IBM Research is detailing its quest to find the smallest number of atoms required to store a bit, the fundamental unit of digital data. The answer is just 12, IBM says — a pretty remarkable stat considering that memory in today's PCs has around a million atoms per bit (by our rough calculations, that's nearly 69 quadrillion atoms for an 8GB machine). By aligning the atoms in two offset rows of six with alternating magnetic orientation, IBM figured out that it could isolate the bit so that it wouldn't magnetically interfere with the bits around it.
The company notes that Moore's Law — the longstanding phenomenon dictating that transistor count will double on integrated circuits every two years — is in danger of falling apart as chipmakers start to reach the physical limits of how small transistors can be fabricated. To keep it going, a fundamentally different approach is needed, and assembling memory from individual atoms definitely qualifies as "different." The long-term goal is to produce much denser storage than you can buy today: it's 417 times denser than traditional DRAM and some 10,000 times denser than SRAM, which means you'd have no problem dropping several terabytes of memory and cache into your computer of tomorrow, though there's no word yet on just how long that'll take to materialize.

There are 38 Comments. Add yours.
man, there’s probably a really good pun in here somewhere about Intel processors and patent litigation. somebody else do the heavy lifting.
Posted on Jan 15, 2012 | 6:39 AM EST reply Recommend (1) Flag actions
Intel patents getting its atoms in a row, found using PowerPC processors to do the math. Whateva.
Posted on Jan 15, 2012 | 12:47 PM EST reply Recommend Flag actions
Well done, IBM. You guys can all each have a cookie!
Posted on Jan 15, 2012 | 6:43 AM EST reply Recommend (3) Flag actions
But only one consisting of 12 atoms.
Posted on Jan 15, 2012 | 6:50 AM EST reply Recommend (24) Flag actions
I’m sure they’ll have a byte at that. fp
Posted on Jan 15, 2012 | 9:41 AM EST reply Recommend (9) Flag actions
Nope, just a bit.
Posted on Jan 20, 2012 | 9:04 PM EST reply Recommend (2) Flag actions
omfg, you guys lol
Posted on Jan 20, 2012 | 11:16 PM EST reply Recommend (1) Flag actions
You can just put your cache in the tip JAR over there.
Posted on Jan 21, 2012 | 1:40 AM EST reply Recommend (1) Flag actions
I agree, and after your cookie please start working on battery technology. It’s 2012, all mobile phones should last at least 24 hours with normal usage.
Posted on Jan 15, 2012 | 1:38 PM EST reply Recommend Flag actions
only smartphone has terrible battery life FYI, I typically get 3 to 5 days of normal use out of a 960mha battery on my LG ENV3: its underpowered processor sips power and the tiny screen backlights dont take a big toll either. I use an iPod Touch for iOS apps over wifi and a real camera + dump phone : instead of a smartphone because battery anxiety is weak. + The iPhone’s sealed battery is lame: sure you can get an external case for it with a battery extenders, but 3G data is slow and overpriced for its speed: and 2GB limit before paying more is useless for my browsing. 4G LTE is better, but those phones gobble battery even worse than 3G devices.
I agree with you that some serious battery innovations are needed to really give consumer electronics the kind of performance that people actually want in their mobile devices, tablets, ultrabooks ect. It would be really nice to have a macbook air that could loop HD video for 24 hours and still have enough battery for a few hours of light browsing/ email/ ect.
Other critical advances in batteries: they need much fast charging characteristics with far longer cycle lives. Imaging if your smartphone could pick up %80 charge in 30seconds, and had a standard battery you could keep and put in your next phone for years. We are not there yet, but we need to get there, not just to make consumer electronics better, but also to enable awesome battery electric vehicles.
Posted on Jan 15, 2012 | 4:12 PM EST reply Recommend Flag actions
Tl;dr
Posted on Jan 15, 2012 | 5:15 PM EST via mobile reply Recommend (1) Flag actions
Total;derail
Posted on Jan 15, 2012 | 7:54 PM EST reply Recommend (1) Flag actions
It doesn’t sound as impressive after reading atomic-scale memory is just 100 times denser than todays technology. I didn’t know we’ve almost reached atomic-scale anyway. Also, atomic-scale fabricating won’t help keeping Moore’s Law alive for long – it doesn’t get any smaller than that (or does it – would quark-scale transistors be possible in theory?).
Posted on Jan 15, 2012 | 6:47 AM EST reply Recommend Flag actions
Maybe qubits could help? Or perhaps one could just use smaller atoms? Maybe subatomic computing? Perhaps we could just learn to use the memory more efficiently.
I like the idea of extradimensional computing. Is that a real thing? It should be.
Posted on Jan 15, 2012 | 7:09 AM EST reply Recommend (1) Flag actions
We’re increasing the capacity for memory while being more efficient with memory. Eventually programs and languages will become more efficient and streamlined.
Energy density becomes less of a problem if we move to cloud based solutions.
Alan Kay said that “Most software today is very much like an Egyptian pyramid with millions of bricks piled on top of each other, with no structural integrity, but just done by brute force and thousands of slaves.”.
Posted on Jan 15, 2012 | 7:46 AM EST reply Recommend (2) Flag actions
Extradimensional computing: the one component needed in our dimension is an accessor which acts as a portal to the other dimension that contains all our components, as large as they may need to be! Muahahaha!
Posted on Jan 15, 2012 | 10:51 AM EST reply Recommend Flag actions
“extra dimensional computing” is actually happening. We’re stacking wafers on top of each other, all inside one package. It’s commonly used when you want to incude different types of component, that need to be fabricated using different techniques. You often can’t use multiple fabrication techniques on the same wafer, so you stack multiple wafers to produce the chip.
As long as you can figure out how to link the wafers together (which is easier said that done), you can stack as many wafers as you like (but in reality you often run out of space for the wires).
Posted on Jan 15, 2012 | 7:21 PM EST reply Recommend Flag actions
Good, once the hardware is optimized (has more to do that just how small the transistors are) then society can start focusing resources on producing optimized software that can bring the best possible user experience out of the hardware.
I would not sneeze at the density increase. With large die area SOC designs, its not too far away that all of the components (Hard Drive, Ram, CPU, Cache, GPU, GDR, IO Controller) will all be integrated into a single chip, This means there will be more room for battery, and the battery will last a lot longer because the SOC designs use way less power then discrete component systems.
Also, the bottle neck to mobile computing performance is not just about the mobile devices hardware: the upstream laggy data networks we connect to with our devices is part of the problem. Providing gigabit speed low power mobile broadband technology like 802.11ac, is going to be part of creating that “improving user experience”
As the iPhone clearly proves, its not just hardware specs that make a great product: the user experience is what matters. A phone with a slow CPU that is highly optimized for great performance is better then a phone with a fast wasteful CPU that is poorly optimized giving laggy performance (android handsets).
Most people (not nerds) want robust, smooth, well polished, easy to use hardware. That is why the iPhone sells so well. People like iOS because it is dead simple and super intuitive to use: stupid simple means an easier learning curve = less time wasted learning how to use something and more time doing what you want to be doing.
Posted on Jan 15, 2012 | 4:20 PM EST reply Recommend Flag actions
This post kinda fell of the tracks near the end.
Posted on Jan 16, 2012 | 1:51 AM EST reply Recommend Flag actions
Where does the 100x factor come from? If we have a million atoms per bit today and 12 atoms per bit tomorrow, isn’t that more like 800000x?
Also, this technology only works at temperatures of little more than 0K (-273°C), so in its current state, it’s pretty much useless.
Posted on Jan 15, 2012 | 6:59 AM EST reply Recommend Flag actions
I wouldn’t say that – if it made sense for Google/Amazon/MS to use this technology in their servers (and shrink everything by 800000x) then they could keep a room at 0K to put it in. I have no idea if it would make financial sense.
Regardless it’s still cool.
Posted on Jan 15, 2012 | 7:15 AM EST reply Recommend Flag actions
Data would be extremely volatile. A power cut followed by a generator failure could destroy petabytes worth of data.
Posted on Jan 15, 2012 | 7:43 AM EST reply Recommend Flag actions
Pfft. 1K is nothing, could almost get there without lasers. Plus, the researchers suggest room temperature devices might only need 150-200 atoms per bit. Obviously with such highly constrained nanotechnology there must be redundancy and not every atom in a chip can be a bitmine so the x100 estimation is prolly quite apt.
Posted on Jan 15, 2012 | 11:01 AM EST reply Recommend (1) Flag actions
Thats a lot of science for a Sunday morning…. I read this a couple of times and still my head hurts.
So very small bits then? (Insert pun here)
Posted on Jan 15, 2012 | 7:19 AM EST reply Recommend (3) Flag actions
There’s a typo, up left on the picture in ‘technolgy’
Posted on Jan 15, 2012 | 7:35 AM EST reply Recommend Flag actions
That’s pretty damn awesome. I just like saying “atomic-scale magnetic memory”. Can you imagine the sales call: “Excuse me sir, would you like some atomic-scale magnetic memory?”
Posted on Jan 15, 2012 | 8:51 AM EST reply Recommend (2) Flag actions
Sounds cool, but if we’re already dealing with atoms, how much further can we really go? Where’s the next step when we fill our atomic hard drives with… not porn.
Posted on Jan 15, 2012 | 11:08 AM EST reply Recommend Flag actions
what was the datum…?
Posted on Jan 15, 2012 | 11:48 AM EST reply Recommend Flag actions
Chris, why are you talking about RAM when this obviously is magnetically switched bits (i.e. not electrically)? Seems like you properly should compare it to current magnetic media. It most certainly is NOT applicable for cache and RAM unless you somehow couple every bit with an electromagnet or something…
Posted on Jan 15, 2012 | 12:20 PM EST reply Recommend Flag actions
Apple gets all the credit for innovation. LOL Go IBM
Posted on Jan 15, 2012 | 12:28 PM EST reply Recommend Flag actions
I’m just gonna say this so the future generation can mock us
There’s no way we will require 8 Tb of RAM in our PC or whatever devices that will replace PC by then.
Posted on Jan 15, 2012 | 1:02 PM EST reply Recommend Flag actions
The lines between a non-volatile solid state storage HD / chip and volatile RAM chips are just going to continue blurring. When 4K movies become standard, if you want to have a collection of films with you in a small form factor (microsd) then we need some serious improvements in data storage density!
Posted on Jan 15, 2012 | 4:25 PM EST reply Recommend Flag actions
die stacking the CPU, GPU, Shared High Speed Memory, Shared Slow Storage Memory and IO Controller into a 3D stack would be a huge leap forward over the discrete component model of today with laggy bus interconnects. Putting all of these components on the same die, will give even more improvements. Part of the story will be about getting the same things done with way less energy (for energy conservation in grid connected computers) and (for better battery life in mobile devices). Most mobile devices have already moved to a semi-SOC system design, much greater chip consolidation is possible. Also, hardware drivers need a lot of work, especially for touchpads and GPU’s. Too much effort has been placed on high rate iterative hardware improvements, whilst software optimization has been largely ignored. Most existing hardware would run a lot faster and give the user a greatly improved user experienced if a lot more effort was placed on tuning and optimizing the interaction between the hardware and the software to produce buttery smooth performance. Look at how browser performance on a given OS has improved greatly just with code/ codec/ driver based improvements. More effort should be placed on working out the bugs and really polishing the user experience.
Posted on Jan 15, 2012 | 4:38 PM EST reply Recommend Flag actions
You sure do write a lot.
Posted on Jan 15, 2012 | 5:20 PM EST via mobile reply Recommend Flag actions
I think you’re right.
You can either have one of two design philosophies:
1. provide a rich range of features, at the expense of user experience
2. provide a great user experience, at the expense of having a wide range of features
Personally I think both philosophies are useful in different situations. But to prevent this from descending into a fanboy war, I won’t list which devices I think fall into which categories – I’m sure you can make your own informed judgements. ;)
Posted on Jan 15, 2012 | 7:34 PM EST reply Recommend Flag actions
Every time I think they have reached the limits of hard disk storage, they somehow manage to up it. I remember when a gigabyte was considered mind blowing; this makes a terrabyte hard disk seem quaint.
Posted on Jan 15, 2012 | 7:44 PM EST reply Recommend Flag actions
This is a very impressive feat and a great story but I’m not sure what it has to do with Moore’s Law (which is incidentally not a phenomenon but a postulation) seeing how IBMs research deals with storage, whereas Moore’s Law talks about transistors.
Posted on Jan 15, 2012 | 8:36 PM EST reply Recommend (1) Flag actions
The math here doesn’t add up. If this new method stores 1 bit in 12 atoms and current tech uses 1 million atoms, then “100x more dense” is incorrect. The correct answer can be found by dividing 1 million by 12.
83333.333333333333333333333333333
It would seem to me that this new storage is not 100 times more dense but in fact over 8000 times more dense.
Posted on Jan 21, 2012 | 3:47 AM EST reply Recommend Flag actions
Something to say? Choose one of these options to log in.