Are there big differences in speed between different hard drives?

Posted on


When buying a (consumer-level) hard disk, I normally only pay attention to storage capacity. Is it worth to compare drive speeds also, and if yes, how can this be done?


If performance is your main concern, it’s also worth considering whether solid state drives (SSDs) suit your needs. They are (usually) significantly faster than (premium) hard disk drives. But SSDs also cost more per GB, so are often used just for the boot/OS/applications drive; HDDs must still be used for bulk data storage.

So, for me, the relatively minor differences in HDD performance are almost irrelevant because SSDs can fundamentally perform better.

Yes, it’s worth it to compare drive speeds if getting your data more quickly holds any advantage for you. For the home user, it’s typically a matter of convenience. For a professional server, faster drives mean faster performance under disk I/O load.

Assuming you’re looking at drives with the same interface, the most common speed measure, and usually the most significant, is the RPM of the platters. Common values include 4200, 5400, 7200, and 10,000 RPM. High end SCSI drives and the like can go up to 15,000 RPM.

You may see values like seek time, access time, and the like. This will usually be similar for drives of similar rotational speed.

Also, you’ll see disk cache sizes, typically a larger cache means it can take better advantage of temporal locality.

Tom’s Hardware has some charts of hard drives benchmarks, using h2benchw. Reads are what you spend the bulk of your time doing, but there are write charts there, too.

Read Access Time

This tells you how responsive the drive is, how quickly it responds to a sudden small request. I believe this is what best gives the impression of hard drive speed. It’s analogous to the “first page out” for printers. Less is better, so the best performer is at the bottom (those boneheads).

Average Read Throughput

This chart will tell you how fast it should make medium to large copies. Using the printer analogy, this would be “pages per minute”. More is better, so the best performer is on top.

Storage Review’s Performance Database has more benchmarks. They’ve slowed down the reviewing pace in recent years, but it can give you an idea as to how a family of hard drives is expected to perform. Ignore the IOMeter benchmarks; they only apply to the heavily parallel access patterns found in servers.

Manufacturer supplied stats are a starting point, but you really have to consult benchmarks to have an idea of real world performance. On the Storage Review benchmarks, notice how the 7200 RPM Hitachi beats a 15K RPM drive on the Office Drivemark test. Just citing spindle speeds wouldn’t tell the whole story here.

Is it worth to compare drive speeds [sic]

For a system drive where the OS resides, yes. It’s one of the slowest data-access components especially considering its high utilization.

For a storage drive/secondary drive, not really. Benchmarks are only useful here to find out if a drive is particularly slow. For this, just look for reliability, good customer feedback, low price and a long warranty period.

Hard disk performance is hard to condense into one number. The primary factors which will affect performance are, in approximate order of importance:

  1. Spindle rotation rate (for example, 5400, 7200 RPM)
  2. Platter linear density (number of bits per inch)
  3. Cache size

Random I/O performance (reads or writes to “random” locations on the disk) is almost completely dependent upon the rotation rate. Sequential I/O performance is dependent upon the product of the rotation rate and the bit density.

You can get 10K/15K RPM drives, but they are smaller (less density!) than 7200 RPM drives. That’s great for a database (largely random access I/O), but they are a total waste for storing videos.

Cache size mostly helps with writes (if you enable disk write caching), because the drive can schedule the writes more efficiently, reducing the amount of seek time per write. It doesn’t help that much with reads because your OS already does read caching with a lot more memory.

Leave a Reply

Your email address will not be published. Required fields are marked *