Slashdot Mirror
Carbon Nanotube Memory on the Way
Cyberherbalist writes "Nantero, a nanotechnology company, is expecting prototypes of products using NRAM technology (nanotube-based, non-volatile random access memory) to be available in 2006. In the article at nature.com, it says that 'the company has succeeded in making circular wafers, 13 centimetres in diameter, that hold 10 gigabits of data.' And they are ten times faster than 'flash' memory."
It's a prototype for Christ's sake....it'll get smaller with time. You ever see a string of iron core memory? That stuff was big too, but it got smaller.
Geek used to be a four letter word. Now it's a six-figure one.
And how thick is a wafer, exactly? compared to say, a disk platter? If they attain the access time mentioned in TFA, it is rather impressive. High-capacity, fast, thin, and preferably shiny things are always impressive.
Nothing but the finest in meaningless drivel
Is there anything you can't do with carbon nanotubes?
So far I have heard that they will be the next steel, the next silicon, the next communication line, the next display medium, the next fabric, the next medicinal treatment. I just want to know is will this change the world in the same fundamental way that mineral oil did in the 20th?
For one, it's first-gen stuff. It will likely gain density quickly in the future. Also, don't forget this is basically NVRAM: way faster than a hard-drive, and way more permanent than DRAM. It fills a unique niche and cannot directly be compared to or replace either of the two. (Well, it could replace hard drives, if it shrunk enough). The day is coming (slowly) when the primary storage on any computer system will probably be some sort of nonvolatile solid-state device. Hard drives with spindles will be for bulk data (music, movies, documents), while the OS goes on the nonvolatile ram which is neccesarily much smaller in size, but more reliable and faster to access. You can do things that way now under Windows or Linux by buying a 1-4 GB-ish solid state flash disk for your root disk (or C: drive) and then putting in a large normal hard drive for all your bulk data, but current SSD technology is overpriced and suffers from various little problems, both of which make it impractical for mass deployment even if the OS vendors put more thought into supporting the setup.
11*43+456^2
...does it scratch easily?
Ever heard of the FLASH hard drives? Well, think about it. 10 times faster than flash drives, no limit to the number of write operations... this has the potential to become the *ultimate* hard drive technology. Silent hard drives, which can be operated no matter the humidity conditions (7500+ RPM hard drives don't operate very well in places with high humidity and heat, i.e. coasts)...
Only one thing concerns me, tho. 2 years ago, Nantero had announced the fabrication of their nanotube memory. I was skeptical then, and I'm still a bit skeptical. What if their prototypes don't work? Will Nantero suffer the fate of so many dot-com's we all (don't) know about?
Only time will tell.
Recent story
If you don't know where you are going, you will wind up somewhere else.
When they got bugs in the system, they could correct the memory by hand with a magnet...
Ah, those were the completely off-topic days.
Justin.
You're only jealous cos the little penguins are talking to me.
Why would you do that?
Current DRAM chips were there years ago: current DRAM chips are around 1Gbit per square centimeter. On a ~5" wafer, this means ~40 potentially working chips per wafer and 40Gbits/wafer, four times as much.
And as far as downtime reduction goes, NRAM would be no good unless the server has time to suspend-to-RAM... so you would still need an UPS or ultra-capacitors to cover this.
For one, it's first-gen stuff. It will likely gain density quickly in the future.
It may gain density, but gain little in reliability. I have a Ph.D. in solid state physics, so I should respond. The carbon nanotubes bend to make connect with an electrode, so something moves. This is usually a bad sign for long-term reliability. Ask telecom technicians if they would like to replace their solid state transistor-based switches with moving switches. Their answer will be that these moving parts wore out.
This memory might not improve to the point that it becomes more reliable than your present NVRAM.
It is very easy for a scientist to produce one working device in the lab. We call these hero devices. The rest of the world does not know this. When engineers get ahold of these claims, though, they tear them up, since the process might not be cheap, reliable or scaleable.
So yes, it might get better, but I wonder if this group and the related scientist have invented new physics. Have they fundamentally changed the way mechanical switches make contact, the way electrons move and are held in capacitors, and the way domains set up in magnetic memory? I think not. I think this is a step backwards towards old mechanical swithces made smaller and reliable memory made unreliable.
Well, unless the server was written using memory transactions, which are starting to look like a good idea for other reasons also. If you had a transactional layer on top of your NVRAM, then you could structure things to allow crash recovery as well; then you could recover from any crash at any time.
See Nantero's press releases; they announced their 10Gbit wafer in May 2003. Their partnership with LSI Logic isn't news either, that was announced June 2004. The fact that they're still signing new partnerships on a steady basis tells me this technology is not a dead-end (yet).
-- *My* journal is more interesting than *yours*...