WD's New Raptor Drive Is a Bird of Prey!

Hard Drive Performance Basics

While the recording technology defines the storage capacity, spindle speed is the most important parameter that influences performance. Not only will high RPMs provide the best data transfer rates possible (expressed in Megabytes per second), but they also shorten average access time (expressed in milliseconds). Here we have to differentiate between the seek time, which manufacturers usually refer to, and to the average access time, which includes locating data, positioning the heads above the appropriate track and waiting for up to a full platter rotation before the required data is accessible by the heads. This idle time is referred to as rotational latency. Obviously, faster spindle speeds will not decrease the seek time, but they very well reduce the access time, as rotational latency will be much lower.

Seek time and access time also depend on a drive’s performance when relocating heads from one platter surface position to the next. Arm acceleration performance and break performance come into the equation; both have physical limitations and this movement contributes to a drive’s noise level, as do spindle motor and friction noise. Although hard drives are sealed, the inside is sealed by the use of small filters. Since an air cushion is needed to have the heads flow above the surface (aerodynamics comes into the equation as well) they do not operate in a vacuum. Performance hard drives for the enterprise segment aren’t optimized from an acoustic standpoint, so they can be easily heard; especially when there is a lot of arm movement going on.

All hard drives come with cache memory, which is used to store read or write data based on complex algorithms. Most drives have 8 or 16 MB cache; some models go up to 32 MB. The cache memory capacity really isn’t a big deal as long as you have some. In a very simple example, a hard drive may continue reading more data into the cache memory even after the requested data has been read, because there is a chance that the following sectors may be requested soon.

All modern hard drives utilize either the Serial ATA (SATA) or the Serial Attached SCSI (SAS) interface. Serial ATA is based on the good old parallel UltraATA protocol, while SAS builds up on the Small Computer System Interface. Both SAS and SATA use the same connectors and offer the same bandwidth (150 or 300 MB/s), but SAS is fully SATA compatible, as it can tunnel the SATA protocol. They both also support similar feature sets such as Native Command Queuing (NCQ), which the drive utilizes to analyze and reorder incoming or pending commands in an effort to process them with the highest efficiency possible. SAS still is more sophisticated, as it allows drives to run on two physical connections for the sake of performance or redundancy.

Related article: Understanding Hard Drive Performance

Drive Energy Efficiency Basics

Physical movement requires energy; hence it doesn’t come as a surprise that high RPM hard drives typically require more power than slower models. However, this also depends on the drive generation, on the platter diameter and on the platter count. A 2.5" 15,000 RPM hard drive with only two platters may very well consume less power than a larger 3.5" 10,000 RPM drive, which utilizes four or five platters. If you compare drives within similar generations (e.g. 2008 hard drive models), it is safe to assume that a high capacity 3.5" terabyte hard drive will always consume more energy than a smaller-capacity 3.5" single platter drive at 250 or 320 GB, and that higher RPMs will require more energy as well.

While PCs are increasingly being evaluated from a performance per Watt standpoint, the hard drive industry must also consider capacity per Watt. 3.5" hard drives turned out to be ideal in this benchmark, as they currently provide up to 1 TB of storage per drive (Serial ATA models) at power consumption of 3-15 W. Smaller capacity, single-platter versions may provide lower power consumption, but at the same time they’re inferior when it comes to capacity per Watt.

Lastly, the drive’s interface contributes to power consumption. We found that parallel interfaces such as UltraATA or SCSI were generally more energy efficient, SATA shows differences whether it’s running on 1.5 Gb/s or 3.0 Gb/s. The faster link speed takes its toll by adding some 0.3-0.5 W to a drive’s total power consumption. While the old Raptor-X at 150 GB wasn’t bottlenecked using SATA/150, the VelociRaptor gets close to the limits of this interface by reaching almost 125 MB/s. Since there is a certain protocol overhead, which becomes obvious when looking at SATA/150 drives that typically max out at 126 MB/s, introducing SATA/300 was an absolute necessity.

Achim Roos