Design Article
The evolution of phase-change memory
Greg Atwood
7/26/2010 6:38 PM EDT
Phase-change memory (PCM) is a term used to describe a class of nonvolatile memory devices that exploit the ability of certain materials to rapidly change phase between two stable physical states. Phase-change memory blends the attributes commonly associated with NOR-type flash memory, NAND-type flash memory, EEPROM memory, and DRAM (Figure 1) and is a leading candidate for the next generation of nonvolatile memory (NVM), expanding the use of NVM in computing and storage systems. As traditional electron storage-based memories such as NOR and NAND flash begin to encounter scaling difficulties, PCM is considered to be the best candidate to continue the scaling of NVM.
 Click on image to enlarge. |
Figure 1: PCM attributes: This new class of nonvolatile memory brings together the best attributes of NOR, NAND, and RAM.
Here are some of the unique capabilities of phase-change memory:
-
Nonvolatile: Like NOR flash and NAND flash, PCM is nonvolatile. DRAM, of course, requires a constant power supply, such as a battery backup system, to retain information, resulting in higher power consumption. DRAM technologies also suffer from susceptibility to so-called "soft errors" or random bit corruption caused by alpha particles or cosmic radiation, an effect not observed in PCM.
-
Scalability: Existing memories, NOR, NAND, EEPROM, and RAM all rely on charge (electron) storage as their memory mechanism. As scaling progresses to ever smaller dimensions, the number of stored electrons reduces to the point where reliability is degraded and continued scaling becomes very difficult. This, then, requires the introduction of increasingly complex structures with diminishing returns. PCM does not use electrons, but instead a physical phase change as a storage mechanism. Stability of the stored phase has been demonstrated to be at least as small as 5 nm. Scalability is one of the major motivations for the development of PCM.
-
Bit-alterable/erase/software: Like RAM, PCM is bit alterable. Unlike RAM and PCM, flash technology requires a separate erase step involving the manipulation of a large block of data in order to change a small amount of information. Bit alterability can greatly simplify the use of the memory and, combined with the lack of a need to erase, makes software management of the memory much easier. In some usage environments, PCM can be as easy to use as RAM.
-
Write speed: PCM is capable of achieving write bandwidth comparable to NAND, but with 100x lower initial latency (time to write the first byte) and with no separate (and slow) erase step required. The lack of a slow erase step before write can greatly increase the overall effective write bandwidth, and the low latency can allow PCM to be used, in many cases, like a direct write memory as opposed to the mass storage "disk" usage model for NAND. The write speed (bandwidth and latency) of PCM does not match the capability of DRAM, although with proper management some DRAM usage can be displaced for infrequent or managed write applications.
-
Read speed: Like RAM and NOR-type flash, PCM technology features fast random read access times. This enables the execution of code directly from the memory without an intermediate copy to RAM. In contrast, NAND flash suffers from long random access times on the order of 10s of microseconds that prevents direct code execution.
The combination of these attributes positions PCM uniquely with an opportunity to provide the next generation of nonvolatile memory with an expanded set of performance capabilities, sitting solidly between today's DRAM and NAND.
Next: The past:
|
|
|
|
|
greenpattern
7/26/2010 10:53 PM EDT
"In 2008, ST and Intel combined their NOR, NAND (ST's NAND), and PCM business to form a new flash company called Micron. The formation of Micron has further accelerated progress in the development of PCM resulting in the first commercial PCM product at the end of 2008."
You mean Numonyx, right?
Sign in to Reply
unknown multiplier
7/26/2010 11:08 PM EDT
The phase change material interaction with nitride or other materials is nontrivial, especially when current density heating (which moves atoms) is needed to activate the mechanism.
Sign in to Reply
Volatile Memory
7/27/2010 11:54 AM EDT
Incredible! Mr. Atwood claims that phase-change memory writes at 100 MB/s, yet the datasheet of Numonyx (now Micron) own PCM Omneo chip shows that it writes at less than 1 MB/s. When will the lies stop?
Sign in to Reply
unknown multiplier
7/27/2010 12:55 PM EDT
Phase change memory really compares unfavorably with spin torque MRAM and ReRAM in performance and compares unfavorably with Flash and ReRAM in cost. So ReRAM (or some call it RRAM) will eat PCM's lunch.
Sign in to Reply
tipotech
7/27/2010 1:33 PM EDT
Looks very interesting , but from what i see Samsung will beat Micron to the market with PCM.
Samsung has stated it will introduce PCM in Mobile Handset first.
Good Luck Micron
Sign in to Reply
Volatile Memory
7/27/2010 7:41 PM EDT
Samsung lied. Samsung's press release promised that the introduction in mobile handsets will happen by June 30th. Never happened. No product on the market uses any phase-change memory, for obvious reasons. Samsung and Numonyx have both lied repeatedly about the status of their phase-change programs.
Sign in to Reply
JAK620
7/27/2010 3:44 PM EDT
Hey you wrote this whole article about the background environment and history of PCM. I would expect that you know about "Micron". Stating that Intel formed Micron in 2008 is the most ridiculous false statement I have ever seen on EETimes
Sign in to Reply
JAK620
7/27/2010 3:59 PM EDT
"Greg Atwood is a senior fellow at Micron Technology, Inc. He received a M.S. degree in physics from Purdue University in 1979, joining Intel Corporation in the same year. "
And you wrote "In 2008, ST and Intel combined their NOR, NAND (ST's NAND), and PCM business to form a new flash company called Micron."?
Wow. This is supposed to be a professional web site for the industry. and you work for Micron?
Sign in to Reply
zman_tekinsil
7/27/2010 9:02 PM EDT
Deception continues. Mr Atwood did not joint Numonyx. He was with Numonyx all the time, from inception. Numonyx was joint venture between Intel and Microm, mainly.
Sign in to Reply
zman_tekinsil
7/27/2010 9:02 PM EDT
Meant Joint Venture between Intel and ST. Sorry!
Sign in to Reply
patrick.mannion
7/27/2010 11:16 PM EDT
Hey all, let's take it easy on Greg here. This was simply a typo. The article was originally submitted to EETimes shortly before Micron sealed the deal on its purchase of Numonyx. (Yes, Numonyx was formed through a joint venture between ST and Intel in 2008.) In the revised version sent to EETimes shortly after the deal was sealed and right before article posted, a global substitution of 'Micron' for 'Numonyx' had been implemented, which caused the error.
I should have caught it, for sure, but these things happen. I have now fixed it. Thanks for spotting this, guys! Much appreciated, as always.
Sign in to Reply
Volatile Memory
7/28/2010 1:08 AM EDT
patrick.mannion: Why should we take it easy on Mr. Atwood? He is ruining Micron's reputation. The article, as published, leaves the false impression that Micron will engage in the same phase-change memory scam that Numonyx/STM/Intel have been perpetrating for years. Mr. Atwood managed to fool at least one member of the EETimes team, Mr. Peter Clarke, who after reading Mr. Atwood's outdated article, concluded that "Micron ... has indicated that it intends to back the phase-change memory technology" - clearly an unwarranted conclusion, if the article was submitted prior to the closing of the acquisition!
Sign in to Reply
patrick.mannion
7/29/2010 4:28 PM EDT
Hi again. The article was originally submitted by Numonyx prior to the finalization of the acquisition. It entered the queue here (we get a lot of submissions). When it came time to post it, the acquistion had closed and we sent the article back for revision, with Micron. The version you see now, is the version Micron sent back to us.
Sign in to Reply
ahshabazz
7/30/2010 2:44 AM EDT
Our memories dont rely on electronics our memories use dark energy - :}
Sign in to Reply
photo recovery
7/30/2010 4:35 AM EDT
The technology is indeed a productivity booster, and there are easy solutions to privacy issues. Phase-change memory blends the attributes commonly associated with NOR-type flash memory,
Regards,
Data Recovery Software
http://www.datadoctor.biz
Sign in to Reply
Helicopter
8/3/2010 1:33 AM EDT
I thought PCM write speed is less than 10MBps...
But if the argument is PCM write speed is 100MBps, then NOR could also be architected to achieve the same.
Sign in to Reply
unknown multiplier
8/3/2010 4:28 AM EDT
They cannot be doing block writes, only individual cells.
Sign in to Reply
olearydq
7/10/2012 11:48 AM EDT
Figure 2 appears to contain errors in the "Properties" column. The amorphous phase should exhibit low reflectivity and high resistance The polycrystalline phase should exhibit high reflectivity and low resistance.
The text and caption state this also.
Regards,
Daniel O'Leary
Fort Worth TX.
Sign in to Reply