Much has been written on what technology will replace non-volatile FLASH memory, and when. Fortunes have been wagered and lost on both questions. We must find a replacement to FLASH since process shrinks are getting more expensive and difficult. As chip making geometries get smaller, the number of writes decreases and more complicated controllers and control software are needed. This article describes the problem:
34nm MLC NAND is good for 5,000 write cycles, while 25nm MLC NAND lasts for only 3,000 write cycles. To account for this drop in individual cell lifespan, manufacturers increase the amount of reserve capacity that replaces worn sections as the drive degrades.
As a front runner to win this non-volatile replacement race, I favor Micron's (NASDAQ:MU) phase change memory "PCM". This is the technology that initially caused me to look at Micron 10 years ago when their CTO, Tyler Lowrey, left the company and began working with the inventor of the base technology, Stan Ovshinsky, the brilliant deceased founder of the now bankrupted and vanished Energy Conversion Devices. 50 years ago Ovshinsky discovered a class of materials called chalcogenides might be appropriate for computer memory in their change from amorphous to crystalline states.
While it's taken a lot of time to develop, PCM is finally here and is out of the labs and getting mainstream press attention like this "Altered States" article from the Economist 4/21/2013.
There are also some great articles here on SA about some of the different horses in the non-volatile replacement derby.
- An article from last year explains a bunch of the different technologies that are trying to win this race.
- SA's great Russ Fischer, a veteran of the memory industry, wrote this about phase change.
- And here is an article which supposed that the last rev of the iPhone might have included PCM. While that didn't happen I wonder if it might yet?
Sanford Bernstein had this to say about the technology in one of their recent mammoth reports ("Mourning Moore's Law, June 2013"):
PCRAM - Potential DRAM or
Phase-change memory (PCM or PCRAM) is a non-volatile type random-access memory that utilizes chalcogenide glass as a mechanism of action. Heat produced by the passage of electric current switches the glass between two states - crystalline (low resistance) and amorphous (high resistance) - and stores data (see Exhibit 236).
PCRAM is emphasized as a potential substitute of NAND and to some extent DRAM, thanks to its high durability, fast read and random-access ability. PCRAM does degrade with use, but it can theoretically endure 100 million write cycles - extremely higher durability than flash. Also, it does not require erasing an entire
block of cells to have new data to be stored in each cell.
However, as the switching process is thermal, PCRAM is very sensitive to temperature and the process requires a relatively large amount of power. It is already commercially available at Micron, but at a very small quantity and sizable premium in terms of price. Despite several advantages, we believe PCRAM is likely to penetrate only into niche markets as its slower performance than
ReRAM and worse scalability than NAND prevent it from becoming a perfect replacement for either NAND or DRAM (see Exhibit 237).
Despite their negative conclusion in the last sentence, elsewhere in the report they do mention that of all the new technologies under consideration, this is the only one in production. In fact the technology is shipping now in the Nokia Asha phone.
So what are the downsides of PCM? Several things are usually mentioned but all are being worked simultaneously, by Micron and others:
- Cost. Cost per bit in production isn't as readily available for PCM as it is for other technologies. We do know that Micron has already taken production parts from 90nm down to 45nm. As with all things semiconductors, more volume and more shrinks will rapidly lower the cost and price. No other on-the-horizon technology is in volume production, let alone beginning the node size shrink that has already begun with PCM at Micron.
- Speed. While PCM is significantly faster than FLASH (10x to 100x) critics say it is still not fast enough to replace DRAM. There are lots of projects to increase speed including this one.
- Power. while PCM uses no power when not being accessed (and it is "radiation hard" unlike FLASH) it does use a significant amount of power to change the state of the chalcogenide. Various schemes involving material and power handling are already succeeding in the lab to handle this, including this one.
- Scalability. Some analysts and observers are stuck in the past thinking PCM can only have two states. Here is a article addressing that which says in part: "IBM researchers have built PCM memory chips with 16 states (or four bits) per cell, and David Wright, a data-storage researcher at the University of Exeter, in England, has built individual PCM memory cells with 512 states (or nine bits) per cell. "
A PCM based SSD (Solid State Drive): One of the applications I'd like to see Micron and their Crucial memory line debut is a PCM based SSD. The thought would not be a new one to them as they have apparently supported this effort at UCSD on just that. This master's thesis project concluded the following:
The write performance of the late-generation raw
flash devices outperforms that of first-generation Micron PCM. With this in mind, I predict that with next-generation PCM from Micron, a future version of Onyx will outperform the ioDrive in both reads (by 2x) and writes (by 5-6x).
This project was done in 2011 and used a hand built FPGA controller and previous generation 90nm chips with smaller buffers and non synchronous reads and writes. I wonder how fast such a system is running now in Micron's labs?
Conclusion. PCM is the only "next generation non volatile memory" in volume production (per the Bernstein report quoted above and other sources.) As such it is already a leader in the race to win the non-volatile replacement derby. Micron has a strong patent position and a great partner, Intel (NASDAQ:INTC), with another part of the patent portfolio. I am very long Micron stock and options and while I am focused on the near term FLASH and DRAM markets and Micron's upcoming tremendous Elpida acquisition, it is nice to know they have a champion entrant running strongly in the non-volatile memory replacement race.
Disclosure: I am long MU, INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.