# 64-bit OSes and software: is there a performance edge? If so, for what applications, and what benchmarks were used to assess performance?

I am not sure if a 64-bit OS will ever show a difference in performance because most applications are not written or designed for 64-bit systems. The only advantage I can see right away is that we can add a ton of RAM to our machine, and that will make our apps run faster.

It is very anecdotal but I have a library that I compiled a 64-bit version of and it runs about 3x faster on 64-bit computers than on 32-bit computers. It does LOTS of math and works almost exclusively with VERY large numbers ... that might explain it really.



Edited at 2008-11-10 12:14 am (UTC)

Not really 'reliable information', but here is my own media use:

the shelf life of magnetic vs. optical media (esp. SATA HDs vs. DVD-Rs & Blu-Ray)?

Both are very seriously exaggerated in my experience.

Optical media:
I haven't archived my DVD collection to BD yet (will only do that when double-layer discs will become available in stores here), but I've just archived ~185 CDs and there were so many discs that were either unreadable altogether or had several read errors it's not even funny. I kept a tally in a SPSS window about various variables that occurred to me (brand, 650 vs 700 megs, labeled/unlabeled, etc) to see which influence the outcome most, but I haven't done the stats yet. I still have about 200 CDs at another location I need to archive and I want to put those in the file too. When I'm finished I'll post a writeup. Based on how it looks, I think the brand of the disc might be even more important than how frequently the disc was used (since this is hard to operationalize, I just made it a simple dichotomous variable of games/everything else because everything else is mostly archived material that does not get used as often as games which usually require a disc in the drive... though these days I just make image files out of everything, but that wasn't so in the CD days).

HDDs:
I have about 5-6 hard disks at any given time, usually decent brands and models, and I have about one hard disk failure per year. (OK this is not 'shelf life' strictly speaking, because they are in computers, but HDDs need to be spun up once in a while even if they are used for archival purposes AFAIR.) Most of them are SATA now, I'll see how long they last, the IDE generation was a disaster.

I would not rely on either. I try to back up everything multiple times and on different media. I think even due to the considerable CD read problems I just experienced I only lost one thing... the full CD version of Loom, the Lucasarts adventure game. Argh! P2P to the rescue... (why can't Lucasarts just sell this for a few bucks like they do with their other adventure games?)

For the optical media connoisseur, there is no better site than http://www.cdfreaks.com for reviews of archival properties of various optical media. This article indicates that ISO is beginning a standardization process which will make more transparent what the cdfreak community does laboriously by hand. My understanding of magnetic media (hard disks) is that the shelf life is mostly affected by the wear of moving parts; this is opposed to magnetic tape, which has its own well-established standards for archivability.

My only understanding of 64-bit performance enhancements came from a study that AMD did with their Opteron line (not able to find at the moment, I'm afraid). They specifically stressed that intensive numerical computation with few cache hits could be improved by up to ~50%, because the Opteron is capable of performing two most 32-bit operations in one clock cycle (applications they tested were, AFAICR, mostly discrete-event simulation and large matrix manipulation). There is also the increased RAM argument, which has a lot of traction from the "no tertiary storage media!" crowd, but I have not seen any empirical data to support the anecdotal evidence.