There is a theoretical limit to how fast information can be read and written to and from a disk, scientifiic journal Nature has reported. And it's 1,000 times what we are currently capable of.

Puting aside the issue of how small the data bits (magnetised areas), how fast they can be read or created? The faster a disk spins the less time each data bit is underneath the head. How fast is too fast where there is not enough time to read or write a data bit?

Nature says it is aproximately 435Gbit/s - that's 435,000 million bits a second. Currently we write and read bits at around 435Mbit/s. We have disks spinning at 15,000rpm in high-end arrays. According to Nature, they could spin at 15,000,000rpm with current areal density but any faster than that means the data read or write will fail. Of course, disks are capable of spinning faster but areal density will increase so that the time any data bit area spends under the head will decrease.

The research has found that if the signal read or write times is 2.3 picoseconds or less then then operation will fail. It takes a beam of light 10 picoseconds to travel across a blade of grass. Since the light is travelling at 186,000 miles a second, 10 picoseconds is not that long. As for 2.3 picoseconds an eyeblink will seem to take years in picosecond time. A picosecond is one trillionth of a second.

Researchers fired magnetised particles at the magnetic material used for disk platter coatings and recorded the time for the magnetic field to switch in a reliable way. As the time period shorthens switching stops occurring reliably and starts becoming random, meaning the data bits can no longer be reliably written or read.

How long before this becomes a factor in disk production? Let's suppose that disk spin speed stays the same then areal density would have to increase a thousandfold for this new limit to be reached. IBM researchers believe areal density is increasing at the rate of 49 per cent a year. Twelve years from now, in 2016, we will hit this limit.

Microsoft is believed to be worried that its next-but-one version of Windows may have to be redesigned to take this research into account.