Does hard drive buffer size matter?

Question

These HDDs will be used mostly for backups and data-storage (pictures, movies) which aren't often read. The price difference is 1.5 Euros in the case of these two:

• HDD Western Digital Caviar Green Power
  500GB, 7200rpm, 32MB, SATA2
• HDD Western Digital Caviar SE16
  500GB, 7200rpm, 16MB, SATA2

The only difference is the buffer size (16MB vs. 32MB). Is the larger buffer size worth the extra money?

Solution 1

It really worths the extra money. Why :

when you try to access a content from hard disk , it takes time to find the content you need. buffer is made from good stuff with high speed access. So when you find a content, it will save them all to buffer through an algorithm .

Conclusion: It makes your HDD disk works faster :) especially with large contents :)

nXqd

Solution 2

http://www.pcguide.com/ref/hdd/perf/perf/spec/otherCache-c.html

As memory prices have fallen into the "dirt cheap" range, drive manufacturers have realized that they can increase the size of their buffers at very little cost. Certainly nothing is lost in doing this; extra cache won't hurt performance; but neither does it greatly improve it. As a result, if interface transfer rate is the "reigning champion" of overrated performance specifications, then cache size is probably the "prime contender". Some people seem to think a 2 MB buffer makes a drive four times as fast as one with a 512 KB buffer! In fact, you'd be hard pressed to find even a 4% difference between them in most cases, all else being equal. Not surprisingly, both external transfer rate and cache size are overrated for the same reason: they apply to only a small percentage of transfers.

The cache size specification is of course a function of the drive's cache characteristics. Unfortunately, manufacturers rarely talk about any characteristics other than the cache's size.

spinning rust HDD performance is a complex thing (lately, determined more often than not by firmware optimizations) and cache size is not a silver bullet by any means.

Rotational speed is a much better predictor of drive performance in my experience: 5,400 rpm is slower than 7,200 rpm, which is slower than 10,000 rpm, which is slower than solid state. Even then there are (rare) exceptions.

That said, at such a low price difference, probably worth it.

Solution 3

It is probably not worth the money, though it depends on your usage.

• When you read large continious files (typical for movies etc) then you are limited to the speed of the actual data transfer of the spindles. Cache will not boost this.
• When you write large files (e.g. isos) then you are limited to the speed of the actual data transfer of the spindles. Cache will not boost this.
• When you write lots of small files then these can fit in the cache and the drive can reorder this writes for higher efficiency. This will help.
• When you read lots of small files and you get lucky then the next file to read may already be in a cached track. It is possible that it helps.

Conclusion: It depends on usage.

Solution 4

For some reason you'll rarely see this mentioned, but the primary benefit of the cache is to reduce the number of time-intensive disk head movements. Picture this scenario, one file is being read, and another file is being written at the same time. Without a cache, the head would be jumping all over the place trying to service these independent file streams. However, with a cache, the file being written can be temporarily stored in the cache, while the other file being read continues to be pulled from disk. If the file being written fits entirely in the cache, the hard drive signals to the OS that the write has completed, allowing the program that was writing the file to continue, despite that the disk cache hasn't yet completed writing the file to the physical disk. Meanwhile, the reading of the first file never stopped. In this way, significantly higher apparent bandwidth was accomplished, to the extent of the cache size. This may not be significant when dealing with 2 MB caches, but newer large capacity drives have 256 MB caches, which is enormous, and even under high read/write loads, will give highly optimized read throughput (very close to the physical disk's rating) while simultaneously supporting full SATA write bandwidth, until you run out of cache. Usually, you'd balance the number of drives in your RAID to match your throughput requirements so you don't typically exceed your cache capacity.

Solution 5

The buffer also saves wear and tear by reducing unnecessarily repetitious reads and writes. Buffer size affects speed and wear more in some apps than others, but any time the buffer is exceeded, you lose both performance and durability. That's why they are increasing it---win win situation.

View more solutions

Comments

[jtreser]

I agree with this comment. I have seen a significant difference when using a WDC 5400 rpm drive vs. a 7200 rpm drive but no noticeable difference with two like speed drives using different cache size with drives running at the same spindle speed.

[ssvarc]

Agreed that at around $3(?) why not. Likewise agreed that there is very little difference in how fast it will go. More important is the spin speed.

[Pacerier]

citation needed

[Pacerier]

@JeffAtwood, What do you mean by "firmware optimizations"? Are there even real harddisk algorithm optimizations in this time and age? Don't they all use the same standard algorithm these days?

[I say Reinstate Monica]

Welcome to Super User. This answer looks good. It could be made better with a little formatting, such as a paragraph break or two.

[Sean]

That's what you get for free. Line breaks cost extra