Slashdot Mirror
"Self-Healing" NAND Flash Memory That Can Survive Over 100 Million Cycles
another random user writes with an interesting use of 800C heating elements to keep flash working longer. It's long been known that heating NAND to temperatures around 250C can restore life, but doing so was practically impossible. From the article: "Engineers at Macronix have a solution that moves flash memory over to a new life. ... They redesigned a flash memory chip to include onboard heaters to anneal small groups of memory cells. Applying a brief jolt of heat to a very restricted area within the chip (800 degrees C) returns the cell to a 'good' state. ... According to project member HangTing Lue, the annealing can be done infrequently and on one sector at a time while the device is inactive but still connected to the power source. It would not drain a cellphone battery, he added."
It's still a long way from commercialization, but if it works on a small scale...
What are the odds they'll let something that can heat up that much on an airplane, once they read this article? :\ More seriously, I assume this is over a very, very small area, and the chip dissipates that heat within a few minutes, and that it would only be warm to the touch for a few moments... but I still gotta ask: Is there the possibility of catastrophic failure? Like if the chip was maliciously reprogrammed to trigger all the heating elements simultaniously?
#fuckbeta #iamslashdot #dicemustdie
I cant wait for the first YouTube guide telling they should put flash memory in the oven to fix it. That will translate into baking cellphones ... and you know someone will try it.
Back around 1976 I was working in a group that used Intel 4004 processors and 1702A EPROM. They found they could get more program/erase cycles out of a 1702A if they periodically baked them in an oven.
To reiterate my comment posted on Ars two days ago when this popped:
So it's sort of a mix between traditional flash technology and the mechanism by which PCM works.
PCM does short pulses of between 400C and 700C to change the resistivity of the chalcogenide material, so generating these temperatures on microcircuitry like this isn't new.
*PCM = Phase Change Memory;
I suspect that 800C isn't out of reach, and the elements can be much coarser given you don't need them to alter a bit.
This story has popped up a few places already, and 90% of the comments are always "800C! But what if it catches fire?"
Yes, the floating gate is heated to 800C, but the volume of the heated area is on the order of a few hundred cubic nanometers. The energy involved in heating a volume that small is, well, incredibly small, and dissipates rapidly into rest of the chip. Your flash memory will not burst into flame. It will not require significantly more energy from your battery, and it will not require special clearance from the TSA to bring it on a plane.
The real challenge here is not coping with high temperatures, but rather balancing the increase in cell lifetime with the increase in die size. If the 100 million cycles number is completely accurate, then there's not much question that this technology will make its way into a lot of flash, but if that upside is only for a few (or even most) of the bits on a die, then things get more complicated
For more info run through the comments from the Ars Technica writeup of the same story: http://arstechnica.com/science/2012/11/nand-flash-gets-baked-lives-longer/
Before everyone gets worked into frothy concern about flash write cycles, keep in mind that the #1 cause of failure & data loss on Sandforce-based controllers is a toxic mess of piss-poor proprietary firmware that's brittle, bitchy, and will brick the drive in self-defense if it corrupts its internal database, then decides you're trying too hard to salvage your data by trying to use something like dd_rescue on it. Oh, and their decision to save a buck by omitting the supercapacitor that's supposed to guarantee that it always has enough power to finish its current write.
Read the Agility/Vertex 2 & 3 forums at ocz if you think I'm making this up. Basically, Sandforce drives have mandatory encryption that can't be disabled to maximize your odds of successful data recovery, but they also employ active countermeasures to detect "hacking attempts" that usually result in the drive ending up in "panic mode".
I wouldn't touch a Sandforce-tainted SSD with a dirty, tetanus-infected pole. They deserve to be sued into oblivion by class-action lawsuits. At the VERY least, they should give us the option of setting our own encryption key (to a value WE know), and a way to rip the bits from a borked drive for offline recovery. The most infuriating thing about data death by Sandforce is the knowledge that 99.99% of your data is *there*, but you aren't allowed to recover it due to their fucked up business policy.
I seriously doubt there will any detectable increase in the surface temp of the chip when this heating occurs, so that part of the story is a non-issue to me. What is much more important to me is extending the life span of the NAND flash, meaning SSD's and other devices using it will no longer have such a short useful life. I have not even considered buying an SSD so far precisely because of that. This technology, if it works as promised, will make me take the plunge! .
The older programmable logic chips of the seventies and eighties, such as bipolar PROMs and PALs, had a metal fuse for every programmed bit. Those fuses would be melted to program the chip. However, those days featured external programmers to do the melting.
The determined Real Programmer can write Fortran programs in any language.
It would seem to me you'd be much better off with a soldering station (digital if you prefer) where you can aim all (carefully controlled) heat directly at the chip in question. It's not like they're stupidly expensive either, and they're enormously useful for all your standard soldering needs.
How is the Riemann zeta function like Trump rallies? Both have an endless number of trivial zeros.
... is a really, really bad idea.
The plastics used in IC packages tend to absorb moisture from the air. Over time this moisture gets trapped in the package. If you heat a plastic IC package that has been in service to 250C there is a pretty good chance it will 'popcorn' thus destroying the chip, and often visibly rupturing the package. It can be prevented by baking the board at 50 - 80C for 12 - 20 hours, but that is going to cause other problems. No one is going to be using reflow ovens to "reset" FLASH cells.
As others have pointed out spot-heating a FLASH block to 800C would take an almost trivial amount of energy, and would only need to be done when a block fails an erase-write cycle, or as an idle task performed on blocks that have been reassigned from the 'in-use' pool to the 'dirty' pool, but have not yet been erased and assigned to the 'free' pool.
The hardware required to implement this on an existing flash design would be almost trivial.