Originally posted by: smitty6
if the bit density is so much greater, could you not then reduce the disk rpm's by a large amount and still have excellent read/write performance? This would reduce noise, power and failures.
Not really. If you break down read write performance, it becomes highly dependent on teh size of file you are reading. There are two primary components to read/write performance:
1) The time it takes the head to get to where you need to be on the platter
2) The time it takes to read the information
On a typical 7200 RPM hard drive, you can read/write at say 30-40 GB/sec. on average. Meidan file size is less than a meg. Let's say the file is sequential and is exactly 1 meg.
At 30 GB/sec it takes 1/30,000 seconds to read that 1 meg, which is 0.033 ms
If you're able to double read speed to 60GB/sec you are reading that section at 0.016 ms.
BUT the dominant factor in this case is the time it takes to get to the data. Even if the armature doesn't have to moveyou still have to wait for the data to pass under the head. One revolution on a 7200RPM drive is 8.33 ms, and on average latency is half that (4.17 ms).
So in this example doubling the sequential transfer rate only decreases the file access time from on average 4.203 ms to 4.186 ms, only a 0.4% improvement overall. Typically there is some actuator travel time and settling time added to that also, which is also on the order of single digit ms readings, so generally total time to get the desired information under the head is in the low teens of milliseconds. But you can see that latency is a large component of the time it takes to get the information desired under the head.
the average latency with a 5400 RPM goes from 4.17ms to 5.56 ms. ~1.3ms, which is on the order of 10% or so slowdown on a typical single file access.
On larger transfers, bit density has a significantly larger impact, and if transferring several gigs, it dominates seek time much as seek time dominates in this example. Because latency is ONLY dependent on RPM, and it is an unavoidable component of the seek time, RPMs won't be decreasing anytime soon.
Also, is the power required by the magnetic head (that much) greater when aligning the bits in the perp fashion?
I don't know for sure. Heads aren't my thing, media is.
The geometry change demands a significant increase in gap width between the poles of the head, however the magnetic field that needs to be generated within that gap is significantly lower.
Kinda difficult to explain without pictures, but longitudinal would be like holding a magnet over the surface of something to magnetize it. Magnetic force decreases rapidly as you get further away from the magnet (square of the distance). The field necessary for switching needs to exist at least as deep as the the media thickness plus whatever the fly height is. So the max field that needs to be generated directly within the gap is many times higher than the field necessary for switching
Perpendicular essentially puts one pole of the magnet from the head UNDER the recording layer, so the field you're using is like using magnetic field that actually runs through the middle of the magnet itself, and not the field lines that emanate from the magnet. So even though the gap is very wide, the max field generated is only slightly higher than the field necessary for switching.
If I had to guess, I would guess that perpendicular recording at the same density as longitudinal probably uses a little less energy, but as I said, I'm not nearly as up to speed on heads as I am on media, and that is essentially just a guess.
Edit:
I should add that power the head uses is pretty much insignificant in the overall power consumption of the drive, if that's why you asked. Mechanical components like the positioning mechanism(s) and spindle motor are going to use the bulk of the power.