Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!cica!tut.cis.ohio-state.edu!netsys!dsinc!wells!edw
From: edw@wells.UUCP (Ed Wells)
Newsgroups: u3b.tech,comp.sys.att
Subject: Re: Are there any redeeming qualities of a 3B5/3B15?
Summary: 80386 vs. AT&T 32100 chip
Message-ID: <61@wells.UUCP>
Date: 13 Jul 89 07:49:33 GMT
References: <1146@vsi.COM>
Organization: Wells Computer Systems Corp., Levittown, Pa. 19058
Lines: 43

Stephen J. Friedl writes:
>      We have a customer with a 3B15 running Sys V Release 3.1,
> and we generally like it a lot (I develop on it about half-time)
> However, my Everex STEP/25 80385 machine beats the 3B15 for every
> benchmark I can find, and this surprises me.  While certainly a
> 25MHz 80386 with a good cache will beat it for CPU-bound tasks, I
> find that an ESDI drive on the 386 beats the SMD drive on the
> 3B15.  I suspect that while the SMD interface has much higher
> bandwidth, the 14MHz WE32100 CPU just can't push it fast enough.
> 
>      Does anybody know of things where a 3B15 will beat an
> 80386 machine (or a 3B2/600 for that matter)?  I would really
> like to avoid AT&T bashing here, but real experience is welcome
> no matter how brutal.

  I have also done some interesting testing.  I have a 16 Mhz cached 80386
clone on the AT bus and a VME/3B system (a 3B on the VME bus made by
AT&T).  The VME/3B is a 18 Mhz version.  Both systems have 4 MB of ram
memory at 0 wait states, hard drives with comparible access times.  If fact,
the VME/3B is 12 mS and the 80386 is about 18 mS.  The 80386 is running
Xenix and the VME/3B is running System V.3.  Althought the operating
systems are different, I put something in memory and monitored its in-memory
progress while I tied the processors up with other memory/disc activity
to try to get an idea as to how its performance was effected.  This
should have given me a good "processor" test.  Neither processor was
busy enough to cause swapping.

  In every case with single tests (while in multi-user mode), the '386
beats the VME/3B, however, on big difference is that the 80386 is known
to fall like a rock on performance whereas the VME/3B has a nice gradual
slope.  I have generally found that the 80386 is normally tied up in
mountains of page faults per second when the performance vanishes.

  I like both systems, however, they both have their pros and cons.
                  .....
  Anybody else have similar experiences?  I'd like to hear from you
if you have.

-- 
=========================================================================
Edward E. Wells Jr., President			    Voice: (215)-943-6061
Wells Computer Systems Corp., Box 343, Levittown, Pa. 19058
{dsinc,francis,hotps,houxl,lgnp1,mdi386,pebco}!wells!edw