Memristor Based RAM Could Be Out By 2009

neural.disruption writes "According to the EETimes, HP is announcing that it 'plans to unveil RRAM prototype chips based on memristors with crossbar arrays in 2009.' I don't know if you remember the earlier story about HP Labs proving the existence of the Memristor that had been predicted in 1971 by Leon Chua, and has the nice property of maintaining a memory of the current that passes by it. This could bring us a new type of small non-volatile high-speed RAM at low cost because of the low complexity of the mechanism employed."

Comments

Security Concerns

[Anonymous Coward]

It seems like this will make recovering encryption keys from RAM much easier. If I understand the article correctly, these devices won't automatically clear themselves or decay like conventional RAM. I'm not quite sure I want this thing in my computer until this gets worked out.

Re:Security Concerns

[fuzzyfuzzyfungus]

OS check on power-on would be plenty safe for error checking purposes. It is a pretty basic application of checksumming. In theory, it could fail; but the odds are vanishingly small.

Against malice, I agree, not useful. Of course, if the attacker has arbitrary read/write access to the nonvolatile RAM, it is game over, period. The logical solution would be to move control over the nonvolatile RAM to the system hardware. Just as the BIOS can restrict the system's choice of boot devices, it could lock or unlock access to the nonvolatile RAM. Lots of fiddly implementation possibilities; but the concept is clear.

The main obstacle to instant-on in computers seems to be the ghastly state of peripheral power management support. All sorts of system peripherals have state that needs to be preserved or restored, some of it dependent on outside conditions(network cards, etc.) and support is uneven at best. The fact that ACPI is said to be a bit of a clusterfuck doesn't much help. Even now, with the various suspend modes, we can get within a couple of watts of what nonvolatile RAM would be like; but support for instant on/off is hampered by peripherals being wonky.

Re:Security Concerns

[epp_b]

You forgot the "get off my lawn!" part at the end of your comment.

Seriously, get over yourself. Instant-on computing isn't just something we want because we're impatient. It's very convenient and productive.

I've stopped shutting my computer down every night in favour of using standby mode. Particularly on a laptop, it works very well. It takes two seconds to go into standby where it's drawing just enough power (read: almost nothing) to retain the data in memory. Then, in the morning when I bring out of standby, it takes only ten seconds until my session from the night before is conveniently restore and I can pick up right where I left off. Sure, a few applications don't like it -- PuTTY's connections are dropped, my email client occasionally balks at not having had a network connection -- but merely reopening a few SSH sessions and clearing my email error log is a whole lot better than a three minute wait from a cold boot and spending another ten minutes getting all my programs running and back how they were (that includes starting up my VMWare virtual machines which, by the way, are completely unaffected by going in and out of standby mode).

Re:Security Concerns

[drinkypoo]

You can also use it to get a recoverable ramdisk, recoverable disk cache, et cetera. Just imagine rebooting and still having files in cache. Your computer could preload for the boot at shutdown time, so you'd get the benefits of a cold boot, and a preload.

Re:Security Concerns

[Shadow-isoHunt]

You know I've posted this on /. like three or four times now and you'd think it'd be more common knowledge by now... but getting encryption keys from RAM is pretty trivial. It's called a cold boot attack.

http://citp.princeton.edu/memory/
http://en.wikipedia.org/wiki/Cold_boot_attack

This attack was sort of one that was under the hat of pentesters and hobbyists until a few months ago when it was rather a do-it-yourself thing, but then McGrew Security made a followup PoC - http://mcgrewsecurity.com/projects/msramdmp/ to the Princeton paper. I played with it right after it came out, and then awhile later threw up a tutorial on remote-exploit. Now, Mati Aharoni's a really smart guy and most assuredly knew about the PoC before I did, but shortly after the tutorial and some discussion on IRC, it's now in BackTrack 3 (http://www.remote-exploit.org/backtrack.html) as a syslinux boot option putting the attack within the reach of everyone.

http://tourian.jchost.net/shadow/liveusb/boot.png

Getting the encryption keys out of the ram dump isn't a point and click operation, but the code's out there and it compiles. People are walking around right now with this on their USB key, and it's the sort of attack that is a real problem that physical access and untrusted users present now. Even without the encryption keys, you've still got the contents of previous webpages, cookies, IM conversations, unencrypted files, and everything else in RAM. Disabling boot from USB doesn't matter much because you can just use a grub CD, and carry around a laptop drive and do dumps on multiple machines. Hell, if you felt like dealing with it you could make it a PXE image... even disabling both boot from USB and CD, most cases in public places(think Dell) can be quickly popped open with the power still on and the BIOS jumper tripped.

Things like this should make you really nervous if you were freaking out about Microsoft's little COFEE ( http://tech.slashdot.org/article.pl?sid=08/04/29/1441215&from=rss ) toy, since it's no more impressive than a customized "Gonzor's Payload" U3 USB Drive ( http://wiki.gonzor228.com/index.php/SBConfig ) with a Microsoft Sticker and this is quite a bit more, well, dirty.

Re:Security Concerns

[miro f]

couldn't you then conceivably use the same non-volatile RAM in the device memory?

Still not holding my breath

[default luser]

They've been saying they'll give us affordable NVRAM without the drawbacks of flash for years, and it still hasn't happened.

MRAM - fast, but not as fast as DRAM. Very low-density.
PRAM - more volatile than flash, because it can change state spontaneously based on temperature (thermally written).
FeRAM - can't be made with cutting-edge processes, and even then can't match the density of flash.
CBRAM - still experimental.

I'll just be surprised if HP can just produce a memory module that is as fast a DRAM, let alone as high-capacity as flash.

Re:Still not holding my breath

[mo]

Every single example you list above is based on the transistor. Sure, there's lots of variations (MRAM using magnets, PRAM using chalcogenide glass, FeRAM using a ferroelectric layer, etc.), but these are all basically: glue stuff on a transistor to store data.

Memristor-based RRAM is different. It doesn't use transistors at all. This is truly a departure from all of the exsting RAM technologies, and while the prospect of RRAM storage is pretty cool, the possibility of analog computing using memristors is even neater. I'm cautiously optimistic that this technology is going to take computing in some interesting directions.

Re:Still not holding my breath

[georgewilliamherbert]

The mechanics of how these work make producing compact high speed arrays easy.

The circuit element is just two stacked planar layers between an underlying and overhead wire. Look at the electron microscope images to see what a row of them looks like... they're no bigger than the contact areas of the wires. A chip of these would be a grid of vertical wires, the active layers, then a grid of horizontal layers. The packing density is approximately wire spacing density.

Speed is good - you send a moderate voltage down one side and see if you get strong or weak signals out the far side, so it's essentially no delay other than speed of electron travel.

Skepticism is one thing. There could be all sorts of gotchas going from a small test area to large chips of this. But the fundamental method of operation is fast and the fundamental area is small, and it works at test scale. This is an extremely promising technology.

Re:Strong AI? Analog circuits rock!

[calidoscope]

And to think that my first analog circuits class was taught by prof Leon Chua...