect that you might have just gotten a dud. Mechanical drives that are dead when you open the box aren't exactly unknown in the field(on the other hand, though, intel has had a couple of really embarrasing firmware issues, and anything that JMicron has cursed with their misbegotten controllers is utter junk, so the field does have some maturing to do).
More broadly, though, size and reliability are actually closely linked with Flash SSDs. It is inherent in the nature of Flash that it will only survive a limited number of writes before a given block of cells becomes unwriteable at best and unreliable at worst. SSD controllers deal with this by trying to spread writes as evenly as possible over the available Flash space, and by having some amount of reserve space that can silently be substituted for failed blocks. The trouble, of course, is that since Flash is expensive, there is a strong commercial imperative to make as much as possible of the Flash you include visible storage space, so you can put a big shiny number on the box, and as little as possible reserve space, since that is hard to brag about. As a consequence, you'll note that cheap consumer SSDs ship with substantially less reserve flash than do the expensive; but reliability focused, enterprise ones(some of which will even let the customer adjust the allocation between storage and reserve).
If you can make Flash denser and cheaper, you'll make it more likely that, for all but the crappiest fly-by-night shops soldering together stuff stolen from nearby dumpsters, adding more reserve Flash is cheaper than processing RMAs and dealing with angry customers. Improvements in the intrinsic reliability of Flash cells would be nice as well, of course; but we are already using vaguely RAID-like techniques to turn quantity into reliability, so improvements in density and cost are almost as good.
Not really. I heard an anecdote once and it was really lousy.
Radio communication does not say it has to be over the air, it means that there is a carrier wave (in the wire) that has the signals put on top of just like radio.
7 years * 12 months/year * 10,000 GB/month = 840 TB of data written/deleted
10,000 Erases * 128 GB = 1280 TB of data written/deleted
It seems like any SSD of appropriate capacity will do that. 10,000 erases is actually extremely conservative, most SSDs advertise 2-3 orders of magnitude more than that. It'd take continuous writing at maximum speed for more than a decade* to kill most modern SSDs. Or at least that's the theory, I'm sure someone has gotten a defective one that died in a month or something.
* 5,000,000 Erases * (256 GB / 100 MB/sec) = 405 years