Buckminsterfullerene Strikes Again - Nanotube RAM

putaro writes "Nanotube based RAM, under development by Nantero, promises to deliver densities of over 1 terabit per cm^2, is non-volatile and faster than current DRAM. The Economist has a nice story. Forget about just kicking DRAM's and FLASH's butt, is this finally the end of magnetic storage as well?"

Comments

Re:too bad

[Smidge204]

"At the moment, Nantero has only a working prototype. But the firm aims to have memories on the market within a year."

A lot shorter than ten years, hopefully. Though I'm skeptical we'll see them commercially available within a year...

I hope it would be compatable with existing memory systems, though. It would be nice to just swap out existing RAM for a NanoRAM module and get an instant performance and capacity boost (Providing the controllers don't become an issue, but that's where AMD's 64-bit chip and it's built in memory conntroller come in!).
=Smidge=

And the skeptic says...

[EvilTwinSkippy]

Like any technology, I will only belive it once I can buy it. DVD-Ram is almost there. I'm still waiting for my reflective-LCD laptop. And where did the fuel cells for PDA's go?

Bitter... No, not me.

Re:too bad

[Fulcrum of Evil]

you can't upgrade the memory speed beyond what your entire CPU supports -- you have to upgrade your entire CPU. Which means AMD has to redesign the CPU to take advantage of faster (or different types of) memory. And Opterons aren't that cheap yet...

Not so. The opteron has 3 hypertransport busses which can be connected to alternate memory controllers - the onboard one is then disabled. What I want to know is whether AMD plans to maintain separate part numbers for each speed/controller combo, or if they're just going to band them, with higher clocked Opterons getting faster memory.

Great for security, too!

[Jade E. 2]

Replacing SDRAM (or RAMBUS or whatever) with some type of NVRAM will require a whole new approach to security. Otherwise, when you go home at night, what's to stop me from booting your computer (off a CD or floppy if it's reasonably secure), or rebooting it if you left it running but locked, and running an app that allocates a couple gigs of memory without initializing it then lets me browse it? Encryption keys, passwords, anything that's cached I could get. (Wouldn't care about anything stored on the disk, or other permanent media, I could get those with this method now.) Or, you could just initialize the memory on boot, but then you lose the advantages of nvram like the ability to shut down then pick right back up where you left off.

You couldn't even track it by user at the OS level (user a has memory x and y allocated, so user b can't use that.) because I could still boot it into a different OS through a removable drive...

Of course, you could just eliminate all caches of keys or passwords... But do you really want to have to re-enter your slashdot password everytime you hit refresh, or click on a link to the comments page, or click to read a reply?

Maybe the solution would be to specify a certain area of RAM that would get initialized on power-up (be it a reboot or just waking up from an NVRAM suspend), and get apps to put any sensitive information in that area... Which would probably require additions to your favorite OS's API, in addition to new versions of a lot of apps...

Just thinking 'out loud' here... Anybody else thought about this?

Screw the memory applications....

[poptones]

If this is that fast (half a nS actuation time) and static as well, the implications go well beyond memory applications. OR gates, AND gates and flip/flops (every single nanotube is a complete f/f) are the building blocks of every CPU out there. What about a 128 bit CPU that didn't need an air conditioner to keep from destruction? A CPU with a 1nS clock cycle time and a few MB of on chip cache?

It's a very cool idea, but I'm wondering why they didn't mention these issues. Is it an unmentioned limitation of the technology, or a limitation of the Economist's journalistic scope?

Re:too bad

[kwerle]

Of course, Silicon RAM is to NanoRAM as a kumquat is to a watermelon

No. Silicon RAM is to NanoRAM as Vacuum tubes are to Silicon. They're both there to do the same thing, it's just that nobody bothers with vacuums (except for a very few special purpose - like audiophiles) because they're old and clunky.

The point is that you would have faster, non-volitile RAM that would fit into existing hardware.

No. The point is that you could easily have so much RAM that it would make retrofitting it into a current system look like putting an spoiler on a model-T.

I'd also imagine that this non-volitile RAM uses next to no power, making it a great potential drop-in replacement for laptops and other portable devices.

Good thinking. Oh, and let's not forget you wouldn't need to spin a disk at 1000s of RPM, which uses some energy as well.

This is not an upgrade. It is a change.

The thing I worry about...

[Wolfrider]

--When components start getting this small, the chances of having an potential error occur go WAY up. What I'd be interested to see, is what they're doing to protect against stuff like cosmic-ray bit pollution and such.

--After all, if the scale is NANO, one cosmic ray or stray electro/magnetic field can potentially screw up a lot more percentage of memory... Massive redundancy, high speed and constant bit cross-checking would seem to be a reasonable requirement for these chips.

--For just one example, look what a few scratches can do to a CDR - or worse, a DVD. If you can't read it (use it reliably XMillion times) it's basically not very useful...