compare
|
1 to 12
|
1
|
</tr> <tr> <td>shifts</td> <td>5 to 14</td> <td>1</td> </tr> <tr> <td>divide</td> <td>>100</td> <td>1</td> </tr> <tr> <td>sqrt</td> <td>>20</td> <td>1</td> </tr> <tr> <td>long->double</td> <td>>50</td> <td>3</td> </tr> <tr> <td>double->long</td> <td>>50</td> <td>3</td> </tr></table>
Amazing huh? Likewise, floating point operations have
been showing 100% speedups over G4 implementations. Because the G4 was a
32-bit architecture, any 64-bit operation had to be split into two
32-bit units, operated upon, and combined again. This may not speed up
your dragging and dropping (yet), but for workstation-esque tasks it
will provide a world of difference at a similar or fraction of the cost
of many comparable offerings from
SGI,
IBM,
Alienware,
Sun,
and HP.
If you’re looking for a huge boast over the G4 in number crunching, then
by all means consider the G5; you will not be ripped off – it’s a
workstation at a high-end desktop price. However, for general usage I
would suggest going with the G4 based lines still available (and
dropping in price daily). I myself am going with the G5 because I can
use a speed boost in FP and long operations, and if I can’t then I will
invent something that will.
-m