Subject: Re: Cache for VS3100/M30/M38? [Was Re: VS3100/M76 is FAST!]
To: None <port-vax@NetBSD.ORG>
From: Allison J Parent <>
List: port-vax
Date: 06/02/1998 11:01:16
<There may be (or not be) cache on the CPU/motherboard. It can be figured
<out by the contents of some internal registers. If there is cache; it is
<a primary 1K I/D cache and a secondary 64K cache. I think that the cache
<is enabled by the ROM VMB, but I'm not sure. Will verify that later.

Good point Ragge.

Cache can be one of three options if that cpu has it.  NONE, turned off.
internal 1k and none external.  Both internal and external.

Liekly more than most want to know but... here goes.

IF any of you with a 486 system have a mind to....
  My results on a 486dx2/50 64k external cache and 486 internal 8k cache.

 Run a benchmark with both caches running (dos or whatever).
   DOS/WIN3.1 norton sysinfo 108 (386dx/33 is 34 as a reference)!
 Run with external cache turned off, note perfomance drop.
   dropped to 39!
 Run with external but no internal cache.
   Dropped further to 16!
 Run with both turned off, really slow now!
   A whopping 6.9...  I first thought it died but waited... and waited.

FYI: a 386sx/16 running the same stuff (no caches there) scored 16!!!
The 386 is a slower processor (more internal cycles needed for a given 
instruction than 486) yet it's better matched to the memory and gives
slightly better results.   As an experiment I dropped the 486 clock
to 20mhz and while generally slower for the first two tests the last two
showed little difference!  The reason is the last two tests depend on 
external memory cycle speed and at 20mhz I could drop wait stats for DRam
to 0 and while the cpu was slow the ram was a tiny bit faster, net results
showed similar results until it became compute intensive.   Generally for 
micros the design and implmentation of internal caches yeilds more speed 
than external as they don't require external bus cycles to get access so 
they often offer gar greater speed improvement than external (note 486 
results).  I have a 386dx/33 with external cache and the difference 
between enabled and none is far smaller as it's an external cache only.

Cache makes a difference but CPU speed and memory speeds interact and in 
some cases it will be memory access that limits the system regardless of 
the CPU speed.