I know, I used to have a bet with my MS-bashing friends to find me an article where he had even made anything
close to that statement. Not to prove them wrong mind you, but I had heard that same claim so often that
I really wanted to find out where it might have originated if true.
I don't know where that quote really originated from. Someone in the industry may have actually said it, but it definitely wasn't Bill Gates. The internet made the quote so famous and oft repeated that people actually think it's a real quote from him.
One issue though, as you already pointed out, drives can be built with smaller platters to create an additional reduction in latency.
Drive latency as defined by drive manufacturers is the amount of time it take the platter to rotate to the proper sector on the track it is searching. Thus, reducing platter size will not reduce latency, only seek time, the amount of time it take the read/write head to position itself over the proper track on the platter. The only thing that will decrease latency is a faster spindle speed.
You already mentioned the Savvio drives, but IIRC the Cheetah line has been taking avantage of smaller platters for several generations.
All 10k drives use roughly 3" platters, while all 15k drives use around 2.6" platters. The Savvio is the first non-portable drive to use the 2.5" form factor which uses platters smaller than typical 10k drives, reducing average seek time. Due to power and space savings eventually all hard drives will move to this form factor.
Your calculations above are all based on keeping the drive volume constant, yes?
Not sure what you mean by drive volume, but absolutely nothing affects hard drive latency but spindle speed. A 10k RPM drive with 4 mile wide platters will have the exact latency as a 10k RPM drive with 1" platters. The average latency spec quoted by hard drive makers is simply maximum latency divided by 2. It's a simple mathematical formula [(1 / (spindle speed / 60)) * 0.5 * 1000] which can be simplified all the way to:
30,000 / spindle speed
Which gives you the average latency in milliseconds. As you can see, the only variable is spindle speed, and you will always have the same average latency for the same spindle speed (or whatever is spinning at that rate).
