Talk:Disk buffer

Rename

The edit comment "I've never heard of "disk buffer" in such context; I think "HDD internal cache" is 100% WP:COMMONNAME here" is absolutely right; the name "disk buffer" is WP:OR (although I do find it more appropriate). Note that the article's content was not written by me, but was split from the cache article long ago.

What alternatives would you suggest? "HDD internal cache" sounds cumbersome and is probably tedious to link without using pipes. Perhaps "disk internal cache"? -- intgr [talk] 13:57, 27 September 2007 (UTC)

"Internal disk cache"? Although this one isn't completely self-explanatory, it could make people wonder "internal to what?" -- intgr [talk] 14:01, 27 September 2007 (UTC)

"HDD cache" "disk cache" or "hard drive cache" would be an improvement. Fractal618 (talk) 17:26, 7 February 2013 (UTC)

Software?

What about software-based caching? Every mainstream operating system maintains a software cache too, regardless of what the hardware is doing. Agree with the comment about 'disk buffer' above - have never in 20 years heard it called this before (not even the manufacturers use the term). Socrates2008 22:44, 25 October 2007 (UTC)

Benchmarking

Tom's Hardware did a test comparing 8MB and 16MB cache. It seems that large buffer sizes are irrelevant after 8MB. Here is the link: http://www.tomshardware.com/reviews/understanding-hard-drive-performance,1557-14.html Maybe someone should include this in the article. —Preceding unsigned comment added by 98.213.141.241 (talk) 08:18, 4 July 2008 (UTC)

The very limited comparison in the Tom's Hardware Guide reference is between 8 and 16 MB cache, on a particular hard drive, in a particular set of tests, not between 8 MB and any other larger caches, and not on all disk drives or under all circumstances. The conclusions about the effect of cache size on performance in general are completely unsupported. The reference to cache effects in Windows 7 refers to solid-state drives only. — Preceding unsigned comment added by JohnOFL (talk • contribs) 01:01, 29 October 2011 (UTC)

Considering these objections, I am removing the 'Performance' section. There are other sources that encourage large cache sizes in hard drives, e.g. http://www.kitchentablecomputers.com/hdrive.php or http://en.kioskea.net/faq/2831-choosing-your-hard-drive or http://www.newegg.com/product/CategoryIntelligenceArticle.aspx?articleId=209 -- not to mention that Other World Computing (well-respected computer folks) sells hard drives with as much as 128MB buffers: http://eshop.macsales.com/shop/firewire/1394/USB/EliteAL/PerformanceRAID. I think there should be a section on cache but I haven't been able to find anything conclusive to cite yet. Begeun (talk) 16:48, 2 July 2012 (UTC)

I have no reason to really trust any of those sources. Who's to say that they're not just regurgitating what they heard? Note also that believing you are right =/= you being right. I would much rather believe empirical sources, ie the Tom's Hardware bench. It's by no means conclusive, but it's better than just some generalizations scattered around the web. And I'm inclined to say JohnOFL is right. Those tests showed that disk buffers for that particular drive doesn't matter. But I'm not sure that increasing buffer size from 32MiB to anything higher would have made a difference, nor do I know what he means by "under all circumstances." I am therefore inclined to say for that particular drive, disk buffers don't matter. -- BlueFenixReborn (talk) 00:42, 22 December 2014 (UTC) Edit: That of course says nothing conclusive about other drives.

I agree with BlueFenixReborn that this is inconclusive. The sources provided by Begeun are all what I'd consider untrustworthy, either clickbait articles or trying to sell you more expensive products. Those Mercury boxes are RAID systems and RAID cache is not quite the same as a disk buffer. The Tom's Hardware article from 2007 is also hardly relevant any more. Let's not recreate this section without more solid research to cite. -- intgr [talk] 12:08, 22 December 2014 (UTC)

Right, there's no point in even remotely comparing those RAID boxes with bare HDDs. Also, how would we reliably benchmark the effects of different HDD buffer sizes? In other words, how to find two modern HDDs that are mechanically and firmware-wise identical in everything except in buffer sizes? Such things simply do not exist; the only reliable way would be to test one HDD first with its whole buffer, and to disable one part of the buffer for the second batch of benchmarks. However, I'm not aware of the ways to do something like that, short of hacking HDD firmwares – what would be a tall order. — Dsimic (talk | contribs) 13:40, 22 December 2014 (UTC)

Well, the good thing is that we (Wikipedia editors) don't need to worry about coming up with a methodology, that's what sources are for. :)

But I have a feeling that the real reason is something mundane, such as larger RAM chips being cheaper simply because their production volume is much higher. -- intgr [talk] 15:36, 22 December 2014 (UTC)

That totally makes sense, as another question (which supports it) is why no HDD manufacturer bumps the disk buffers further so, for example, 4+ TB HDDs come with large 256–512 MB buffers as a "killer" speed improvement feature? They've probably tried that already and concluded that operating systems already do enough of the caching, resulting in no perceptible speed improvements. Also, much larger buffers would increase the dangers associated with write caching (or buffering, in this case :). — Dsimic (talk | contribs) 22:16, 22 December 2014 (UTC)

I would like to see some sort of performance section in this article. At least it should provide an example of when 64mb cache trumps a 32mb cache. (http://www.overclock.net/t/664870/32mb-vs-64mb-cache) (http://www.tomshardware.com/forum/261964-32-justifying-64mb-cache) Fractal618 (talk) 17:34, 7 February 2013 (UTC)

text picture mismatch

disk buffer (often ambiguously[citation needed] called disk cache or cache buffer[citation needed]) is the embedded memory in a hard drive

Why is the the picture next to that sentence claiming disk buffer is on the SCSI host adapter? That is obviously not "embedded memory in a hard drive". --Xerces8 (talk) 15:11, 28 April 2011 (UTC)

I replaced the image with the back of a hard drive. If you want, you could make an image with the disk buffer chip circled.--RaptorHunter (talk) 21:17, 28 April 2011 (UTC)