From owner-freebsd-chat  Sun Nov  9 10:44:43 1997
Return-Path: <owner-freebsd-chat>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id KAA27353
          for chat-outgoing; Sun, 9 Nov 1997 10:44:43 -0800 (PST)
          (envelope-from owner-freebsd-chat)
Received: from srv.net (snake.srv.net [199.104.81.3])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id KAA27348
          for <chat@FreeBSD.ORG>; Sun, 9 Nov 1997 10:44:40 -0800 (PST)
          (envelope-from cmott@srv.net)
Received: from darkstar.home (tc-if3-36.ida.net [208.141.171.141])
	by srv.net (8.8.7/8.8.5) with SMTP id LAA06239;
	Sun, 9 Nov 1997 11:44:33 -0700 (MST)
Date: Sun, 9 Nov 1997 11:44:01 -0700 (MST)
From: Charles Mott <cmott@srv.net>
X-Sender: cmott@darkstar.home
To: Alfred Perlstein <perlsta@cs.sunyit.edu>
cc: chat@FreeBSD.ORG
Subject: Re: IDT processors?
In-Reply-To: <Pine.BSF.3.96.971109123631.18731B-100000@server.local.sunyit.edu>
Message-ID: <Pine.BSF.3.96.971109112749.24278B-100000@darkstar.home>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-chat@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

[moved to -chat from -hackers]

On Sun, 9 Nov 1997, Alfred Perlstein wrote:
> from what i've heard the IDT chip's FPU is garbage, and it's integer
> performance is terrible.

This from the IDT data sheet:

  "The floating-point unit is designed to maximize clock frequency and to
   minimize chip size while providing adequate levels of floating-point
   performance for typical desktop use. Some floating-point instructions
   are pipelined, but some are only partially pipelined."

So perhaps Alfred's description of the FPU is on the mark.  IDT admits
they are targeting the low-end, socket 7 part of the market.  Still, most
people who have done some digital circuit design take IDT seriously.  I
wonder if somehow slot-1 is Intel's equivalent of the IBM microchannel
bus?  Intel's competitor's might be wise to establish an alternate socket
spec, basically one with growth capacity to 128 bit data paths.

It is now possible for a team of 20 people to design a Pentium clone in
two years or less.  Intel has an incentive to make their technology as
obtuse and complex as possible (just like Mircrosoft has an incentive to
continuously add glop to their OS). IDT says in their documentation that
they dumped most of the more obscure Intel instructions into slow
microcode and basically concetrated on a decent RISC core. 

Charles Mott

[I'm finished with posts on this non-FreeBSD subject]