On Tuesday 22 May 2007 21:01, Tom Miller wrote:
> ...
>
> Now memory isn't the only factor, but is good enough for this
> discussion.

In large part, memory _is_ the only factor. Or, at least, a very large 
one. Memory bandwidth, that is. Processors speed increases have 
continued to outstrip that of both primary and secondary storage, 
though on a per-CPU (core) basis, that has pretty much come to an end. 
Now more CPUs (or cores) are being placed on a given chip or board, and 
the effect is the same.

So for many (I'd venture to say most) applications, memory bandwidth vs. 
demand is the dominant performance factor. Thus, 32-bit systems will 
quite often outperform 64-bit ones. Unless, of course, you really need 
something that is only possible with the 64-bit system (very large 
virtual address spaces, that is).


> Things that use a lot of math also benefit with 64bit 
> pointers, as the math is just faster.  So unless you are running a
> mission critical DB or some vector graphics stuff, 64bit won't help
> either.

Don't you mean a 64-bit memory bus? Bigger pointers are all downside. If 
you must address that much memory, then there's no alternative, but 
it's all overhead, regardless.


> ... 


Randall Schulz
-- 
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to