IT Jobs
Intel set for first public demo of PRAM
New technology to feature in IDF conference.
By Sumner Lemon, IDG news service
Published: 09:56 GMT, 16 April 07
Intel's chief technology officer Justin Rattner is set to give the first public demonstration of the company's PRAM (phase-change RAM) technology at this week's Intel Developer Forum (IDF) conference.
PRAMis a type of non-volatile memory being developed by Intel and several other companies. The prototype chips are seen as a replacement for flash memory and, perhaps, DRAM. Unlike DRAM, non-volatile memory like flash can store memory when a device's power is turned off, but flash memory chips write and read data more slowly than DRAM, have less capacity, and are also more expensive.
"We're thinking phase change is looking promising and we're going to go into production with that part in the second half of the year," said Rattner.
PRAM is based on chalcogenide glass, which can be altered using the heat generated by an electric current. Heat changes the physical structure of the glass to either a crystalline or amorphous state. Each of these states has a distinct electrical resistance that is used to represent the ones and zeroes needed to represent stored data in binary terms.
PRAM looks set to offer better read-write speed and durability than flash memory, which works by trapping electrons in a memory cell. Over time, electrons inevitably become trapped in these cells and can no longer be removed, rendering the memory chip useless.
"It's not clear there's actually a wear out mechanism because you're just moving this material, chalcogenide, between phase states," Rattner said. "Inherently, that's not a destructive process."
Intel and other companies are counting on PRAM to replace both NOR and NAND flash memory to generate the demand required to produce the new memory chips in volume, and drive down costs.
Flash memory is used in cell phones and other mobile devices, but can also be used in PCs. Flash is also at the heart of Intel's Robson flash memory cache technology. Robson, and similar technologies, like hybrid-disk drives, use flash to improve system performance. Robson and hybrid-disk drives try to anticipate which files will be required by a user and load them into the flash-memory cache, reducing the number of times information must be read off the hard disk.
PRAM could also be used in this way, and Intel is investigating whether the chips could take on an even bigger role. "Ultimately we have to answer the question, is phase-change memory a replacement for DRAM? It seems to have the performance characteristics," Rattner said.
Even if PRAM proves to be a practical replacement for DRAM, that's doesn't necessarily mean that PRAM will replace it completely. "Is it a new level in the hierarchy? Or will DRAM move into more of a new level of the cache hierarchy and phase change will actually be the main memory as we think of it today?" Rattner said.
.gif)
Add your commentComments
Bob | Published: 21:52 GMT, 19 July 2007
Can PRAM be made in high enough density to replace hard disks, rather than DRAM? Doesn't DRAM have access times in the 3 ns range vs. 5 to 30 ns range for PRAM - seems that DRAM still has a slight edge in pure speed; but PRAM based solid state drives would be superior to hard disks in every respect.
burt | Published: 18:38 GMT, 11 June 2007
deep thought: that's an excellent analogy to genetic algorithms. I've been thinking along those lines for a while but never thought to make that relation. I'm pumped to see pram take off. I think they're dumping a lot of money in this because they definitely need a more reliable technology before "flash on the board" and hybrid or solid state drives can become ubiquitous. take away that "spinny thing" and replace it with something that has the perf characteristics closer to DRAM and that makes even your low end systems scream.
Deep thought | Published: 23:41 GMT, 04 June 2007
Developing new technology is something like a search process. Technology moved towards flash memory in the past because it seemed flash is a good type of nonvolatile memory. So moving towards flash was like moving towards a local minimum in the search process :-) Maybe PRAM technology will provide a better minimum (=less cost, better miniaturization etc.), but it can be proven only by searching towards that direction (=research and development efforts). R&D efforts need money and making money from PRAM is only possible if PRAM becomes competitive against flash memory. If we think of the recent momentum of flash technology, it seems PRAM will not ever be able to catch flash. So we are stuck in a local minimum in the search badly. Technology is also being developed like evolution process, and evolution (=genetic algorithms) does not necessarily provide the best answer. It only provides a good answer. And, if they can make money from flash technology, it is a good answer :)
Hugh | Published: 21:57 GMT, 20 April 2007
RAM is great already as a cache to cut down disk reads. Flash is cheaper, so that makes it plausible. What is really needed is a non-volatile write cache. That way synch and similar operations needed to commit disk transactions could be done much much more quickly while still being safe. The first place to put these would be on the disk side of the SATA interface. (Current disk buffers are volatile.)
Bill Davidsen | Published: 15:09 GMT, 19 April 2007
I find it more interesting as a replacement for small disks in some applications, such as routers. Operation from flash is possible today, but the relatively short life presents reliability issues. I'm trying to replace moving parts, but not with something even more likely to fail!