![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
whats with 7,200rpms?
- Thread starter Jason711
- Start date
unhappy_mage
[H]ard|DCer of the Month - October 2005
- Joined
- Jun 29, 2004
- Messages
- 11,455
Increasing the rotational speed makes for bigger motors, stronger platters, and more heat generated. 7200 right now is the balance between performance and problems.
SCSI drives go up to 15000 rpm. They work hard to get there, and those drives aren't cheap.
SCSI drives go up to 15000 rpm. They work hard to get there, and those drives aren't cheap.
As the sticky on the anatomy of storage performance suggests, increasing spindle speed reduces rotational latency and increases transfer rates, two very good effects. However, it also increases heat and noise, as well as either reducing reliability or increasing cost (or both 
), all of which are bad things. Furthermore, it is difficult to achieve high density on faster spindles.
Now, let's look at the positives of increasing spindle speed. We have the reduced rotational latency...however, the drive's firmware will likely cache the entire track as it goes by while the drive is waiting for the sector that was requested to pass under the read/write heads. Since locality in desktop performance dictates that other data on that track will be needed (SR notes that 20%+ of requests in their productivity DriveMarks fall on the same track as the previous request, as much as 50% depending on the density of the drive), it is highly unlikely that that cached track will be wasted, meaning that the impact of rotational latency on desktop storage performance is fairly small. On server drives, this is not the case, as requests are quite random and caching an entire track is of little value.
As for higher transfer rates, the makers so far have had better luck increasing the number of sectors per track to increase transfer rates. This is a win-win for the makers, as it does not involve compromise on heat or noise, allows them to make bigger drives, and allows them to make cheaper drives at existing capacity points.
As you can see, raising spindle speed is of little value to desktop drive makers...the gains from it are small or better accomplished with other methods, and the drawbacks are pretty substantial.
), all of which are bad things. Furthermore, it is difficult to achieve high density on faster spindles.Now, let's look at the positives of increasing spindle speed. We have the reduced rotational latency...however, the drive's firmware will likely cache the entire track as it goes by while the drive is waiting for the sector that was requested to pass under the read/write heads. Since locality in desktop performance dictates that other data on that track will be needed (SR notes that 20%+ of requests in their productivity DriveMarks fall on the same track as the previous request, as much as 50% depending on the density of the drive), it is highly unlikely that that cached track will be wasted, meaning that the impact of rotational latency on desktop storage performance is fairly small. On server drives, this is not the case, as requests are quite random and caching an entire track is of little value.
As for higher transfer rates, the makers so far have had better luck increasing the number of sectors per track to increase transfer rates. This is a win-win for the makers, as it does not involve compromise on heat or noise, allows them to make bigger drives, and allows them to make cheaper drives at existing capacity points.
As you can see, raising spindle speed is of little value to desktop drive makers...the gains from it are small or better accomplished with other methods, and the drawbacks are pretty substantial.