Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!crdgw1!crdos1!davidsen
From: davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr)
Newsgroups: comp.sys.ibm.pc
Subject: Re: CACHE on 386 and GRAPHICS PERFORMANCE.
Summary: It depends
Message-ID: <427@crdos1.crd.ge.COM>
Date: 21 Sep 89 11:52:50 GMT
References: <12718@s.ms.uky.edu>
Reply-To: davidsen@crdos1.UUCP (bill davidsen)
Organization: GE Corp R&D Center
Lines: 33

In article <12718@s.ms.uky.edu>, munish@ms.uky.edu (Munish Mehra) writes:
|  I am putting together a 386 which will be used for graphics along
|  with other things.
|  
|  My question is that does the cache make a significant difference
|  and if so how much ?

  It depends on the speed of the memory you use with the system. If you
have 60ns interleaved column static you will not save much. If you have
slow memory you will save a lot. Two reference points:

System A: AMI MB, 32 bit 0w/s on MB, 16 bit 1w/s memory on AT bus.
Without cache the 16 bit memory is 35-40% slower. With cache enabled
10-15% slower.

System B: AMI 25 MHz with 4MB 60ns, 32 bit 0w/s on board. Turning on
cache saves about 5%.

  If you will be running fast interleaved memory the cache will be of
less use than if you have to save a few bucks and go with slower memory.
You can pay for the cache, or pay for the memory and fancy circuitry.

  Another consideration is that if you run DOS and use extra memory for
RAM, EMS, etc, you are not *executing* in the memory above 1MB, so a
system which has a fast 1MB on the MB may be fine without cache. If you
run UNIX or OS/2 you take a much larger performance hit.

  Sorry there's no one yes or no answer.
-- 
bill davidsen	(davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen)
"The world is filled with fools. They blindly follow their so-called
'reason' in the face of the church and common sense. Any fool can see
that the world is flat!" - anon