Intel, Numonyx claim phase change memory breakthrough

Status
Not open for further replies.

Jos

Posts: 3,073   +97
Staff

Intel and Numonyx, the chipmaker's flash memory joint venture with STMicroelectonics, have released details on what the companies hope will be a breakthrough in the development of phase-change memory (PCM). Specifically, the companies developed a new manufacturing process that would allow stacking several memory/selector layers atop each other so memory can be packed more densely in a given volume.

The design paves the way for non-volatile memory chips with larger capacity and greater energy efficiency than current memory types while using a minimum of die space. While such stacking is the goal, yesterday's announcement was of a working 64Mb, single-layer version of the new memory architecture -- multiple layer variants are still on the drawing board.

Phase-change memory stores data in small cells of chalcogenide, a special compound that can change physical states between crystalline and amorphous with the application of heat. The chips can run faster and with greater longevity than conventional transistor-based NAND chips. What's more, phase-change memory is "RAM-like" in that bits can be changed individually, not only in blocks as is required by NAND, yet it is also non-volatile so power isn't required to keep the data in memory.

These attributes could allow the merging of DRAM memory and storage into one high-speed, high-bandwidth architecture. Unfortunately, such a leap is still a long way off according Intel developers.

Permalink to story.

 
That does seem like a promissing technology indeed! SSDs as they are today are already getting quite interesting (Intel's 34nm X25-m for instance). This will be even better.
 
good news indeed if it can be incorporate into affordable unit, then maybe finally we can get cheap ssds.
 
Although it will take time to bring it to average users, the technology they use is quite interesting.
 
I've been busy trying to get every bit of news I can about these new types of memory. That we have come so far is simply amazing to me.

The merging of HDD/SSD with ram would be absolutely monumental to say the least. I can not wait for the day to come when this is possible, everything would run so fast...
 
I was just thinking the same thing, Colonel Lance... Imagine systems with no hard drive architecture as it exists now, just a processor and a ram disk that is instant on and fast as lightning... Add in the optimization and simplification you could do in the entire computing process, and the power consumption savings, and there could be an interesting evolution on the horizon for computing (particularly mobile platforms). Conventional HDD/SSD drives may end up relegated to the roles of mass archiving, offline or portable storage, or just another backup method.
 
yeah if this goes onto be something substantial then no more booting woes,be it windows boot or firefox cold starts :|
 
I love yet hate seeing news bits like this. It makes me super excited about future technology, but at the same time, I just want to see it implemented already!
 
long way off huh? who would be surprised if they had that technology working right now and aren't gonna release it until the slow advancement phases has made their potential profit margins
 
Yeah, I wonder what the cost of devices that sport this new technology will be. Oh, and to the Staff, isn't Paragraph 2, Line 2 supposed to say "...minimum of dye space."
 
This is pretty interesting stuff. There seem to be quite a few new things upon the horizon. USB 3.0, Light Peak, Phase Change Memory...I hope I don't get left out in the cold when I buy a computer soon, only to find out that these new technologies are in computers Only a couple years later. At least the article seems to imply years before Phase Change memory is available.
 
Yes, a long time to go. What I see from using non-volatile high speed memory, the first thing, is faster wake and sleep state transitions for PCs. The ACPI spec acknowledges that the only method to store state data for various peripheral devices distinct from CPU and memory is to use the hard disk or other selected mass storage devices. Primarily this method meant that memory would increase overtime in unpredictably high capacities and that more and more state information would be stored and written to the hard disk in one way or another, but perhaps using a type of memory that would stay consistent despite memory's ubiquitous volatile nature would allow nothing more than perhaps the moving of a few small data structures; the operating system could resume at substantially more responsive speeds. On the matter of merging the physical concept today that conveys John von Neumann's idea, there are already similar systems on several ASIC devices in terms of addressing. In contrast to merging the entire two major modular designs of the componentry, those ASICs simply merge memory and storage in the same address space; The physical merging of these two components in my views would introduce a larger more expansive relationship with the application layer and the driver layer, relinquishing the isolating design of interfacing with the common serial controller and through the hard disk's controller.
 
This sounds awesome, I just hope that the prices of these when they do come out isn't insanely high like imo SSDs are. Oh well I guess, technology prices fall fast and hard usually.
 
Zeromus said:
but perhaps using a type of memory that would stay consistent despite memory's ubiquitous volatile nature would allow nothing more than perhaps the moving of a few small data structures; the operating system could resume at substantially more responsive speeds.

In a perfectly implemented system where RAM and storage are the same, there would theoretically only a need to have a very small section that would imprint the current data within the processing core when a system was shut off - it could be virtually instantaneous on and off if the memory is persistent.

But, therein lies the rub... In order for a system like that to work properly, applications will have to do very effective housekeeping on their own resources. A memory leak in a program today just usually results in your RAM getting packed full and slowing your system down, until the program crashes (or the system does). A memory leak in a program on a computer where the memory is also the storage could lead to a massive amount of garbage being stored in your persistent storage space, effectively ballooning up until there is no room left to operate. Either the programs will have to be very efficient, or a larger overhead monitoring system will have to be in place to prevent runaway memory leaks - either of those situations could potentially slow down the lightning quick potential of an integrated system.
 
This is quite promising since if they can get it to market soon enough it should help drive down the costs for SSD's which we all will appreciate I'm sure. :)
 
It is always nice to see what the distant future holds. Sounds like it is still some time before we see memory of this caliber.
 
I can't wait until I can select how much RAM I can have by dragging a slider on a partition manager XD
 
It would be even more impressive if they could take care of there defective ssd solutions now instead of focusing on something they hope will take away the problems.
 
We've been hearing about this stuff for awhile now, which I'm always happy to hear about a 'new way' of doing things. There was a news post a few months ago about Phase Change Memory: https://www.techspot.com/news/34593-samsung-to-initiate-mass-production-of-pram-in-june.html

Based on some of the reading I've done, one of the *problems* with this type of memory is the 'phase change' itself takes far longer than the random access time of current DRAM solutions... That makes for more than 10x the latency.

This is probably my own lack of understanding too, but in order to change GST (or whatever material may be used) from a crystalline substance into a liquid, heat must be applied. I remember wikipedia mentioning something like 600*C...... If we're going to be dealing with that kind of heat in our computers, it sounds like there is going to be a whole new overheating battle on the way. That heat has to go somewhere and even if it is delivered on a microscopic scale, it can't possible be cooler than what we use currently.
 
This is one trend changing initiative. Increasing memory capacity with same size, is one of the evergreen research fields. By this, I think we can get more out of our RAMs.
 
Status
Not open for further replies.
Back