| 1 |
On Fri, 2007-07-20 at 09:43 -0500, Dustin J. Mitchell wrote: |
| 2 |
> On Fri, Jul 20, 2007 at 09:21:10PM +0800, P.V.Anthony wrote: |
| 3 |
> > The reason for this questions is that there are some information on the |
| 4 |
> > net that says that there is no much difference between them. |
| 5 |
> > Is that true? Thought that 64bit is always better. |
| 6 |
> |
| 7 |
> Building a system 64-bit buys you: |
| 8 |
> - wider integers (so math with 64-bit integers is faster) |
| 9 |
> - wider pointers (so an application can have a *lot* more address space |
| 10 |
> allocated to it) |
| 11 |
> - bigger binaries and data structures (so more RAM consumed) |
| 12 |
> - future-proofing (in a few years, 32-bit hardware will not be |
| 13 |
> available new) |
| 14 |
|
| 15 |
And a few more for amd64 vs. x86, specifically (don't apply to |
| 16 |
sparc/sparc64 or mips/mips64 or...): |
| 17 |
|
| 18 |
- Twice as many registers (which can be a big win for some workloads) |
| 19 |
- More than ~1G of RAM without HIGHMEM (which is a win in memory |
| 20 |
access speeds, if you're using that RAM) |
| 21 |
- More than 4G of RAM without HIGHMEM64 (which is a *huge* win in |
| 22 |
memory access speeds, if you're using that RAM) |
| 23 |
|
| 24 |
Downsides are lack of support for many (most?) binary-only packages. |
| 25 |
|
| 26 |
In my limited speed testing, my 64-bit installs were all faster for my |
| 27 |
general use cases (basically desktop) than 32-bit on the same hardware. |
| 28 |
My server is also 64-bit (I run lighttpd/php/netqmail/mysql), and it's |
| 29 |
rock solid, but I never did any performance testing on it. |
| 30 |
|
| 31 |
Daniel |
| 32 |
|
| 33 |
-- |
| 34 |
gentoo-amd64@g.o mailing list |
Archives