lkml.org 
[lkml]   [1997]   [Mar]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: MMX Technologie


On Sun, 23 Mar 1997, Keith Rohrer wrote:

> Dunno about AMD cache sizes.

Have a look at http://www.amd.com/html/products/pcd/mpf/pres1.html and
follow the arrows. The main features are:

Cache are 32kB each (I and D), but only 2-way set associative.
Instruction fetcher performs predecoding and stores the additional
information along with the instructions in I-cache.

Floating point multiply takes only 2 cycles (5 on PPro).

Floating point add takes 2 cycles (3 on PPro).

Much lower branch misprediction penalty than PPro.

Fits into a Pentium socket.

But I don't have any to check their claims. Anyway it looks good even after
removing marketing hype.

Gabriel

\
 
 \ /
  Last update: 2005-03-22 13:39    [from the cache]
©2003-2011 Jasper Spaans