I read somewhere that smaller (in terms of disk space) hard drives are faster than equivalent but bigger hard drives. How true is this? In other words, say I have two hard drives. Both are of the exact same brand and specs, but one is an 80GB while the other is 500GB. Which would be faster? Or does storage capacity not have any effect on speed at all?
A generalisation isn't useful, but mostly when talking similar models/same series I'd say the larger drive would be faster due to higher data density in some way (be it more platters and heads, or just denser platters).
The bigger model would likely be newer as well and could benefit from firmware and other production improvements.
This is even more true in the world of SSD where the larger capacity options are generally faster due to more parallelism. On the other hand, sustained throughput isn't always the important factor of an SSD compared to mechanical drives but rather the low-latency on small random access read/write - which will be the same in most scenarios regardless of the number of chips.
Size is but one of numerous considerations in determining the actual realized performance of a drive.
Rotational speed is one of the factors that determines the write rate. A 15k RPM drive would likely be faster than a 10K RPM drive of the same specs and size. (Assuming all things are equal which they are not in most cases)
The next thing to consider is the expediency in which the voice coil can move the read/write heads for a seek or continued file access. The latency introduced by the moving voice coil read/write head is perhaps the most significant source of delays in the read/write process.
The electronic controller board and what connectivity BUS it supports is also another significant determination of speed. A good example is the various versions of SCSI disks which supported higher and higher speeds with every revision to the scsi standard. SAS drives offer aditional performance over SCSI,IDE and SATA because of increased BUS bandwidth.
The number of platters is indeed also a factor but not the most critical performance consideration.
High data density is desirable, as it has a positive impact on data transfer performance: the more bits the drive can read concurrently, the faster it is. As a result, a new 3.5" 7,200 RPM hard drive always outperforms an older model. However, access time doesn't benefit from higher storage densities, as the head positioning cannot possibly be accelerated without putting substantial mechanical strain on the components.
In general I agree everyone else's answers. Given two hard drives with all else being equal the drive with greater data density will outperform the one with lower data density.
I can think of two scenarios where a larger drive capacity is a detriment to performance. In both cases it is not the drive that's the bottleneck but the file system.
- Formatting the drive
This is simply a matter of common sense. Since formating touches every byte on a drive a larger drive capacity will take longer to format. Since this is usually only done during an OS installation its not really a problem. In most cases its unnecessary to perform a full format operation anyway.
- Bumping into the limitations of the file system in-use
The best example of this was the point when drive capacities started to push the limits of the FAT file system. Without getting too technical FAT was designed for disk capacities a fraction of the size of its theoretical limits. FAT16's limit was around 2GB but as partitions approached this limit not only did they waste significant amounts of space but the overall performance of the file system degraded. FAT32 broke the 2GB barrier and performed better than FAT16 but ran into the same problem when drive capacities started approaching its theoretical limit (its around 2TB but this would be laughable to even attempt)
Each file system has different best and worst case running conditions. Modern file systems are designed to at least maintain performance if not improve it as drive capacity grows at the expense under-performing on small drives. A reasonable trade-off considering drive capacities are continuing to grow.
A hdd has a few platters. If both 80g and 500g has the same number of platters. That would mean the OS installed would fall in several platters on the 80G where as 1 or 2 platter on the 500G.
Each platter would have its own read and write. So on the 80G it is serviced by more heads than the 500G. So it is faster.
Does storage capacity affect hard drive performance?
Purely based on storage capacity: No.
However a bigger drive (as in, with more storage capacity) tend to be more modern and faster. So in practise: often yes.
From a mechanical point of view, and assuming rotating rust:
- A drive is faster is it spins faster (more RPM).
This means more platter spinning below the the r/w head in a period of time and less time waiting for a sector to arrive under the head.
- A more modern drive often has higher data density. Think of it as reading a book (at the same speed). When you write in smaller letters you can read/write more of them at the same distance.
- More modern drive tend to be bigger and have other technological advantages (e.g. faster head movement).
- Drive with more platters are often faster because a head switch is faster than waiting half a rotational time. And more platters more bigger capacity.
For spinning rust/traditional hard drives drives there's a few elements that would determine speed.
platter size - a 2.5 inch HDD would be faster than a 3.5 inch hdd, with all other things being the same. There's a smaller 'surface area' to seek, and quite often better read speeds. This is probably more true with enterprise drives since a typical 2.5 inch drive is laptop optimised.
cache size would affect the 'burstiness' of transfer speeds. Bigger ram caches are better, and some modern drives have a huge ssd cache. This might be a critical factor since a newer drive is likely to have a bigger cache. If we has a SSHD - a hybrid with a large nand cache, you'd certainly see a difference there.
rotational speed affects seek speed and throughput, faster is faster.
More platters means greater throughput (since data can be grabbed of each of many platters) but in some cases seek times are affected, since the heads move on the same actuator.
Interfaces are important. You're probably on SATA - though 80gb and 250gb are around the the time we switched from PATA. SAS drives have somewhat more efficient encoding, wider data pipelines, and in some cases can be faster. The latest generation of drives are hung directly off the PCI bus, and are ludicrously fast. I doubt that's in the scope here tho.
In this specific case though, chances are a more modern 250gb drive would have better performance charecteristics than a 80gb drive of the same rotational speed
WIth SSDs, cache is still a factor. However number of channels and the fact that SSDs are random access storage means that all other things the same a bigger SSD typically would be faster, and split reads between nand chips.
It depends, but often bigger drives tend to be faster.
There are a lot of factors that affect drive performance as people have pointed out:
Platter density - larger drives will typically have platters that have more information per track (single stripe around the disk), this will result in higher sequential read and write speed. Also, tracks near the outside of a disk will contain more data per revolution because, well, geometry. The data is the same density, but there's just more space out there and everything is rotating at the same rate.
rotational speed - drives that spin faster will have higher sequential read speeds, all else being equal, but they will also have a lower seek time, which is generally dominated by waiting for the platter to rotate under the read head, thus they will also tend to have higher random I/O
cache size - larger drives tend to have more cache, which tends to produce random read/write speeds, as the cache can handle bursts better, but can also give the drive more opportunity to re-order writes more efficiently as the platter rotates
interface/newer technology/etc - larger drives tend to be newer drives and as such are generally just made faster by improvements in scheduling algorithms, interface speeds, materials science, etc.
All of these factors will have some effect on the performance of a RAW disk, but we don't use our disks raw, we format them with filesystems and put stuff on them and this has a BIG effect. A larger disk will tend to have more freespace, and thus should be fragmented less, and fragmentation can kill drive speeds because it turns what would be fast sequential I/O into slow random I/O. For SSDs (at least those based on NAND flash) it's even worse because an SSD can only write each block in-order. In order to reclaim space from a block the SSD must move data from one block to another, excluding things that have been deleted. This means that on an SSD with very little free space the drive can spend more time moving stuff around to make space than it does actually handling writes. This also have the effect of burning more of the drive writes and reducing the lifespan of the drive.
I'd suggest running test on_the_same_drive rather than on two different models, no matter howsoever similar they may be.
HDDs are very slow media compared to processors and even a single link different in IO path could mean higher or lower performance.
What I'd suggest is use LVM on Linux to simulate a lower capacity drive on a HDD. Let's say around 10-15% of total HDD capacity. Run one cycle of workload you wish to test.
Then expand number of blocks in logical drive using LVM to full capacity. Flush linux block cache. And again run your workload cycle.
In my personal experience, performance(10% HDD) >> performance(100% HDD). I'm afraid i don't have performance data readily at hand. I'll run a couple of tests & post results here.
What everybody's answer here is focusing on is disk density. One factor that is missing from almost everybody's answer is 'length of rotation'. If you have lesser number of blocks, you traverse less length on disk surface, your rotation time goes down tremendously.
While disk density does give you more data per sq inch, in my experience length of rotation carries much higher share in disk latency. I've seen performance worsening by 10 times when going from 10% HDD capacity to 100% HDD capacity for same density. Even if higher capacity drives offer 10 times density (which i doubt), the disk would still have to rotate more to cover that many more blocks.
So the effect of higher density is somewhat undone by higher block count.
For practical example, look at SAS drives. What do you think is commonly available capacity for SAS HDDs? It is 300GB. In a world of 1TB capacity SATA drives, enterprises are dumb enough to sell drives with 30% capacity at super premium prices.