Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ho95b.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!ihnp4!houxm!ho95b!wcs
From: wcs@ho95b.UUCP (Bill Stewart)
Newsgroups: net.arch,net.unix,net.micro.att
Subject: Re: AT&T 3B-2's
Message-ID: <406@ho95b.UUCP>
Date: Wed, 8-May-85 20:55:01 EDT
Article-I.D.: ho95b.406
Posted: Wed May  8 20:55:01 1985
Date-Received: Thu, 9-May-85 03:19:03 EDT
References: <677@mtuxo.UUCP> <514@houxj.UUCP>
Organization: AT&T Bell Labs, Holmdel NJ
Lines: 29
Xref: watmath net.arch:1165 net.unix:4454 net.micro.att:11

> To get data sheets on the WE 32000 processor in the 3B-2/300,
> I think you are supposed to call (800) 372-2447.
> 				Bill Dietrich
> 				houxj!wapd

Actually, we just announced the 32100 series, which is the successor to the
32000 series.  The early 3B2s used the 32000; I think some of the later
3B2/300s used the 32100.  Certainly anything new will use it.
The 32100 series, by the way, includes a floating point chip, which takes
care of the main performance lose in the 3B2 (good software has done half
that job, though.)

There have been some comments recently about FFTs, and how it probably
wouldn't make sense to build a special-purpose machine to do them.  We have a
chip called the DSP-32 Digital Signal Processor, which is designed for
fast repetitive floating point calculations (4 or 8 MFLOPS, depending on what
you count).  Basically it's designed for doing FFTs and similar signal
processing for communications equipment.  I think we sell them?

I don't do chips, but the phone number above is probably the right people to
call for 32100 documentation.
		Bill

Disclaimer: Sorry if this is too commercial, but it seemed germane to the
discussion.
-- 
			Bill Stewart	1-201-949-0705
			AT&T Bell Labs, Room 4K-435, Holmdel NJ
			{ihnp4,allegra,cbosgd,vax135}!ho95c!wcs