Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!elroy.jpl.nasa.gov!decwrl!sgi!shinobu!odin!patton.wpd.sgi.com!jmb
From: jmb@patton.wpd.sgi.com (Jim Barton)
Newsgroups: comp.sys.sgi
Subject: Re: 4D4X0 ??
Message-ID: <1991May2.012911.10176@odin.corp.sgi.com>
Date: 2 May 91 01:29:11 GMT
References: <15089@darkstar.ucsc.edu> <100054@sgi.sgi.com> <3985@uc.msc.umn.edu> <100399@sgi.sgi.com>
Sender: news@odin.corp.sgi.com (Net News)
Organization: Silicon Graphics Inc.
Lines: 33

In article <100399@sgi.sgi.com>, jeremy@perf2.asd.sgi.com (Jeremy Higdon) writes:
|> From: jeremy@perf2.asd.sgi.com (Jeremy Higdon)
|> Newsgroups: comp.sys.sgi
|> Subject: Re: 4D4X0 ??
|> Message-ID: <100399@sgi.sgi.com>
|> Date: 30 Apr 91 04:26:16 GMT
|> Organization: Silicon Graphics, Inc., Mountain View, CA
|> 
|> In article <3985@uc.msc.umn.edu>, ken@uh.msc.umn.edu (Ken Chin-Purcell) writes:
|> > Jeremy Higdon writes:
|> > 
|> > > The 4D/400 family are 40 mhz R3000 machines.
|> > 
|> > Has either bus (Iris or VME) gotten faster?  We're hungry for
|> > throughput....
|> > 
|> 
|> The VME and MP buses run at the same speed, but with the 1MB second
|> level cache, you will probably get a better cache hit rate; thus, you
|> shouldn't need as many MP bus accesses.
|> 
|> 

Now wait a second. The MP bus SUSTAINS 64Mb/s read bandwidth from memory, and
32Mb/s write bandwidth to memory/IO. The IO DMA engine is built to run at 
32Mb/s, so that is the theoretical peak for IO bandwidth.

The VMEbus PEAKS at 32Mb/s in the best of all possible worlds. Measured IO3
(POWERChannel) VMEbus performance is 24-28Mb/s with fast IO controllers.

-- Jim Barton
   Silicon Graphics Computer Systems
   jmb@sgi.com