Phase Change Memory Points To Future of Storage 70
An anonymous reader writes
"A UC San Diego team is about to demonstrate a solid state storage device that it says provides performance thousands of times faster than a conventional hard drive and up to seven times faster than current state-of-the-art solid-state drives. The drive uses first-of-its-kind phase-change memory, which stores data in the crystal structure of a metal alloy called a chalcogenide. To store data, the PCM chips switch the alloy between a crystalline and amorphous state based on the application of heat through an electrical current. To read the data, the chips use a smaller current to determine which state the chalcogenide is in."
Re: (Score:2)
Hardly skunkworks. PCM chips are in Micron's (Numonyx) catalog. I considered using one in a design last year, but they were too expensive, mainly because I only needed a wee bit of non-volatile storage and these chips only come in 128 Mbit density.
You can buy them from Digikey at $4.57 each. But you have to buy them a tray at a time (576 parts per tray).
Re: (Score:2)
Why thank you for this incredibly useful comment!
Shades of the Foundation! (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
Great for Self-Destruct Drives (Score:2)
But Seriously, wouldn't external heat sources pose a problem for such a technology? I don't want my drive erased next time my electricity goes out during the summer.
Re: (Score:2)
>just need a hair dryer!
Honey?
yeah?
Can I borrow your EZ Bake Oven?
But daddy, the police took it last time...
--
BMO
Re: (Score:2)
If you have issues with your house going 600C without AC you may have other issues. Not to mention high heats have just as detrimental effects on magnets.
Application of heat doesn't sound too hot... (Score:3)
To store data, the PCM chips switch the alloy between a crystalline and amorphous state based on the application of heat through an electrical current.
That seems like probably not a very good idea, and I'm sure it will end up being one of the major hurdles to this technology really getting off the ground. What happens when the memory is heavily used (leading to buildups of heat in the memory that could cause unintentional bit changes)? Obviously the heat used to flip the bits must be dissipated very, very quickly, and that's a pretty challenging problem in its own right. Also, what happens to system temps when you use this memory? Many systems are already difficult to keep cooled, and adding another source of heat could be a particularly bad idea. Heck, what if the increase in system temps leads to memory faults? After looking over the Wiki, the last question seems unlikely to be a problem, but heating something to >600C seems like it could make it very hard to use this tech in heavy use situations.
Re: (Score:2)
I had a PDA with NVRAM and I never noticed it getting more than slightly warm - and if the memory did get significantly hot, it would have drained the battery very fast.
Re: (Score:3)
Re: (Score:2)
I saw this same principle being used on CDs: in the crystalline state, it reflects much more light than in the amorphous. IIRC, to change into crystalline, it required more than 130 C, but take it with a grain of salt. What I'm sure is that it was a lot hotter than 50 C, otherwise I would've thought the same as you.
Re: (Score:1)
Re: (Score:2)
Fortunately, the smaller the memory element, the less heat it takes to raise its' temperature. And fortunately, or unfortunately, depending on your point of view, the smaller the memory element, the faster and easier heat dissipates from it. I'm sure that one of their criteria is to select materials that do not spontaneously switch states at the temperatures typically fo
Re: (Score:2)
It's not as much heat as you might think it is, it's just very concentrated. Same reason you can use a laser to heat something hotter than the surface of the sun without even warming the room noticeably.
Re: (Score:2)
Make that 8.8 gigabits per second compared to 6Gb/s hard drive. Doesn't sound revolutionary. There must be something in the other details that make this exciting.
The only way for the speed comparison to make sense so that we get a 1000x improvement on conventional harddisks and only a 7x improvement on flash is if speed is referring to latency instead of bandwidth, which is correct even if counter to normal marketing material. If the stated bandwidth is for a small element then you can add X of those elements to your drive to multiply the bandwidth by X, so possibly the bandwidth could be pretty good too.
similar issues to flash (Score:2)
Re: (Score:3)
It's bit-erasable. You can write a single word to PCRAM, without having to do the read-erase-rewrite cycle that Flash requires over a large cell. This means that you can just map a PCRAM device straight into RAM and use it as (slowish) memory. This is great for things like applications - the entire binary can be run from the mass storage device, it doesn't need copying into memory (execute in place). For a mobile device,
Re: (Score:2)
It would also get rid of much of the weirdness of SSDs. Most of that is because of the largish erase regions compared to block size for a filesystem and the time it takes to erase a sector.
Moneta (Score:3)
So, what's up with the name of the system, "moneta", is it what I think it is (a Russian word for 'coin'), or is there something else at play here? I am a bit confused, because in the article, (which I am sorry to admit I read), it mentioned a bunch of names, but none that were Russian sounding.
Re: (Score:2)
It's also Russian, though it easily might have come into the language from other language, that's certainly a possibility, OK, but why call this tech a 'coin'?
Re: (Score:2)
OK, that makes more sense than this Coin [google.com]
Re: (Score:1)
Financial Crisis of 2008 explained in 2006 in detail in a video. [tinyurl.com]
Offtopic: Thanks Roman_Mir, that was an excellent vid. Very interesting especially from a foreigner's point of view (kiwi)
I encourage others to view at least the first five mins. I started out with this in mind and was hooked, ending up watching the entire hour-long clip.
Re: (Score:2)
Sure, he has a radio station and a bunch of businesses around the world doing investments and now even banking it seems, so if you liked what he had to say in that vid, you should research some more.
Re: (Score:1)
...so if you liked what he had to say in that vid, you should research some more.
That seems wise. What I've learned today is clearly the tip of the iceberg and some further investigation on my part is in order.
Collectively we younger Western nations seem to fit the 'short-term thinking' mould pretty well; meanwhile those nations that take the long view are just starting to reap their rewards. Those rewards are going to come thick and fast, aren't about to slow down anytime soon and will not include the West.
Re: (Score:2)
Possibly an interesting compilation [slashdot.org] I put together. Just fun to look at.
Re: (Score:2)
Moneta was the Roman goddess of memory.
It was also the name I gave to an set of in-house backup scripts I wrote a few years ago. If only I'd trademarked it!
Re: (Score:2)
Re: (Score:2)
Well, 'Roman' (with the stress on the second syllable) is a common first name in Russian as well :)
Re: (Score:2)
Why is that a surprise, my first language is Russian, not Italian.
Re: (Score:1)
Check this out on Wikipedia:
http://en.wikipedia.org/wiki/Moneta
Goddess of memory...
Density? (Score:2)
I actually skimmed the article and I didn't see any mention of data density. Anyone know how it compares?
Re: (Score:2)
Thank you, Mr. AC!
Interesting but... (Score:2)
Re: (Score:2)
Intel did succeed, but sold the assets to Numonyx.
Aside from heat issues and built up (Score:1)
Sounds like it works fine until you lose power.... (Score:2)
Re: (Score:2)
If it works as described with the PCB generating heat to change the substance from a crystal to a liquid, once that heat is removed via loss of power, all would change back to a crystal, and you would lose all your data. This may be fine for RAM, but not for storage.
Actually I think you'll find it's the same material and a similar principle to CD-RW disks. The difference is that in CD-RW they are heating it with a laser, and reading it back optically. In PC memory, they are probably heating it electrically, and they are using a change in resistance rather than a change in reflectivity to read it back.
Re: (Score:2)
Re: (Score:3)
It's not switching between solid and liquid. It's switching between crystaline and amorphous solid (stated in the first wikipedia intro, second link in the article). It takes a small amount of heat to make it switch between the two states, and they have different properties that can be measured, but both states are static in the absence of heat. What you described would make no sense at all, since if the PCB knows to provide heat to certain places then it already has external memory.
Of more interest is
Interesting comparisons (Score:2)
A UC San Diego team is about to demonstrate a solid state storage device that it says provides performance thousands of times faster than a conventional hard drive and up to seven times faster than current state-of-the-art solid-state drives.
I don't know why, but the speed comparisons in the summary amuse me. "This is THOUSANDS of times faster than a conventional hard drive — absolutely phenomenal speed gains, faster than anything else ever seen in the conventional hard drive world , speeds which will blow your mind straight out of your skull with how much blindingly faster they are than conventional hard drives, literally THOUSANDS of times faster — and kinda sorta faster than solid-state drives."
Re: (Score:2)
It reminds me of this:
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
>>Yes, I didn't know solid state drives were already hundreds of times faster than hard disks.
Heh, you're right. Consumer grade SSDs aren't hundreds of times faster than a HDD. Maybe one order of magnitude faster at sustained combined read/write (my SSD benched in at only 3x faster than my new HDD), but the real gain is in latency. Maybe that's what they meant by "faster". =)
Re: (Score:2)
They are, if you measure in I/O operations per second (IOPS), which is how enterprise-grade storage is often measured.
What exactly have they built? (Score:2)
I'm struggling to understand what these researchers have made, exactly. They certainly didn't invent phase-change memory, and the article states that this "Moneta" uses memory modules from Micron Technology. The wikipedia article mentions Samsung started shipping modules last year, ready for use in mobile applications. So clearly PCM has been available for some time. So perhaps Moneta is an actual device available for end users? That would be exciting!
Re: (Score:2)
"We've found that you can build a much faster storage device, but in order to really make use of it, you have to change the software that manages it as well. Storage systems have evolved over the last 40 years to cater to disks, and disks are very, very slow," said Swanson. "Designing storage systems that can fully leverage technologies like PCM requires rethinking almost every aspect of how a computer system's software manages and accesses storage."
So you're right, they didn't invent PCM, but they're coming at the problem with the assumption that PCM will become commonplace and then looking at the problems that come after that. All the speed and capacity in the world are only helpful up to a point if your software stack and memory bus are bottlenecks.
Re: (Score:2)
I've spent all of two days now reading about PCM, but here's an observation: The lessons learned in making NAND flash work as a high-speed storage medium are applicable here as well. Many of the problems are the same, with the need for wear-leveling and optimization of write performance. The solutions appear to be somewhat different though. Their wear-leveling algorithm does not at all resemble the complexity of a typical FTL and I think that's the point.
Dealing with the problems of getting this technology
Re: (Score:2)
Actually, it's 40 chips for 640MB on a DIMM. The sample they demonstrated was 10GB in total.
Interesting (Score:1)
Interesting, this is similar how cd/dvd-rw works, where they use a laser to do the state change.
Told oyu it was coming (Score:2)
OUM/OVM memory for storage applications. The IOPS should be quite nice once done properly.
Re: (Score:2)
see: http://www.physorg.com/news/2011-05-samsung-mass-30nm-class-gigabyte-memory.html [physorg.com]
Samsung has had 32GB/40nm sticks out for over a year, 16GB out for 2 years, now they are about to ship the 30nm 16/32GB modules with lower power consumption. Price per GB is bad.
Memory4less has:
Samsung 16GB PC3-8500 DDR3-1066MHz ECC Registered CL7 240-Pin DIMM
~$950, other speeds for more $
32GB sticks run ~$2,150 and up
= about 6 to 7 times the cost/GB of vanilla 4GB sticks on newegg, or about 3 to 4 times the cost/GB of 8GB