Hi jvbig,
There are three primary parameters that determine how effectively a disk operates. Throughput, seek time, and "rotational latency" (nomenclature ?). How the applications require data from the drive determines which is most important. To put things simply (ie at the level I think I understand them <g>. Think of throughput as the difference between DMA-33, DMA-66, and DMA-100 drives. If everything is optimal the 100 is 50% faster at moving your data between the disk and memory than the 66, and 3 times as fast as the 33. But first it has to get to that data. The data resides in groups (clusters) in concentric rings (tracks) on one or more surfaces (heads). The heads move together, so it does not really matter which surface the data is on. However, it does take the head a discrete amount of time to move from one track to another (seek time - under 9 is good), and once it gets to that track it takes another discrete amount of time for the desired cluster to rotate under the head (latency - a direct function of the 5400 vs 7200 rpm).
Now it gets tricky. A 20 gig drive with 1 surface should be exactly the same as a 40 gig drive with 2 surfaces, all else being equal. But a 40 gig drive with one surface should be faster than a 40 gig drive with 2 surfaces all; else being equal. If the tracks are closer together then the seek time should be less and if there are more clusters per cylinder, the throughput should increase given the same rotational velocity (although, the latency for accessing the first desired cluster will be the same). So theoretically, one could have a 5400 RPM drive with 2 surfaces being faster than a 7200 RPM drive with 4 surfaces. Of course, if the throughput on the two drives was the bottleneck then both would be similar in response.
Then there are the things that have nothing to do with the drive. For instance, a badly fragmented drive is constantly doing seeks all over the place and a tuned 5400 will outperform the fragmented 7200. If one primarily loads executables that then do most or all of their work in RAM (ie Excel) then it really doesn't matter if it takes 1/2 a second to load on a "fast" drive vs 1 1/2 seconds to load on a "slow" drive. On the other hand a random access 10 GIG database will definitely notice the difference. Then there are the caching algorithms of the OS. On a smaller Database with often used sections that can be cached in RAM, the disk won't matter much.
And then there's environmental factors. A 7200 RPM drive may generate more heat and/or be louder than a 5400 RPM drive. Three years ago, every time I got a new computer, I would take out the 16X/24X/32X CDROM, put in a 4X/6X/8X CDROM, and sell the faster one because they made so much damn noise when they were accessed. (That does not seem to be the case now days; although I'm not sure since I've only used DVD the last 18 months. (I also always get an R/W CDROM drive, but since I only use them for recording - at 4X - I haven't noticed). Speaking of which, the DVD's work just fine with cdroms, at least the "pressed" ones. I find they are sometimes recalcitrant with CDR and CDRW disks although that varys by DVD brand as well as CDR/CDRW blank brand.
Anyway, this is what you get for asking <g>. I'm sure any myths/misconceptions I've perpetuated will be quickly cleared up by this good and knowledgable thread.
Good luck,
Steve |
