The 2.5" vs. 3.5" RAID Challenge

Conclusion: 3.5" Wins In Performance, Loses In Efficiency

It was pretty obvious that a four-drive RAID array containing 2.5" drives would not be able to compete with one using 3.5" drives. The performance difference caused by the slower rotational speed and smaller outside circumference of the 2.5" platters is too huge, and the access times make a difference in the I/O benchmarks too.

The 3.5" drives were considerably superior in the IOmeter runs. Looking at the database and web server tests in RAID 1+0, the WDs score between 100 and 200 I/O operations per second, and even more under workstation and file server conditions. The 2.5" Fujitsu drives lagged between 10 and 40 percent behind. When running a redundant RAID 5 array, the controller load becomes an issue, and as a result, the gap between 3.5" and 2.5" drives was less noticeable.

The 2.5" drives cannot compete with their bigger brothers in either access time or throughput benchmarks. In access time the numbers were 18 ms for the 2.5" array, versus 14 ms for the 3.5". For data transfer speeds, the result was 91 and 169 MB/s for maximum read speed of the 2.5" and 3.5" arrays, respectively, and 40 and 79 MB/s minimum read transfer performance.

Clearly, the 3.5" array trounced the 2.5" installation. From a performance point of view, 2.5" drives should only be deployed for server use if they outnumber a comparable 3.5" array. In this context, we need to underscore that it won't be possible to save costs by going this path. Also, you will have to cope with the smaller 2.5" drive capacities, since the largest models store only 100 GB, while 3.5" desktop drives hit 400 GB.

However, there are still some arguments that speak in favor of a 2.5" RAID array. These relate mainly to considerations such as storage density, energy consumption and noise levels. Up to ten 2.5" hard drives can be fit into a 1U rack mount device, while only four 3.5" can be accommodated into removable frames. As you can easily imagine, ten 2.5" drives should be able to outperform four 3.5" disks in terms of I/O performance.

Looking at the drive specifications reveals the 2.5" drives to be much more energy efficient. When idle, they actually beat their 3.5" brothers in power consumption by a factor of ten times! When comparing activity power consumption, the 3.5" drive still require four times more energy. If we decided to start checking performance per watt, the 3.5" drives would be smashed.

Last but not least, all 2.5" drives at common rotation speeds are extremely easy to handle since they do not require any special cooling. For this reason, we consider a 2.5" RAID array running 4-6 drives as a pretty applicable solution for high-end desktops and some workstations. There would be hardly any noise or cooling needs, and the capacity reached should be enough for most application scenarios; checking our benchmarks will give you a sense of what to expect performance-wise. The only drawback would be the total cost of such an array.

  • If you look closely you will see that this review compared 5400rpm 2.5" drives with 7200rpm 3.5" drives.

    Which makes it completely useless and flawed. I seriously can't believe Tom's did that. Maybe if there had been 5400rpm 3.5" drives included, some useful information could be gleaned from the tests.

    This entire article should be deleted just to save face, if not disk space. This article's very existence makes me embarrassed for Tom's Hardware.
  • Agree
    This is ridiculous
    the outside speed of the platter !!
    I cannot believe it .. you wait one turn never mind where your data are located and at 7200rpm your platter may be 1 meter in diameter it is not going to change anything it will stl be one rouind trip
    You will wait one turn ( 1/7200th of a second )
    It ain't go faster nor slower mechanically
    The heads are another story