Memory and Storage - Part 2: New Memory Technologies

by [Published on 2 Aug. 2007 / Last Updated on 2 Aug. 2007]

Data storage up-and-coming technologies

If you missed the other part in this article series please read

If you would like to be notified when Russell Hitchcock releases the next part in this article series please sign up to the Real Time Article Update newsletter.

In my last article I explained to you how some of the more common memory storage technologies work. All of these technologies have found their place in today's typical network. In this article I will discuss technologies which have not yet found wide acceptance. In some cases these technologies are still only found in laboratories, in other cases the products available today do not fully reflect the potential of that technology.

Molecular memory

So what is wrong with the memory storage technologies explained in my last article? Well, nothing. The motivation for developing new memory storage technologies is that we are quickly reaching the limits of how small and how fast we can make these things while users are demanding more capacity and better performance. New technologies will soon be needed. Is that technology molecular memory? Maybe. What makes molecular memory attractive is that even large molecules are very small and could provide a memory density many times greater than current technologies.

To hold a bit in a molecule is, in theory, quite simple. You simply add or subtract electrons to the molecule. The difficult part is reading and writing the bits. You cannot exactly have a copper wire leading to a molecule; the size difference would be inhibiting.

In order to access the molecules for reading and writing, some researchers have been arranging arrays of molecules around tiny nanotubes capable of carrying electric charge. This method is shown below in figure 1. Other researchers are trying to manipulate bits via radio waves. Did you say radio waves? Absolutely. Pretty cool eh! They do this by creating an electromagnetic pulse of a specific frequency which would alter the charge of the molecule. To read the bits another pulse of a different frequency is then created. The effect the molecule has on this second pulse can tell you what the first pulse did to the molecule, hence allowing you to store and then read a bit.

Figure 1:
A diagram of a molecular memory device from

As you can see, molecular memory holds great promise to provide the user with a large memory density. Currently, molecular memory is still in the laboratory stage so we will all have to wait a few years to find out what this technology can really do for us.

Phase change memory

Unlike molecular memory, phase change memory is available today. In fact, the technology behind phase change memory has been around for decades. In the 1960s Stanford Ovshinsky invented a way to crystallize amorphous materials, that is, materials without a crystalline structure.

As mentioned in my last article, CD-Rs, and CD-RWs work by a laser changing the opacity of a small region on a disc. What changes the opacity is the fact that the material changes from amorphous to crystalline, or vice versa. This is the same technology invented by Ovshinsky. Ovshinsky actually made a prototype CD-RW in 1970!

The difference between CDR technology and phase change technology is that with phase change memory the crystalline state of a small region is altered with an electrical current not lasers. Since we are not using lasers to read and write the data we do not refer to the opacity of the region but rather the resistivity of the region. Once the region is changed to either crystalline or amorphous, the resistivity of the region can be measured and based on the amount of resistivity the region is considered to be either a '1' or a '0'.

As a side note, you should now be able to see that electrical resistivity is really quite similar to opacity. Where a resistive material does not allow much electricity to flow through it and an opaque material does not allow much light to pass through it. You should also know that opaque materials actually reflect light. You may not realize that resistive materials also reflect. More correctly, it is the impedance of a material which reflects the electricity. Resistance is one aspect of what makes up impedance; the others are capacitance and inductance. In many applications, limiting this reflection by impedance matching is a major design consideration.

Phase-change memory has the potential to replace flash memory in only a few years. How does it compare to flash? Like flash, phase change memory is a non-volatile random access memory making it well suited for both code execution and data storage. In 2006 IBM, along with Macronix and Qimonda, announced research results which stated that they had designed, built, and demonstrated a prototype phase-change memory device. This device was 500 times faster than flash while using less than half of the power. The prototype device was also far smaller than flash memory.

Another sign the phase change memory will soon be upon us in great numbers comes from Intel. In April of this year Intel announced that they would begin shipping samples of phase change modules in 128mb sizes. Look for these modules to appear in your electronics soon.

Holographic memory

Many people think holographic technology is a futuristic, far off technology, but it is available today from InPhase technologies. Of course, it is not widely available and is quite expensive. This will soon change because there are a lot of advantages to storing your data on holographic memory.

Figure 2:
A holographic memory device from

Holographic memory works by shining two coherent beams of light onto a light sensitive medium, a data beam and a reference beam. The three dimensional interference pattern created by these two beams of light is stored as a hologram. This interference pattern can be read by shining only the reference beam of light onto the interference pattern; the resulting beam resembles the original data beam.

This type of three dimensional memory means that we can store and access memory pages at a time. It also means that holographic memory devices will have an incredibly dense memory capacity.

Because of these advantages, I feel that it is inevitable that holographic memory will become a powerhouse in the tertiary storage arena. I am not convinced however, that holographic devices will ever reach the popularity of CDs and DVDs.

Magneto-resistive RAM

Magneto-resistive Random Access Memory, like holographic memory, is available today. In July 2006 Freescale semiconductor announced the world’s first mass produced MRAM product. Even today, one year later, there are few MRAM products available. Those that are available are expensive, are of low memory density, and are of interest to only a small niche market. Within a few years I believe that MRAM will become much more popular.

Like hard drives, MRAM stores data in magnetic storage media, making MRAM a non-volatile storage medium. This is an important feature of MRAM which will allow it to compete with DRAM and SRAM which lose their data when the power is no longer applied. Although not as fast as SRAM, MRAM chips provide faster read and write times than DRAM. MRAM does however, have a much higher memory density than SRAM. This will give engineers designing future CPUs another option as they will have to choose between a large, slightly slower cache, and a smaller, slightly faster cache. Personally, I think MRAM will win in the end, as it takes much longer to put data into the cache from your hard drive than to read data from the cache. So the less often you have to load the cache the better performance you will see.

Watch for MRAM to take on flash memory within a few short years. This is because MRAM memory, while much faster than flash, is also cheaper to make. I feel that the cost advantage of MRAM will be what propels the MRAM chips to take on the flash market, while the increased speed will be a nice side benefit to the customer.

In my next article I will set away from talking about the memory itself and discuss memory addressing techniques. These techniques are extremely important for both speed and error correction. I know you will enjoy it.

If you missed the other part in this article series please read

If you would like to be notified when Russell Hitchcock releases the next part in this article series please sign up to the Real Time Article Update newsletter.

Featured Links