It is not very often that one of those “it will be out in 3-5 years” storage technologies actually reaches the market, so I was very interested in seeing that Freescale Semiconductor is now shipping MRAM chips. The chips are a very modest 4mbit capacity, but as they are being manufactured on a very large 180 nm process, capacity will go up as more modern fabs are devoted to manufacturing. So, what is MRAM and why the hell am I excited about it?
MRAM is a persistent memory storage technology with no moving parts. In this way it is a bit like flash memory, which is most commonly known to people as the technology behind those little USB thumb drives that you can carry on your keychain, and a bit like your hard drive, which is most commonly known to people when it crashes and eats all your data. Unlike flash, MRAM stores data magnetically, like a hard drive. Unlike a hard drive, it has no moving parts, so it is much more reliable and uses up much less power.
One of the chief problems with flash technology is that it has always lagged behind hard disk technology in terms of storage capacity. The highest capacity hard drives these days are 750 GB. The highest capacity USB flash drive one can readily purchase in a storage is 4 GB. Another problem with flash is that all flash memory can only be written to a certain number of times before a particular memory block goes bad. This can be dealt with somewhat by keeping a lot of spare blocks free and to spread the writes equally over the entire memory using “wear levelling” techniques. Still, the fundamental problem remains that if you frequently write/update data (for example, an OS’s paging file) on a flash based drive, it will die a quick death. Some specialized embedded Linux distros have been developed to minimize writes by mounting themselves read only. One early such distro is Pebble Linux, which sadly is very dated at this point. Rob Flickenger took the Pebble, adapted it for his company’s Linux based WiFi routers, and then re-thought the concept and came up with Pyramid Linux, which is based on the more modern Ubuntu Linux. MRAM could make it possible in the very near future to create reliable embedded devices with full r/w abilities.
The other major benefit of MRAM is that being a chip based technology, it should have very short seek times and be able to handle random i/o patterns, which would cause a hard drive’s r/w heads to thrash back and forth like in a dog watching Wimbeldon. Also, as a chip based technology, it should be far more reliable, due to the lack of moving parts and be virtually impossible to physically damage. In my line of work, I am often meet new clients due to failed hard drives. While I am happy to be able to eat and pay for gadgets as a result, I would frankly rather be doing things other than data recovery and implementing backup solutions. I probably see 2-3 hard drive related clients in an average week and hard drives by far remain the most common physical cause of computer hardware problems. With hard drive technology finding its way into an increasing number of electronic devices like PDAs, MP3 players, cell phones, digital video and still cameras, gps navigation systems, etc. one can only expect hard drive related problems to increase, particularly as the hard drive technology used for micro-drives is quite a bit less robust than their desktop cousins.
MRAM has the potential to offer all the benefits of magnetic hard disk technologies with all the benefits of flash based technology. There is also speculation that MRAM may be fast enough to replace traditional random access memory in a variety of applications, making it the near universal memory techology. MRAM is on the market now. Whether it has a future is anyone’s guess. The major sticking points, imo, will be:
– whether the market will pounce due to the benefits, or will stay with the devil it knows.
– Will fabricators divert capacity from manufacturing know profitable but low margin products and risk trying to manufacture higher capacity modules? MRAM is not very interesting in it current .5 MB per chip capacity. In theory, magnetic technologies can scale to far smaller sizes/greater density that flash based products. This has been demonstrated by the ever increasing capacity of hard disks relative to flash devices. We will have to see if MRAM follow a similar pattern.
– The current cost of the MRAM chips is around $25/chip. I have a hard time imagining that this part is very attractive given its small capacity and high price.