Print 45 comment(s) - last by Spyvie.. on Mar 12 at 3:58 PM

Phase change memory wafer manufactured at 90nm
Intel claims it will mass produce phase change memory before the end of 2007

This week Intel privately shared parts of its roadmap for memory technologies through 2008. Intel’s progress on phase-change memory, PCM or PRAM, will soon be sampled to customers with mass production possible before the end of the year.

Phase-change memory is positioned as a replacement for flash memory, as it has non-volatile characteristics, but is faster and can be scaled to smaller dimensions. Flash memory cells can degrade and become unreliable after as few as 10,000 writes, but PCM is much more resilient at more than 100 million write cycles. For these reasons, Intel believes that phase-change memory could one day replace DRAM.

“The phase-change memory gets pretty close to Nirvana,” said Ed Doller, CTO of Intel’s flash memory group. “It will start to displace some of the RAM in the system.”

For its implementation of phase-change memory, Intel has since 2000 licensed technology from Ovonyx Inc.. The Ovonyx technology uses the properties of chalcogenide glass, the same material found in CD-RW and DVD-RW, which can be switched between crystalline and amorphous states for binary functions.

Every potential PCRAM memory maker thus far licenses Ovonyx technology. According to Ovonyx’s Web site, the first licensee of the technology was Lockheed Martin in 1999, with Intel and STMicroelectronics in the following year. Four years after that, Nanochip signed an agreement.  Elpida and Samsung were the next two in 2005, and Qimonda marks the latest with a signing this year.

IBM, Macronix and Qimonda detailed last December its recent developments on phase-change memory. Researchers at IBM’s labs demonstrated a prototype phase-change memory device that switched more than 500 times faster than flash while using less than one-half the power to write data into a cell. The IBM device’s cross-section is a minuscule 3 by 20 nanometers in size, far smaller than flash can be built today and equivalent to the industry’s chip-making capabilities targeted for 2015.

Intel’s initial phase-change technology, however, is already a reality, as the chipmaker revealed that it has produced a 90 nanometer phase-change memory wafer. At the 90 nanometer process size, the power requirements to write are approximate to that required for flash. Intel said that its early test work shows data retention abilities of greater than 10 years even at temperatures of 85 degree Celsius.

Intel touts PCM as a “new category of memory,” as its attributes are distinctly different, and typically superior to many of the memory technologies today as it combines the best attributes of RAM, NOR and NAND. Intel wouldn’t give a firm date on the availability of its phase-change memory as several details still need to be finalized after the sampling process.

“We're going to be using this to allow customers to get familiar with the technology and help us architect the next generation device.” Doller said. “We're hoping we can see [mass] production by the end of the year, but that depends on the customers.”

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

By Mudvillager on 3/9/2007 5:38:15 AM , Rating: 5
With that many big companies backing PCM/PRAM there's no doubt that this will replace NAND.

500 times faster than traditional NAND = system RAM and storage becomes one. Probably the biggest breakthrough since the computer was invented.

RE: Promising
By nurbsenvi on 3/9/2007 5:57:25 AM , Rating: 1
Damn right!

RE: Promising
By TSS on 3/9/2007 7:03:40 AM , Rating: 3
I'm not convinced yet this will bridge the gap between data storage and system memeory. it's a step in the right direction but no replacement. for it to actually do that it would have to rival RAM's speed with the HDD's capacity & price.

remember its great to have a 3.5" HDD of 2 TB's from this stuff but if that costs 5000+ bucks thats not going to help any.

However, if they manage at least 40GB's on a HDD sized disk for about €100 (i can get a 320GB disk for that) with significant (speedy gonzales signficant) speed improvements that come close but not quite to DRAM, i would buy it.

RE: Promising
By zsdersw on 3/9/2007 7:17:12 AM , Rating: 2
You don't have to be "convinced" of anything.. you just have to recognize the potential, which is all that matters with technologies that are a few years down the road.

RE: Promising
By fliguy84 on 3/9/2007 7:37:04 AM , Rating: 2
Can't wait for my 1TB nanoSD card :)

RE: Promising
By MrEMan on 3/9/2007 9:03:12 AM , Rating: 5
Yeah, like recognizing the potential of Itanium a few years back...

RE: Promising
By zsdersw on 3/9/2007 10:39:40 AM , Rating: 2
If you think this is "like" the potential of Itanium, you're a fool.

RE: Promising
By zsdersw on 3/9/2007 10:45:28 AM , Rating: 3
And let's not forget.. Itanium's potential as an architecture was never its problem. The problem was the marketing, the apathy toward it by the x86 market, and the path-of-least-resistance that x86-64 offered.

RE: Promising
By deeznuts on 3/9/2007 1:06:12 PM , Rating: 3
I'm not an Itanium expert, but I thought Itanium and x86-64 are two completely different markets. x86-64 competed with the Xeons right? Itanium was built to run some very specialized programs I thought.

RE: Promising
By zsdersw on 3/9/2007 3:42:10 PM , Rating: 2
Itanium (EPIC) and x86-64 are different markets, but Intel once envisioned that EPIC would be the 64-bit market.

RE: Promising
By darkpaw on 3/9/2007 4:20:17 PM , Rating: 4
Yah, cause they sorta underestimated the willingness of everyone to rebuild all of their x86 based programs to a completely different architecture.

It was really a bad idea from the start, since it had 0 backwards compatibility.

RE: Promising
By zsdersw on 3/9/2007 6:28:05 PM , Rating: 2
The architecture was not a bad idea at all. Attempting to take it where it wouldn't fit right was the only bad idea.

RE: Promising
By iollmann on 3/11/2007 12:09:42 AM , Rating: 3
AMD is to blame here, as they admitted when they introduced the x86-64 ISA. Transitioning CPU architectures is possible. Apple has brought its users kicking and screaming (or drooling impatiently as the case may be) through two CPU architecture switches over the years, and is in the midst of a 64-bit transition, which will likely play out without a hitch. x86 has no backward compatibility features for PowerPC. PowerPC had only minimal similarity to mk68k

The difference between the Windows and MacIntosh experience is that in the Mac case, users had no choice. With each switch, Apple lost a few old customers, but picked up a bunch of new ones.

Apple also controls the operating system. I have no information about Microsoft's level of enthusiasm for EPIC. Microsoft is in a position that they could convince business users that they have no choice. (Evidence of industry boycotts of Vista notwithstanding.) However, whether they wanted to go through the effort to write and support a emulator is another question. Microsoft is playing a defensive game in the operating systems market. Bold moves don't do much except endanger their position of dominance. Itanium is likely nothing but a big giant expense for Microsoft.

RE: Promising
By oTAL on 3/12/2007 11:05:31 AM , Rating: 2
Two points are worth mentioning about Itanium...

1. It is STILL a promissing technology... anyone who can't see that needs a pair a glasses... will it ever deliver on the promise? No one really knows... (anybody who thinks he knows is wrong...)

2. Yes, It competes with x86-64. Two different architectures can compete the same way macs and pcs compete for a place in your desk. The latest failures for Itanium came when Opterons and low priced Xeons (prices forced by AMD competition) started eroding into existing and potential costumers.

One of the largest semiconductor companies asked my company for an opinion on the possibility of architecture shift on some software we deliver (decision support software, high power necessities - large databases). Lots of different SOs and architectures have been used by this company - the code I work on has flags (currently never used) for stuff like Alpha, Spark, etc. They use HP-UX, they have unix systems, windows systems and I believe they may still have some Solaris stuff. These technologies compete with each other! Maybe with higher inertia since there are very large investments at stake, but this competition exists.
When we were asked we answered - there are some choices... each have ups and downs... our opinion is still that x86 is here to stay and it requires the smallest investment on almost every level...
Still, I can see Itaniums being usefull for many things... and they do offer blistering speed for optimized code... it's just too much of an investment to buy expensive hardware and develop expensive software...

On a side note...
AMD is to blame here, as they admitted when they introduced the x86-64 ISA.

I'm going out on a limb here and say that AMD is probably quite proud of carrying that blame... ;)

RE: Promising
By Hoser McMoose on 3/11/2007 1:42:13 PM , Rating: 2
And let's not forget.. Itanium's potential as an architecture was never its problem. The problem

Let's see, the Itanium was designed to minimize transistor use at a time when transistor count was quickly becoming a less important factor. It was designed to use a simple and elegant instruction set that ended up being neither simple or elegant. It was supposed to fix problems that didn't yet exist at the expense of not breaking a major issue (good backwards compatibility) that DID exist.

I suppose you could argue that a lot of these failures were more due to the implementation of the architecture rather then it's basic ideas, but to say that Itanium failed JUST because of "apathy" is 100% false.

Itanium is purely an example of something that looks good on paper, but when you take a really good look at ask the hard questions, it weaknesses become readily apparent. This was true as far back as the late 90's and it hasn't improved any.

It remains to be seen if the same is true with this memory or not. Again it looks good on paper, but there are still some hard questions that need to be asked. The key is how fast is the memory, how dense is it and especially how expensive. Those three factors will help determine whether or not it has a place in the computer industry and just what the place will be. Fortunately Intel does seem to be in the process of trying to answer these questions, so perhaps by the end of the year we'll have some more concrete information.

RE: Promising
By zsdersw on 3/11/2007 3:07:33 PM , Rating: 2
Show me where I said anything about Itanium's situation being the result of "just" one thing or another.

RE: Promising
By Oregonian2 on 3/9/2007 3:10:58 PM , Rating: 2
Phase-change memory was touted to revolutionize memory technology in the 70's (last century). Had "potential" to revolutionize the world, etc. Been around for a long time. Sounds like it may actually make it to commercial product which is good (finally). But I think I'll reserve enthusiasm at least until after it exits being the vaporware that it's now (and has been for about 30 years).

RE: Promising
By zsdersw on 3/9/2007 3:45:40 PM , Rating: 1
Had the potential to revolutionize the world? It still has it.

RE: Promising
By GTVic on 3/9/2007 6:41:18 PM , Rating: 1
wioll haven be

RE: Promising
By zsdersw on 3/10/2007 8:35:50 AM , Rating: 2
Umm.. what?

RE: Promising
By macmikey on 3/12/2007 11:08:02 AM , Rating: 2

Douglas Adams lives!

RE: Promising
By rodch on 3/10/2007 1:29:05 PM , Rating: 2
I vaguely recall a late 70s Electronics article about magnetic bubble memory (also touted by Intel at the time): "The technology of the future - Always was, always will be"

RE: Promising
By zsdersw on 3/11/2007 6:25:48 AM , Rating: 2
The inability to see the forest for the trees often manifests itself in the use of the past to wrongly or irrelevantly downplay the future.

A lot has changed between the late 70s and today/tomorrow.

RE: Promising
By Cogman on 3/9/2007 8:52:12 AM , Rating: 2
This sounds like great news. hopefully the prices will be less then DDR. The problem they will have is convincing the average user of the benifits that it has. Most people not know that NAND degrades over time, and they usually only use it for storing their word docs and music. If this tech can become popular, however, We are going to see some interesting changes with regards to memory.

I think I read somewhere that PRAM has read/write speeds comparable to SRAM (this may not be true, it could be one of the other Future memory types) if this is true, then say hello to CPU's with 16, 32, 64, 128, 256, 512MB of cache! What I do know for sure is that it will have speeds the rival DRAM, how you would make it into a flash drive, I'm not sure (as USB would be a bit too slow for it maybe using an eSATA port would be the answer) Imagine though, Vistas ReadyBoost should actually give some big performance increases when coupled with one of these drives.

Again, price is what will sink or sail this product. One things for sure, I will probably end up buying one.

RE: Promising
By ElJefe69 on 3/10/2007 9:43:27 PM , Rating: 2
on AMD, would be cool to go directly to an ht channel thingie.

If caches were 512Megabytes, programs would go insanely fast.

By ksherman on 3/9/2007 8:23:07 AM , Rating: 2
they can mass prodeuce before the end of the year... Wow. Usually this is something they say will be availible in like 2016 or something. I for one am glad to see new a new flash RAM take over as I never had the feeling that NAND flash was really capable of giving us SSDs that are 'good' enough.

Intel is saying that they can mass produce by the end of the year, but it ultimately depends on the customer. What does that mean? Are there people out there that dont want this memory out on the market?

RE: If
By therealnickdanger on 3/9/2007 8:37:43 AM , Rating: 2
I found this kinda of puzzling as well, but that's only because I still haven't quite gotten used to Intel being so open and forthcoming about future projects. Basically all they are stating is the obvious. "If they build it, will people come?" I can almost ensure that the "comsumers" they refer to are flash memory vendors and government/military types. I doubt this tech will be affordable to the mainstream for a while. However, we won't know until they release it!

RE: If
By Cogman on 3/9/2007 9:05:59 AM , Rating: 2
it isn't going to be THAT expensive. One of the prime qualities of PRAM (according to the Wiki) is the fact that it is cheap to make. Initially, you will probably see prices like 100 - 200 for 512, but that will drop easily before 2016.

RE: If
By Scabies on 3/9/2007 10:02:29 AM , Rating: 2
Yeah right, you know that people are going to buy it anyways, provided the price difference between a 1gb flash drive and a 1gb pcm drive is acceptable. The average consumer cares nothing about performance, and will settle for function and price advantages as priorities. Note this does not apply to must DT regulars.
In fact, I imagine most people are totally unaware of how flash memory works, and that it has a somewhat predictable lifetime. Revolutionary technology means nothing to them, since what is currently available 'works' until someone says "think of it as flash 2.0" and/or says "they cost the same, but this one is better"

(also, someone mentioned eSATA. speaking of lifetimes, how often can you plug/unplug from eSATA? I know SATA (normal plastic) connections have less than a hundred connects/disconnects before the terminal starts breaking...)

RE: If
By jak3676 on 3/9/2007 4:11:54 PM , Rating: 2
Looks great, but with mass production possible this year, I would have liked to have seen some discussion on price. How does it compare to NAND, NOR or traditional DRAM? If its too early to look at end consumer pricing, perhaps they can discuss the complexity of manufacturing. If Intel can print wafers using their current 90nm manufactuing equipment that will at least give us an idea about price.

By Rad T on 3/9/2007 8:32:19 PM , Rating: 2
It is an exiciting technology and it will probably come at a price not many are willing to pay. I am a bit suspect of it when it comes to reliability. Have you ever had data evaporate from a CD/DVD? If this is based on the same substance found in CD-RW, it may not exactly age gracefully...

RE: Reliability
By TomZ on 3/10/2007 5:32:13 PM , Rating: 2
Have you ever had data evaporate from a CD/DVD?

No, I haven't. Have you?

Anyway, even if there is a reliability problem with CD/DVD that I've never heard of, it doesn't mean that the same problem would exist in PRAM.

RE: Reliability
By Spyvie on 3/12/2007 3:58:40 PM , Rating: 2
Yes, though don't think "evaporate" is the right word, I have seen several optical disks "go bad"

I have in my collection a copy of Saving Private Ryan, not a scratch on it, a perfect condition store bought DVD that wont read anymore on any DVD drive.

I wouldn't get ahead of yourselves just yet
By Egglick on 3/9/2007 10:58:11 AM , Rating: 3
PCM sounds like great stuff.....much better than current flash memory certainly, but I'm not sure I'd call it a "replacement for system ram" just yet.

First of all, we have no idea what its latencies will be. It might have great throughput, but that doesn't mean it can hang with DRAM on latencies and access time.

Then there's the issue of write cycles. 100 million might sound like an incredibly huge number.....until you realize how many times system ram is written to and erased. Used as system ram on a heavily used system, this stuff might only last a couple years. Maybe less.

By jkostans on 3/9/2007 11:12:59 AM , Rating: 2
I was thinking the same thing regarding system ram. 100 million write cycles is nothing, and would be dead in less than a few hours if heavily written to at 800MHz. Maybe I'm wrong though.

SSD's anyone?
By sonoran on 3/9/2007 11:58:52 AM , Rating: 2
Bring this down to $1/GB, package it in an SSD, and my slow old mechanical hard disk is out the window!

And we've all heard of Robson technology (flash cache to reduce hard disk accesses, coming to a laptop near you soon). Imagine using this instead. Much faster, lasts much longer. Oh yeah.

The other big plus - being solid state the density follows Moore's law as long as it can keep scaling to smaller process technologies. That means 2x the storage in the same space, or the same storage at 1/2 the cost, about every 2 years. What may be expensive today is affordable in 2 years, and cheap 2 years after that.

RE: SSD's anyone?
By PandaBear on 3/9/2007 1:08:02 PM , Rating: 1
NAND is at $0.50 / GB and it is still too expensive to replace HD at the moment.

RE: SSD's anyone?
By ksherman on 3/9/2007 1:41:59 PM , Rating: 2
What? $0.50/GB for Flash? Where do you get those prices?! I think its closer to $10/GB for the SSDs!

By KernD on 3/9/2007 11:25:50 AM , Rating: 2
The BIOS simply overwrite the start of memory with the OS start, so that would be no problem, and then you have your OS's memory manager which initialize to memory is all free except for the OS boot, so it would consider all the memory to be empty and allocate on demand, and in software you must never assume that memory is all at zero after an allocation, so you wipe it out or overwrite it, so there is no problem there.

By ksherman on 3/9/2007 1:43:08 PM , Rating: 2
Better yet, nix the BIOS. We are moving that way, are we not?

Thats worth talking about
By MadAd on 3/9/2007 4:03:47 PM , Rating: 3
.....could be turned off and then turned back on immediately or 10 years later and start right up where the user left off

.....not be subject to critical data loss when the system hangs up or when power is abruptly lost power required to maintain memory

.....cost advantages over conventional solid-state memories such as DRAM or Flash

.....smaller die sizes without the increasingly exaggerated topologies

Well that sounds enough to get me interested, if you throw in the possibility of additional bandwidth too, im there!

Intel will be sued by Ramtron
By holoman on 3/10/2007 9:14:44 PM , Rating: 3
Chalcogenide is a ferroelectric material I am sure Ramtron who has over a hundred patents is waiting in the wings for Intel to bring this out.

I feel a major lawsuit in the works !

Intel's initial capacity?
By psychobriggsy on 3/9/2007 2:35:33 PM , Rating: 2
I read elsewhere that Intel's PCM chip will be a 16MB device. Is this correct? As a first generation technology low capacities are expected, and should scale up rapidly I'm sure.

"The IBM device’s cross-section is a minuscule 3 by 20 nanometers in size, far smaller than flash can be built today and equivalent to the industry’s chip-making capabilities targeted for 2015"

This is promising however.

As others have said - make it cheap and it'll take over. What's the point in making flash if you can make something that's 500x faster, smaller, and uses less power as well?

invested in phase change
By goldenroad on 3/10/2007 3:35:17 AM , Rating: 2
TYLER LOWERY and Ward Parkinsonfrom from the old Micron own (30%) , INTEL (30%) and the shareholders of (Energy Conversion Devices (40%) are the owners of Ovonyx. If anyone wishes to invest in phase change memory you must buy shares of Energy Conversion Devices, ticker (ENER), Dont forget to thank me next year when your RICH! RICH! RICH!.
Stefan Lai recently left Intel to Join Ovonyx.
Things are heating up.

"You can bet that Sony built a long-term business plan about being successful in Japan and that business plan is crumbling." -- Peter Moore, 24 hours before his Microsoft resignation
Related Articles
Solid-state Drives Ready for Prime Time
November 15, 2006, 3:45 PM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki