🚀 go-pugleaf

In article <5k5m9a$ad3@catapult.gatech.edu>, ccocsjo@acmey.gatech.edu (Who?
Me?) wrote:
> And actually, for a lot of things, MMX might be slower.  The floating
> point registers are used for MMX to, and it has to do some kind of
> context switch to go from MMX to floating point and vice versa.  I
> believe the numbers I saw were that it takes about 50 cycles to make
> this switch in one direction, and 35 cycles in the other direction.
> In addition, if the software you are using does not explicitly take
> advantage of the MMX instructions, there is *no* difference between
> using an MMX chip and a non-MMX Pentium (except the price you pay for
> it of course).  If you are running Linux, BSD, Solaris, etc then there
> is probably no point to buying an MMX chip at the present time.  I would
> suspect there will be a lot of programs that use MMX for Windows 95 soon,
> I am not sure about Windows NT.


Exactly.  Its a lot of hype from Intel, but it means very little to most
people.  I laughed when I saw someone in my department looking to buy a
P200+MMX, what a waste of money.

One more quick fact:  The MMX aware version of Adobe Photoshop was created
by Adobe and tweaked by Intel.  In the demos of how fast MMX is they
compared a Penitum to a Pentium+MMX running the MMX tweaked version of
Photoshop.  Of course it was faster on the MMX machine, but only when
applying certain filters within certain parameters, when the parameters
changed the difference in performance dramatically decreased.

Buy a PowerPC based Macintosh.  MHz for MHz they are about twice as fast as
Pentium (PPC604e vs Pentium) and the OS is superior to Windows95, as well.

Dean

-- 
"They load the clip in Omnicolor, They Pack the nine,  ____
They fire it at prime time... Ya brain dead!  Ya       \ _/__
gotta fuckin' bullet in ya head!"                       \X  /
                -Rage Against The Machine               Bi\/
Dean H. Saxe is dean@cerevisiae.biology.emory.edu         S/M