FPSLabs : Shoes Explains Hard Drives

Thomas “shoes” Gribble takes the time to show us how hard drives work and what we should look for when selecting one for our systems.

One of the most crucial parts of any computer is often overlooked by gaming enthusiasts and the sort when choosing parts for their computer. I could not possibly count the number of times that I have seen people asking for opinions of their soon-to-be setups with cutting edge processors, high speed RAM, and video cards that cost more than my whole computer, while it seems like they buy a hard drive with what is left over of their bank account. Your hard drive, like the three other aforementioned components, plays a very large role in the performance of your computer.

In order to understand what makes a good hard drive, it would make sense that we need to understand how one actually works. To do this, I will explain each part of the hard drive and how they interact with each other.

Spindle: The spindle of a hard drive is essentially the axle on which the platters spin. Because every platter is fixed to the spindle, all platters rotate at the same, constant speed.

Platter: Every hard drive has at least one platter, most of them have more than one. A platter is much like the thin, glossy tape that you can find inside of a floppy disk. The difference is is that the platter is rigid and spins at much greater speeds than a floppy disk. Platters are manufactured to incredibly precise tolerances in order to avoid the possibility of them breaking themselves apart due to vibration under high RPMs. They have a very crisp mirror-like finish which is optimal for the read/write heads to retrieve and store information on. Each platter is double sided as well, and as such, there is a read/write head positioned for each side of each platter.

Read/Write heads: These electromagnetic tips are positioned at the end of boom arms that are also connected to their own spindle, and because of this, they all move in unison rather than independently. Read/Write heads do not touch the platters, instead they float on a cushion of air that is created on the surface of the platter by its rotation. As the surface of the platter passes underneath, the heads read or write magnetic patterns off or on to the disk. The patterns they create can also be removed and rewritten at any time.

Cache Buffer: The amount of cache a hard drive has can have a vast affect on its overall performance. Cache is basically just a dedicated bit of Random Access Memory that allows you to retrieve and write commonly accessed files at electronic speeds. For a quick and easy demonstration of what cache does, we can utilize our internet browsers. All internet browsers have a cache of varying size which you can control. First, reload this page. Depending on your internet connection and the current load on GotFrag, it probably loaded up pretty fast. Then, go to your internet options and find the button for clearing the cache, and do so. Now, reload the page again. You will notice that it takes significantly longer (again, depending on your internet connection). This is because the cache does not have this information temporarily stored so that it can access it at electronic speeds, and instead has to download all the content (text/images/etc) through the internet connection again, which is a slower process. The same is true in hard drives. When you commonly access a file or application, it will be able to load MUCH faster than it would if you were to have to mechanically retrieve it via the read/write heads. Cache also serves as the “buffer” between the motherboard and the platters. All information, whether on its way in or on its way out, goes through the cache buffer.

Controller: The final part of the hard drive is the one which tells the information where to go. When an application tells the RAM on your computer to send the current file information to the hard drive, it sends with it a signal that determines where on the hard drive platters the information should be stored. Upon receiving this signal, the controller directs the read/write heads over the specific part of the platter and the information, which is temporarily being held in the cache buffer, is written to the disk.

Another important concept to grasp is how the information is organized on the hard drive. There are “rings” on the platter that are referred to as tracks. These tracks are divided in to many many small sectors. Information is stored into the sectors, which hold 512bytes of information each. Usually we use the term clusters, which is a group of sectors, because sectors are so small that it is unlikely that data will be stored in just one sector. In addition to this, it is important to note that due to the spindles being circular, the circumference of the tracks increases as you move outward from the spindle. The tracks therefore can hold more sectors as they move outward. This also means that the amount of sectors being used per track on the outside of the disk is far less than it is toward the spindle, which means that there is less information per square unit on the outside of the disk than there is toward the spindle. This is referred to as “areal density”. The more areal density the better; more information is available within a smaller area, the heads have to move less to find the information, meaning less seek time, meaning better performance.

So how do the heads know where to get data when you ask for it? Well, in modern drives, there is usually one side of one platter that has somewhat of a table of contents. It has all the track information for all of the other platters, and the controller knows to tell the read/write heads where to position themselves to get the desired data.

Interfaces: There are really three main kinds of interfaces for hard drives: IDE (PATA/ATA), SATA, and SCSI. IDE and SATA are targeted to mainstream clients while SCSI is mainly an enterprise interface for servers and the like. Pretty much every single motherboard comes with support for the IDE interface, and understandably it is far and away the most widely used interface today. The problem with IDE, however, is that it has been around for a long time. Only one IDE drive can send data through the interface (motherboards usually have 2 IDE channels that make up the interface) at one time. In contrast, multiple SATA drives can send information through their interface at one time, allowing for much higher efficiency. More and more motherboards come with native support for SATA drives. SATA is an attempt to remove the disadvantages of PATA without complicating the whole data transfer process. SCSI drives on the other hand, have very few motherboards supporting their interface, which means that you will likely have to get a controller card to set up a SCSI hard drive. However, a fourth interface called Fibre Channel, which you can think of as an advanced version of SCSI, is definitely the most technologically superior interface, and is the undisputed king in the multi-tasking and heavy development fields. Fibre channel’s cost and complexity is such that it is pretty much exclusively used in enterprise servers and high-end workstations.

Seek Time: As hinted at before, seek time is simply the amount of time it takes for a read/write head to move over the desired track on the platter.

Rotational Latency: Once the read/write heads are over the desired track, it is up to the disk to spin to the right sectors on the platter. Rotational Latency is simply the time it takes for this to happen.

Spindle Speed: Spindle Speed has a direct impact on the rotational latency of the drive, because of course it is how fast the information is flying around on the platters.

The Cache/RPM Debate: Everyone seems to be arguing over which makes the biggest difference. This is likely the result of the competition between two drives from different manufacturers that are aimed at the enthusiast market: the Western Digital Raptor, and the Maxtor Diamondmax 10. The reality of the situation is that the 16mb cache afforded by the Maxtor drive is only useful when doing repetitive or recently done tasks. However, when such tasks are being done, which actually is quite often, the speed is superior to the Raptor (This type of data transfer is called burst transfer). Another advantage the Maxtor has is Native Command Queuing (NCQ). This allows the drive controller to reorder the read/write commands sent to it in a way that the ones stored on tracks closest to the read/write head are accessed first, instead of going across the whole radius of the platter and back again, which helps in multitasking but can hinder performance in pure load times. The Raptor on the other hand, makes up for in brute force what it loses in cache. For random read/writes there is no match for a 10,000 RPM drive in the consumer market. For those that are set on having the absolute highest performance storage using the SATA interface, get one of each, then you have the best of both worlds.

Other things to take into account when buying a hard drive are noise output, heat tolerances, warranties, and cost per gigabyte. So get out there, interpret the numbers, and decide which hard drive best meets your needs.

Popularity: 5% [?]

Support FPSLabs! Share and Enjoy: