Gus said:
There's a 40% performance gap between the 10,000 rpm Cheetahs and the gaggle of
7,200 rpm disk drives, and there's a 33% performance gap between the 7200 rpm
Medalist Desk Pro and the rest of the field. Performance defined in terms of average
access times and sustainted transfer rates.
I'm confused ... are you equating spin speed with average access times and sustained transfer rates?
Average access time is average seek time plus average latency. Average seek times of modern drives are typically 7 to 14 mS, depending on the drive. Spin speed has no effect on it. Average latency is one-half the revolution time. Revolution times are 6mS, 8.3mS, and 11.1mS for 10,000, 7200, and 5400 rpm respectively, so average latency is 3mS, 4.17mS, and 5.55mS. So best case, a pair of 7mS seek time drives, one at 10,000 rpm and one at 7200 rpm, the average access time for the 10,000 rpm drive is 10.5% faster (1-(10mS/11.17mS)), not 40% as you claim. The same comparison for 7200 versus 5400 rpm gives best case 11% ((1-(12.55/11.17)), not 33% as you claim. Also, these percentages drop significantly with the more typical 9-11 mS seek times.
Now, with regard to sustained transfer rates, there are 3 components of transfer rate: spin speed, sectors per track, and skew (skew is how many sectors the next track is offset, so that after a sequential track switch the first sector of the next track is ready to be transferred). So spin speed is a direct factor of transfer rate, right? Wrong! In reality, the speed of the silicon often limits you. Think of it this way: in the 10,000 rpm drive you only have 6mS to write the data, in the 7200 rpm drive you've got 8.3 mS to write the same amount of data, so you've got to write 38% faster to record the same amount of information on the 10,000 rpm drive. If your silicon isn't up to the job, you have to settle for fewer sectors per track, and the transfer rate advantage goes away, in addition to lowering capacity. Do drive makers really design drives with the spin speed outrunning the silicon? Not always, but I've seen (and done) it many times. It's a marketing thing to be able to claim higher spin speed. Too many people just assume spin speed equals performance, and in reality, it ain't necessarily so. |
