2

If I take a 500GB hard drive and a 1TB hard disk otherwise identical, why is the 1TB drive not twice as fast as the 500GB drive?

If they are both spinning at the at the same rate, and the the 1TB drive has twice the data density (as it must), it should have a transfer rate of twice as much (although seek times will be comparable) or am I missing something.

2

2 Answers 2

3

Hard drives don't necessarily differ in the data density but also in the number of platters, for example. It's not uncommon for one capacity jump to just have double the number of platters instead of double the data density.

Furthermore, there are several factors affecting the data transfer rate you get from a hard disk. The most prominent one is probably the controller in the drive itself. If that one is limited ot maybe 80 MiB/s then more data density gains you nothing.

4
  • 1
    If there are twice as many platters, why cannot they be striped - they are all rotating at the exact same speed - would it not be trivial? eSATA2 supports transfer rates of upto 3GBits/s which would roughly translate to 300MBytes/s, if the 500GB drive runs at this 80MB/s, then the 1TB drive should be able to achieve up to 160MB/s.
    – Alister
    Oct 26, 2009 at 23:30
  • 2
    The drives could only max-out the 3GB/s SATA limit during a perfect (defragmented) sequential read, of a very large file. This only occurs very rarely, so the realistic speed is worse. And every time the drive needs to do any random read/write, or move position, there is a relatively huge delay for seeking.
    – sblair
    Oct 26, 2009 at 23:43
  • @Alister: I believe most drives will stripe their data across platters to maximize performance. Trying to do so in the most optimal manner for small or fragmented files doesn't hold up very well though.
    – Jimmy
    Oct 27, 2009 at 6:40
  • 1
    @Alister: One of the factors which triggered the hard-drive capacity explosion was the development of drive head controls which can actively follow synchronization marks which are written more "deeply" than the data track. Even if one one particular day the heads would line up to read tracks from every disks whenever they were lined up for one, effects like differential thermal expansion and aging would mean that wouldn't remain the case.
    – supercat
    Dec 17, 2013 at 23:20
3

If they are both spinning at the at the same rate, and the the 1TB drive has twice the data density (as it must), it should have a transfer rate of twice as much

Actually, the capacity is proportional to the area, whereas the transfer rate is proportional to the radius. Assuming that the number of tracks increases in proportional with the amount of data per track, the transfer rate should increase with the square root of the increase in capacity.

So increasing from 500GB to 1000GB you double the density per area of the drive, but the density per track is only increased by a factor of 1.414 (the square root of 2), so the transfer rate is only increased by a factor of 1.414.

I suspect the real factor may be a bit higher because it may be easier to increase the density per track than the number of tracks.

4
  • Good insights, however usually when a vendor releases a series of drives, the capacity is dictated by the number of platters. However this could be a factor when comparing drives from different manufacturers.
    – Alister
    Jan 7, 2014 at 5:36
  • 1
    @Alister - yes, the capacity is also dictated by the number of platters, but you specifically mentioned data DENSITY. The area increases with the number of platters, not the density. So if you change the capacity purely by changing the number of platters, you contradict your own "the 1TB drive has twice the data density (as it must)" claim - the density of information per unit area (the measure normally quoted for a technology) is precisely the same.
    – user31438
    Jan 7, 2014 at 7:55
  • 1
    @Alister - anyway, if you expect multiple platters to be read at once, I'm sure that's not the case. The drive heads have to follow little deviations in track position to access the data. These deviations vary over time (due to temperature etc) and from surface to surface. So I don't think it's practical to read multiple surfaces at the same time - doing so would need a much more complex and expensive mechanism - though I don't really know. You do gain, though, by needing fewer seeks as more data is in the same cylinder of tracks - a seek isn't the same as just tracking deviations in a track.
    – user31438
    Jan 7, 2014 at 8:10
  • 1
    @Alister - actually, I think early hard disks did read all surfaces at the same time, but that was a long time ago. Possibly so far back that they were still called Winchester disks. In a way, it's a bit like the parallel vs. serial cable fallacy - naively you'd think more wires means faster, but with wire lengths and temperatures varying slightly, capacitance and crosstalk between the wires etc, you get practical issues that increase costs and/or slow things down.
    – user31438
    Jan 7, 2014 at 8:22

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .