Japanese Scientists Develop Long-Life Flash Memory

schliz writes "Flash memory chips with a potential lifetime of hundreds of years have been developed by Japanese scientists. The new chips also work at lower voltages than conventional chips, according to the scientists from the University of Tokyo. They are said to be scaleable down to at least 10 nm; current Flash chips wouldn't be usable below 20 nm."

Comments

Awesome

[NoobixCube]

This will be a huge boon to the UMPC form factor. SSDs are still far too expensive, and regular laptop hard drives eat through batteries in a single-digit matter of hours.

Re:Awesome

[utopianfiat]

10nm get you anyting you want baby, me so info-dense, baby, me so info-dense. Me store you long time.

What is the point?

[damburger]

Given that we tend to dump flash memory whenever a larger and more compact one comes along, and transfer our data, what use is there for a flash chip that will keep data for 100 years but be obsolete in 2?

Re:What is the point?

[dintech]

Archival. Once it's archived you can forget about it. For example, your local library doesn't convert all that old microfilm just because it can. It would only do it to put it onto a more stable storage medium.

Re:What is the point?

[jacquesm]

it's to facilitate the new profession of 'data archaeologist'. People that will be sifting through the digital detritus of the pre-AI era two hundred years from now.

Looking for the rosetta's stone that will enable them to translate 'flash' into 'realmedia' ;)

Re:What is the point?

[$RANDOMLUSER]

The calendar time isn't important, it's just a headline. The real news is the number of write cycles going from ~10,100 to ~100,000,000 cycles, thereby making it usable in things like swap memory. By marking bad cells, much like bad sectors on hard disks, you also don't have to discard the whole chip if a single cell fails - like you do if a single cell fails in a RAM chip.

Re:What is the point?

[Firefalcon]

It's still better than the lifetime of most other electronic storage media. Obviously conservation efforts (i.e. duplication) would have to be made (at it's half life of 50 years I'd guess), but the same applies to film, paper, etc.

The advantage of digital media though is that multiple identical copies can be made, without any loss that can occur when duplicating analogue materials, and the cost of multiple digital copies over an extended period is almost certainly going to be considerably less than the cost of performing restoration and preservation on, for instance, a several hundred year old manuscript.

Re:What is the point?

[MyLongNickName]

I would hardly call 100 years archival. In some exceptional cases its within the memory span of a single human individual.

Ummmm. Yea. I am going to have granny memorize my last ten years of photos, movies and financial records.

Fact is, I have struggled with a good method for backing up all of this. I've basically settled on mirroring with a remote FTP site. It works, but with my horrible upload speed, initial synchronization took 48 hours plus. Quarterly updates take a couple hours. And the other pain in the butt is I have to encrypt my financial info as I don't trust it being in the hands of a third party.

Now if I had a medium that were 99% successful at retaining info for 20 years, I would backup to two manufacturer's media, and stick it in my safety deposit box.

I don't have that degree of confidence in any low cost storage media yet.

So for archival, yes, this is a wonderful advance.

Re:What is the point?

[Swizec]

Oh gawd, you're gonna torture your grandchildren through ALL that archived stuff? Crazy, I can hardly deal with the tens of old-school analogue photos that have survived through the ages, let alone having a grand parent silly enough to think archiving all of their photos was a good idea.

Re:What is the point?

[jimicus]

Archival. Once it's archived you can forget about it. For example, your local library doesn't convert all that old microfilm just because it can. It would only do it to put it onto a more stable storage medium.

At least until the technology changes so much that you can no longer buy anything that will read it, cf. the BBC's Doomsday project:

http://www.iconbar.com/forums/viewthread.php?newsid=937

Re:What is the point?

[MyLongNickName]

I'm assuming you are under 30 and haven't lost a grandparent. Now that I have lost a few family members, I wish I had more photos, more memories to look through. Perhaps it is a case of you don't miss something till it is gone.

I will pull up the digital photo album of old vacations, and my kids love to remember what we did. At some point, my kids will become uninterested as I did when I was younger. But as some point, I know they will enjoy revisiting them.

I sure as hell don't want that to not be possible because my hard drive crashed.

Re:What is the point?

[jim.hansson]

i have heard stories from universities that they sometimes get help requests from people that have data stored on mediastypes that nobody has readers for anymore, and after a little hunt in basements and other places they find the hardware, then there is the problem of software. There is already companies specializing in this sort of things

Re:What is the point?

[jacquesm]

The move where storage is going 'online' will mitigate this to some extent, at the same time it will create a larger problem is something goes wrong with all that online storage.

Storage reminds me of the situation around energy generation. If you all generate your own energy and consume it on the spot then there will be lots of outages, but small ones. If you do it centralized then you get less outages, but *MUCH* larger ones.

I fully expect something similar to happen to online storage, it will seem to be more reliable because on average it will be better than storing your data locally, but when it goes it will go bigtime.

That's when the data recovery guys will have a field day.

Re:What is the point?

[nmg196]

Because currently an SSD will not last the life of your computer. At some stage prior to your processor wearing out, the disk will fail and you will lose data. At the moment, a mechanical HDD is still less likely to fail than an SSD.

We use them here at work in firewall applicances and I've so far yet to see an SSD last for longer than one year when the disk is used heavily to log network traffic. SSDs are absolutely rubbish for high usage (high read/write cycles). If you made one into a Usenet server for example, I doubt it would last a month.

Long life SSDs would also make the ultimate backup medium.

Re:What is the point?

[nmg196]

> I would hardly call 100 years archival

You might not, but everyone else certainly does.

What do you think "archive" folders in Outlook are for? Emails older than 100 years?

Many companies archive financial records, which are then permanently destroyed after 5 or 10 years. There is very little you'd want to archive for much longer than this in the business world. Archived data is simply anything you don't foresee needing to use again. Even if you last used it last week - you might as well archive it if you know you're probably not going to need it again.

I've got no idea why you're comparing archiving data to human memory. No human can remember every single byte of every backed up file on their PC for even one minute - let alone 100 years. I assume that bit is simply flamebait as it's a ridiculously stupid comparison.

Re:What is the point?

[thrillseeker]

What do you think "archive" folders in Outlook are for? Emails older than 100 years?

You never worked for my old boss.

Re:What is the point?

[Hal_Porter]

Regular flash works just fine for swap. If you write nonstop at top speed to a standard chip, you'll wear I'd out in about fifty years. Thus I don't understand why we should care about an even longer lifetime.

That used to be true with SLC chips. It's not true with MLC.

http://www.storagesearch.com/ssd-slc-mlc-notes.html

It's a simple matter to plug new data for MLCs into the calculation I did for the worst case wear-out process for flash SSDs - which I called the Rogue Data Recorder.

Instead of the 64GB example I used then, I'll assume the MLC SSD has 128GB capacity. MLC SSDs have more capacity than SLC. And more capacity means longer operating life - before cells wear out.

I'll still use the 80M bytes / sec sustained write speed - because the fastest MLC products (in Feb 2008) can already do that. (Meanwhile the fastest SLC products have moved up in the world and are about 50% faster.)

The next factor is where we hit the big problem... Instead of a write endurance rating of 2 million cycles (for the best SLC) - I can only use a figure of 10,000 for MLC. MLC has a much lower rating due to the complex interaction of discriminating multiple logic levels reliably coupled with the intrinsic failure mechanism of wear-out.

Plugging these numbers in the same calculation gives an estimated MLC flash SSD operating life (at max write throughput) which is 6 months! (instead of 51 years for a 64GB SLC SSD).

All the affordable SSDs I've seen from Intel and Samsung are based on MLC flash because it costs much less per bit. Down to $2 per GB in fact. SLC currently costs 2-4x as much. E.g.

Here are the average prices for flash

32Gb 4Gx8 MLC 9.27
16Gb 2Gx8 SLC 15.61
16Gb 2Gx8 MLC 3.97
8Gb 1Gx8 SLC 6.31
8Gb 1Gx8 MLC 2.34

SLC is 2.7x more expensive for 1Gx8 and 3.9x more expensive for 2Gx8. So it's not surprising that most SSDs are MLC based. But if you write at full speed to them they will die very quickly.

Incidentally look at the price of 4Gx8 MLC. $2.31 per gigabyte. Pretty damn cheap.

Re:What is the point?

[Free the Cowards]

It doesn't matter what you write, because the logical sectors are not linked to the physical sectors on any reasonable flash drive. The controller circuitry holds a mapping which it adjusts as time goes by to evenly use the entire device no matter what your write patterns are.

As for "not much cheaper", this must be a new meaning of "not much" that I was previously unaware of. Taking a quick sample on newegg.com, I find an 8GB flash drive for $32, and 8GB of RAM going for around five to six times that. The flash drive, in addition to being vastly cheaper, is also much smaller and consumes much less power.

Re:What is the point?

[Anonymous Coward]

I guarantee you that the $32 8GB flash drive is not what you call 'reasonable'. Flash drives which actually do distribute wear across the whole drive no matter what your write pattern is are considerably more expensive. As a general rule, if it's a cheap USB flash key, it's built using MLC flash chips (only about 10K erase/write cycles per erase block), is slow (MLC is slow), and implements a fancy form of defect remapping rather than real wear leveling.

The problem a lot of people have is that they hear about the concept of wear leveling and then assume it's cheap and easy to implement properly, since it sounds so simple. It's not. Try thinking about how you actually have to do it sometime. For example, how do you store the table or list which maps logical sectors to physical sectors? That alone is a very thorny problem, because if you store it in some of the physical sectors, you have to make the table itself subject to the wear-leveling algorithm, as the table is going to be the most frequently-written data of all. So you end up needing to store the table (and probably user data too) using some kind of log-structured file system. Now you have to chase pointers just to figure out where user data is located, and as high-density flash is not quite as random-access as people sometimes think it is (it's very oriented towards reading/writing large blocks rather than random access to individual bytes), you'd better have a controller with a lot of RAM so you can cache large chunks of the table; otherwise performance goes in the toilet.

TLDR version: true wear-leveling flash disk controllers tend to need lots of RAM and/or a second nonvolatile solid state memory with better lifespan than the main flash array. These things are expensive and therefore not found in most flash drives other than the SSDs built to the same form factor as hard disks.

if you write real small

[jacquesm]

Stone tablets will last even longer!

Re:if you write real small

[somersault]

Unfortunately, even God can only fit 5 commands on a single stone tablet.

Re:if you write real small

[jacquesm]

the fact that any three year old can do better is probably one of the stronger proofs that god, indeed, does not exist.

Re:if you write real small

[Chrisq]

Unfortunately, even God can only fit 5 commands on a single stone tablet.

Its just as well. Imagine what Sunday School would have been like if Moses hat taken a PDA up mount Sinai and come back with the 65,536 commandments.

Re:if you write real small

[Thanshin]

It's not just like 'thou shalt not kill', each command is quite verbose.

Thou, hereby referred to as THE SINNER, shall not, under any circumstance, unless with the express permission of thy god, hereby referred as GOD, attempt to willfully, negligently or otherwise end the life of another...

Re:if you write real small

[Redrover5545]

come back with the 65,536 commandments.

Somebody hasn't read Leviticus or Deuteronomy.

Re:I thought that all flash was already long lifed

[ZombieWomble]

This is one of those wonderful headlines where they convert the big scary numbers into a nice friendly unit and completely miss the point. What's interesting about this memory is not that it could be locked away and would be stable, but that it's much more stable under repeated use (100 million writes as opposed to tens of thousands). So they've presumably taken some arbitrary number of "writes per year" and divided to get their 100 year figure.

(Bonus exercise for the reader: Calculate the lifetime of these chips in libraries of congress written!)

Or...

[DrYak]

god, indeed, does not exist.

...Or is even less skilled than a toddler.

Re:Or...

[Wite_Noiz]

Blimey, you guys are cynical.

The guy (or gal) was etching those stones using a friggin' lightning bolt from his cloud in the sky... that's pretty damned impressive.

Read / write cycles

[Dan East]

The summary does not specify exactly what is meant by "long-life". That refers to the current limitation of flash, where individual bits have a physical limitation to the number of times they can be modified. This "new" flash uses some sort of integrated "wear-leveling", so that all bits are utilized equally. Also, when individual bits (or more likely, groups of bits) are worn out they are retired. So instead of a failure, the capacity of the flash would decrease as write cycles exceed the physical limitations. Of course, if wear leveling was performed perfectly, then pretty much the entire array would fail at once, right?

The article doesn't address other important aspects, like read / write speed.

It does say that current flash memory is limited to 10k writes, which is low by at least a factor of 10. Modern flash should withstand at least 100k writes, and I've seen claims of over a million here and there.

Re:Read / write cycles

[courteaudotbiz]

Of course, if wear leveling was performed perfectly, then pretty much the entire array would fail at once, right?

I have 4 wheels bearings on my Chrysler, and even though all of my wheels rotate with a perfect synchronism, only one of my bearings fails at a time, and the other ones don't follow ther brother in the next few miles...

Remember that at this scale, only an atom of difference could make some of those individual bits fail a year before the other...

Re:Read / write cycles

[xtracto]

A bit of research from the original AIST site bring quite a lot of info.

The from the original tech report:

Shigeki Sakai (Leader) et al. of the Novel Electron Devices Group, the Nanoelectronics Research Institute (Director: Seigo Kanemaru) of the National Institute of Advanced Industrial Science and Technology (AIST) (President: Hiroyuki Yoshikawa) in collaboration with Ken Takeuchi, Associate Professor of the Graduate School of Engineering, the University of Tokyo (Univ. Tokyo) have demonstrated that the use of ferroelectric gate field-effect transistors (FeFETs) as memory cells dramatically improves the performance of NAND flash memory. The FeFET, the newly developed memory cell, can be programmed and erased as many times as 100 million or more and with programming voltage of less than 6 V, whereas the conventional NAND flash memory cells have ten thousand program/erase endurance cycles with approximately 20 V programming voltage. It has been assumed that conventional NAND flash memory can be downsized to 30 nm at the minimum, whereas this novel memory cell will meet the needs of the next 20-nm and 10-nm technology generations. And thus, this memory cell is expected to be used in a next-generation, high-density, high-capacity nonvolatile memory.

Results of the research was reported at the 23rd Nonvolatile Semiconductor Memory Workshop (the 23rd IEEE NVSMW / the 3rd ICMTD 2008) held in France, May 18â"22, 2008.

The key might last 100 years...

[Viol8]

... but will there be anything still able to read it in 2108? Even today finding something to read a laserdisk or some old style floppy disks is an issue and thats only 30 year old tech!

Re:The key might last 100 years...

[Viol8]

"I expect in another 100 years we'll have direct-from-brain transfer of information to and from our implanted minicomputers"

Meh , that crops up all the time in sci-fi and futuroligist stuff. I'm not convinced. The technology may become available but I doubt many people apart from a few techno fetischists and body piercing types would really want a machine plugged into their body full time, much less their brains.

Re:The key might last 100 years...

[Amorymeltzer]

You clearly haven't heard of the iPod, it seems.

Re:The key might last 100 years...

[Alioth]

Using electronic memory is far easier than say, a disc. With a disc, you need a lot of precision mechanical stuff in addition to electronics. With a semiconductor, you don't need all the mechanical stuff. It would take me about 45 minutes to make a circuit on breadboard to read a ROM that was made in 1975 onto a modern MacBook Pro. To homebrew a laser disc player would probably be two years work.

That's what flash looks like under a microscope

[Joce640k]

And that's why it wears out. Apparently.

Umm .. MRAM anybody?

[djtachyon]

I just received some samples of military grade MRAM recently. 4MB, "infinite" writes, "infinite" lifetime, -55C - 125C operating range, lower power than DRAM, and 35ns cycle times.

Fairchild has been making MRAM for awhile now.

http://en.wikipedia.org/wiki/MRAM

Re:Umm .. MRAM anybody?

[dave420]

Because that 4MB chip is the only MRAM chip currently produced that's commercially available. 4MB. This isn't 1992 any more. That's not impressive :)

Domesday, not Doomsday

[Mathinker]

It's Domesday, not Doomsday. Judging from Wikipedia, the Domesday Book was, well, kind of like the first British census?

Thanks for the interesting link. One of the things which stuck out was:

Sadly, it is unlikely that Domesday will become available for the general public to use. The contents of the discs are heavily tied up in copyright - parts are owned by the BBC, the Ordinance Survey, and possibly the Local Education Authorities and schools.

Another example of how the inflexibility of copyright strangles reuse and archival of information.

This is not flash ram

[infolib]

First, the real links. I don't know why the blogger didnt't include them, and I don't think this should have gone on the front page without them. Oh well, there's always the comments...

Novel Ferroelectric NAND Flash Memory Cell Demonstrates 10000 Times More Program and Erase cycles than Conventional Memory Cells (AIST press release, surprisingly science-dense).
Highly Scalable Fe(Ferroelectric)-NAND Cell - contribution to the Non-Volatile Semiconductor Memory Workshop, 2008 (you may have access to only the abstract).

This is NOT flash ram, it's ferroelectric RAM. This doesn't matter much to the consumer who can use it much the same way, but it's a different principle. Apparently they've (semi-)tested 100 million r/w cycles, and expect that it can hold data for 10 years (extrapolated from some curve). Besides, it uses a lower voltage than flash, and they expect it to scale down further. Nice. It even looks like it might work. SSDs for teh win :-)

Or...

[Nerdposeur]

...Or, considering that Moses was 80 years old at the time, and was supposed to show these commandments to an entire nation from a mountain top, maybe God was just smart enough to use a LARGE FONT.

What's the "Stick it in a Drawer" lifespan?

[RailRide]

All I've heard so far is debate on how long flash RAM will last while being constantly thrashed with read/write cycles. But what I'd like to know is how long data can be expected to remain intact on a typical flash drive that you just throw a bunch of files on and subsequently stick in a drawer and forget about, or at very least infrequently read from.

I keep a lot of files (mostly art projects) backed up on 2 or 3 seperate hard drives, but while any current project is progress, they tend to reside on a flash drive. Oh, they get rewritten to a couple of dozen times between scanning and completion, but once complete (and backed up) I tend to just leave the project folder on the drive, and when the drive gets close to filling up, I stick it in a drawer and buy another one. Should I expect to be able to call those "retired" drives backups as well, or will the integrity of the data likely deteriorate after a few years like a late model 3.5" floppy

---PCJ