New type of memory may make HDDs and SSDs obsolete!

A write-up about it here.

It would explain a few things, but it has a few holes in it, too.

But anyway. Woudl this electron memory have to worry about wearing out? SSD has a lifespan of thousands of flips before you start running into read-write errors, and HDD's have moving parts that wear out. Stupid Western Digital drives... The armature motor died, so the heads no longer could move. Lost a lot of data on that drive.

Happiness is zero-gee with a sinus cold.

I think that one of the points made is that it doesn't wear out (technically, all solid state components have a finite life, but the wear would not be a direct function of its use), but I haven't read the full article yet.

edited 14th Jan '11 2:13:23 PM by Fighteer

"It's Occam's Shuriken! If the answer is elusive, never rule out ninjas!"

More along the lines of The Matrix, Barkey, except without the ridiculous justification. Look up digital physics on Wikipedia. /derail

[1] This facsimile operated in part by synAC.

Hell, CDs depend on quantum physics for their operation, since CD/DVD/BluRay players use lasers, and lasers are a purely quantum phenomenon.

the future we had hoped for

Reading the article, they say the important property of the memristor is that it displays hysteresis. If some of you remember back to physics class, maybe you remember when studying magnetism, you'll remember something about Curie temperature and seeing a graph that looks like a loop in the shape of a Diet Sprite symbol. The rate of change of the magnetism depends on which way round it is flipped.

It's a funny turn of events because that hysteresis is a bit of a problem in magnetic hard drives but here, it's the basis of its function. I think that if you were to put the values on this graph◊ force would become voltage and extension would be the distribution of charge.

Looking at the particular materials and their description of it, it seems rather unrevolutionary: you have your dopants, your charge buckets as it were, in sandwiched layers. You apply a current and the buckets move: different numbers of buckets = different resistance. The really unique thing here is that it makes use of the coupling between the current and the motion so that you get a way to move the buckets. This is how it retains memory- the moving of the buckets without a charge would not happen.

Now, I'll have to look at racetrack memory. I hope I like it as much. If only because then it lets me say "Hmm, well I like memristors but I also like racetrack memory. There's only one way to find out...FIGHT!".

For now, it's nice enough to have an SSD. So something beyond that is very exciting. It's good to see Moore's law continue to be in place and not utterly destroyed. And technology keep on steadily progressing.

And if there's a person or people in charge of that simulation, that means God exists. Or even Gods. That's an interesting thought.

Well, if by simulation, you mean what I think you do, and that's a Star Ocean 3 hypothesis of the Universe.

Genkidama for Japan, even if you don't have money, you can help![1]

Claims of increased storage in terms of spatial density from a new type of solid state device always strike me as hilariously dumb. Duh, of COURSE you can fit dozens of terabytes into something smaller than your fingernail, we've been able to do that for decades with any type of chip-based storage. The question is, can you store more data in a solid state device for the same price?

Far more interesting to me, among SSD vaporware, is something I first saw a story about some time around 1998 called MRAM. It's (apparently) faster than DRAM now used as main memory, cheaper and more power efficient than flash, has a longer read/write lifecycle than HDDs, and can store data just fine without external power. If it ever materializes, it would replace every type of data storage for every purpose, ever, possibly even including the SRAM that CPUs use as internal cache.

Eric,

No it wouldn't, because it's several tens of orders of magnitude off the upper limit on storage. The technology always marches on.

edited 15th Jan '11 2:25:15 PM by Yej

Da Rules excuse all the inaccuracy in the world. Listen to them, not me.

Question: was this from a technological blog or general news? If it's general news, I wouldn't pay any attention to it.

Fight smart, not fair.

Various different places. For the present state of affairs in brief, The Other Wiki is probably your best bet: Racetrack memory, magnetoresistive random access memory.

Eric,

Racetrack memory was reported on by the Scientific American a couple years ago.

Blind Final Fantasy 6 Let's Play

Well, I'm looking forward to the eventual price drop on SSD's, anyway. I'd get one now, but they're still a bit spendy for me. Wait a year, and they'll be cheap enough - the cutting edge of today will be main-stream in a year, and old-hat in five, as far as computers go.

edited 18th Jan '11 10:11:30 AM by pvtnum11

Happiness is zero-gee with a sinus cold.

And that's what I love about technology!

"I can't imagine what Hell will have in store, but I know when I'm there, I won't wander anymore."

That makes me sad and feel old. New bottom lines machines are as good as my high grade machine from two years ago.

Fight smart, not fair.

Hey, maybe quantum bits will take off and render our current concepts of what a computer looks like as obsolete as the abacus.

"It's Occam's Shuriken! If the answer is elusive, never rule out ninjas!"

Sorta, you still need the classical cpu for a lot of things. Qubits are only good in reducing certain exponential problems to polynomial times but are otherwise hellish expensive and capable of being wrong (non-deterministic). I see it as being an add-on to your CPU that is later an integrated part, much like the ALU.

If you spend enough time on the processing, can't you reduce the chance of it being wrong to an arbitrarily small fraction?

Da Rules excuse all the inaccuracy in the world. Listen to them, not me.

Yes, but it's still probabilistic.

Been a while since I looked at the complexity zoo, though, shit's confusing

edited 20th Jan '11 10:55:07 AM by Tzetze

[1] This facsimile operated in part by synAC.

I think for most things it does, it is already reasonably small. However, if you have a cheap classical processor to do the same thing but faster, why not just do that? Hence why i think a CPU will have both.

Yeah, but when that probability is 2^260,199 to 1 against, it's essentially deterministic.

Da Rules excuse all the inaccuracy in the world. Listen to them, not me.

But it could take enough time to get down to such a low probability that it would make more sense to use a deterministic processor, and in many applications the speed increase mightn't be worth the (still remotely) possible inaccuracy. I don't know enough specifics about nondetermistic algorithms to know how often that comes up, though, all I remember is Shor's... how many are even written?

[1] This facsimile operated in part by synAC.

I can only suggest a peak at Quantum Algorithm Zoo.

....I barely understand any of that, but that kind of speedup seems like it could push out classicals, damn.

[1] This facsimile operated in part by synAC.

Theoretically- we can break RSA but it's alright because we've also made quantum encryption. Practically, we have quantum encryption and we can factorise 15. Bloody engineers don't have the bits to overthrow the paradigm.

The thing is that really, it's not its own experimental field yet. You have people working on a dozen different implementation possiblities rather than one and each implementation possibility is a field of study in its own right. It feels like we're waiting for a jolt to starting things snowballing. The guys we need to design the coding language and the memory and the keyboards don't have a reason to join in yet.