Size as important as rpm for HD speeds!

[Diatribe]

For all those wondering how fast one rpm iteration is over the other, many people forget to consider HD size.
HD size is as important as rpm for the speed of the drive, or better:

The important aspect is the amount of data you have on the drive in relation to its capacity!

A 4200rpm 200GB drive at 40% capacity is often faster than a 7200rpm 100GB drive at 80% capacity. (Both are filled with 80GB for the mathematical challenged )


Obviously if you fill up the 4200rpm drive up to 80% it is going to be slower than the 7200rpm drive but it all depends. A 4200rpm drive will probably run at the same speed being filled up 50% as a 7200rpm drive being filled up 80% giving you the same speed at 25% more usable storage.

So next time you wonder which one to buy consider the size too.

And a link if you need the hard evidence :http://www.barefeats.com/mbcd7.html

 

[robbieduncan]

Obviously if you fill up the 4200rpm drive up to 80% it is going to be slower than the 7200rpm drive but it all depends. A 4200rpm drive will probably run at the same speed being filled up 50% as a 7200rpm drive being filled up 80% giving you the same speed at 25% more usable storage.

Not necessarily. This will depend on a number of factors including data fragmentation and the number of platters (which is another way of saying the density of the data on the drive).

Fragmented data is bad m-kay. This goes some way to explaining why as the drive fills it gets "slower". Whilst OSX is good a defragmenting on the fly not all files get defragmented this way. And as the drive fills up there is a greater change that 2 files that are requested sequentially are further away on the disk in linear, not rotational terms which is a real speed killer.

Platter density is also important. If we assume that the platter density is constant then a drive of twice the capacity requires two platters. This sounds great as we now have 4 read/write heads instead of 2 but in reality it's not always that great as a file can end up spread linearly over all 4 platter resulting in 4 seeks instead of 2 max on the single platter drive (this almost never happens though). But if platter density goes up then the data gets crunched closer together. This is almost always good as this will result in fewer linear head moves, and smaller moves at that which should be faster.

So without knowing about the fragmentation status and platter density in the drives tested we cannot draw absolute conclusions.

 

[Diatribe]

So without knowing about the fragmentation status and platter density in the drives tested we cannot draw absolute conclusions.

Well fragmentation depends on the file sizes you're dealing with. Everything above 20MB will cause fragmentation. Keeping it around a certain fill level will reduce fragmentation.

Platter density will almost always be higher in the higher capacity drives, especially with perpendicular recording.

And besides, it wasn't meant to be absolute, just something to consider

 

[CanadaRAM]

The reason why a drive at 80% full is slower than the same drive at 40% full, is that the data is written from the outside in. The first tracks written are on the outside tracks, where Pi*D will tell you that the length of track (that is, the perimeter) is much longer than the inside tracks. Since the speed of the drive is constant (1/5400 sec per rotation for example) in the same amoount of time, much more data passes under the heads on an outside track than an inside track during one revolution.

This is also why a 3.5" hard drive will smoke a 2.5" hard drive of the same specifications -- the larger diameter drive means that most of the data is on longer tracks than the 2.5" drive's fastest outside track.

It means, as noted above, that on two 2.5" drives, one with 120 Gb and one with 200 Gb, with one platter, at the same RPM, the 200 Gb drive will shovel more data on and off because there are 40% more bits passing under the heads. If one drive is 5400 RPM and one is 4200 RPM, you lose 23% to rotation speed. So theoretically the 200 Gb 4200 RPM drive is faster. But that's only during sustained transfers of contiguous data. MOST of a hard drive's active time is spent seeking the point where the data starts, (and in the case of fragmented files, seeking each successive cluster). That's where the 4200 speed exacts a penalty, in higher latency, as it takes 23% longer for the data being sought to come around under the head, so when the job it to read and write many small bits of data, the 4200 rpm is going to be quite a bit slower.

What this means, too, is if you want to ensure the highest performance from one particular area of your hard drive, then partition it. The First partition is the fastest. Use it for your System and Scratch/VM files. The last partition created is the slowest, use it for backup and archival files you seldom access.

(BTW, virtually all laptop drives are single platter because of the 9.5 mm height restriction. Gains in capacity have to come from increased areal density, not stacking platters.)

Thanks
Trevor
CanadaRAM.com

 

[robbieduncan]

I have to agree with both posts above. I was simply trying to ensure that people were aware of all the facts available and did not simply take the results as read without considering the other possible factors.

These forums are great as the depth of technical knowledge here is amazing

 

[Diatribe]

I have to agree with both posts above. I was simply trying to ensure that people were aware of all the facts available and did not simply take the results as read without considering the other possible factors.

These forums are great as the depth of technical knowledge here is amazing

Yeah, one of the reasons I love this place.