h a l f b a k e r yPoint of hors d'oevre
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
(by the way, in which category do benchmarks go?)
If you bought a computer in the past few years, it probably has a 64 bit CPU. The x86-64 architecture (sometimes referred to as AMD64 or EM64T) was invented by AMD as the successor to the X86 architecture, and has since been picked up by Intel. It
is in AMD's processors since the Athlon 64 and in Intel's since the Pentium D.
But unless you're a power user, you're probably not running a 64bit OS (examples of 64bit OSes include Windows Server 2k3 x64, XP Pro x64, Vista x64, and 64 bit versions of Linux, OSX, and *BSD). And even if you are a power user, you probably have some 32bit apps and drivers kicking around in there.
As a power user, the lack of attention that x86-64 is getting annoys me to no end. Ideally, I would like my 64bit processor to be running 64bit code, no? We've been at this since the Atari Jaguar, right? And in theory, the less 32bit drivers and apps you use, the faster things will be (in practice the difference is negligible, but for power users its the thought that counts).
So the idea here is a simple benchmark, yet another number power users and companies alike can brag about to anybody and everybody. This would simply be the percentage of 64bit binaries on ones system, a higher percentage would be considered better than a lower. Accuracy past the decimal point is a nice thing folks could fight over as well (any benchmark needs a niggling point). This could be called a RU64 number.
Calculating the benchmark would be simple. The program simply needs to look at the headers of all the binaries on the system and check for 64 bittyness. The portable unix "file" utility is one example of a program that can do this. It would probably, then, be easy to knock out a proof of concept shell script using find, file, grep, and awk.
[link]
|
|
Interesting - how much does it matter though if your 64-bit hardware is running 32-bit code? I mean, is it half as fast as it might be?
I'm running MacOS which is really BSD Unix in disguise - I've no idea if it's 64-bit or not. |
|
|
I have another idea: 64bit printer drivers. They DO exist, but haven't caught on yet. |
|
|
There is a 64bit edition of OS X, you can tell which you're using by running uname -a |
|
|
See [link] for a pretty good 64bit OS enthusiast forum. Lots of info, software and drivers there. |
|
|
I'm doing a big hardware upgrade on my notebook pretty soon, so I decided to upgrade the OS, too. |
|
|
I started out trying to get XP x64 to work. I couldn't find all the drivers I need, though. I did find most of them. It doesn't help that Fujitsu (the notebook is a T4220) US's policy is to not support XP x64 or Vista x64 AT ALL. Fujitsu Europe did, however, have all the drivers I need to set up Vista x64 (with the disclaimer that they won't offer tech support for that configuration). |
|
|
So I'm working on installing Vista x64 business (which I am currently downloading, it's one of the perks of being in college; free software from M$). I'm gonna strip it down with vlite till its somewhere between standard Vista and Server 2008 in bloatedness. Also I'll slipstream the updates and drivers in, as well as all the basic software (Firefox, Daemon Tools, 7zip, etc), that way I'll have a very convenient install disc for later whenever it breaks. |
|
|
This will be dual boot, I'm trying to concoct a 64bit edition of Linux Mint XFCE (by adding the XFCE stuff to the regular 64bit edition). XFCE for me has more to do with preference than speed, it doesn't make much of a difference, but Gnome and KDE just rub me the wrong way. For convenience purposes it will be installed through Wubi as a file on the NTFS partition. |
|
|
As for the hardware, I'm getting a 120gb OCZ Apex SSD and 2x2gb of DDR2 6400 (4 gig sticks of ddr2 laptop ram are crazy expensive, so no 8 gig for me). If that all goes well I'll upgrade the processor to a new 45nm one, whichever is the best with a 800mhz FSB (the motherboard won't support 1066mhz), and get a 500gb drive (there's a newly released Seagate that's really good) for the optical bay to make easy backups. |
|
|
I put my commodore 64 away years ago. |
|
|
## "...grandchildren on your knee, Vera, Chuck and Dave" ## |
|
|
//I mean, is it half as fast as it might be?// |
|
|
Not necessarily, it depends on what it's up to. My left hand is a 5-bit device and depending on how I choose to represent integers on it; I can count to 5, or 31 without having to refer to another memory address. (The only opcodes I can reliably perform on my left hand are INC and DEC (though DEC can be tricky sometimes) and I don't have a carry bit - tongue?) Anyway, if I wanted to count to larger numbers without resorting to Paging (writing something down on a Page) then I'd benefit from a second memory address - at which point, things would get significantly faster as I wouldn't have to keep jotting things down all the time. |
|
|
Well, speaking from experience, 32bit drivers make a 64bit OS take longer to boot up. (note that only userland drivers may be 32bit, the rest must be 64bit) |
|
|
//count to 5, or 31// if you're actually *counting* something then presumably you have something to count, therefore 32 not 31.... or if that was a word-choice typo then 4 not 5. </pedant> |
|
|
[edit] 6 not 5 (not 4 not 5) |
|
|
No, I'm not sure I understand what you mean - it is both 31 and/or 5, as if I don't have any fingers 'out' then the value is zero in both cases, If I have all fingers out then I have counted either to 5, or to 31. |
|
|
5^2 = 32 possibilities. However, there must be a value reserved for 0, so you can only count to 31 with one hand. |
|
|
[zt][Sc] good point, well in that case you can only count up to 15 or down to -15... that will give you +0 and -0 to play with too. ;D (I've never understood how -0 is supposed to work... or +0 for that matter) |
|
|
"Do you have bananas ?"
"Yes, we have no bananas." |
|
|
Generally, -0 is skipped and we count from -15 to 16. |
|
|
Unlike computers, which generally count from -16 to 15.* |
|
|
*At least, the 5 bit, 2's complement ones do. |
|
|
Well I *did* flunk digital logic... |
|
| |