Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!wuarchive!mit-eddie!rutgers!bellcore-2!bellcore!messy!mo
From: mo@messy.bellcore.com (Michael O'Dell)
Newsgroups: comp.arch
Subject: Re: seeking general purpose workstation w/large memory
Keywords: memory RAM huge big
Message-ID: <25941@bellcore.bellcore.com>
Date: 6 Aug 90 19:03:10 GMT
References: <1990Aug6.181339.21156@caen.engin.umich.edu>
Sender: news@bellcore.bellcore.com
Reply-To: mo@messy.UUCP (Michael O'Dell)
Organization: Center for Chaotic Repeatabilty
Lines: 28


Hmmm.....

The last time I checked 32 bits of address would only get you
4 gigabytes, minus whatever the OS wanted out of the
user virtual space, since most Unix systems put the kernel
in the uppermost or lowermost part of shared virtual space.

I guess on a 386 or an RS/6000 you might do a bit more,
but if you compute how much POWER that large a RAM array
would consume (and how warm it will make your office!)
you will discover why no workstation has
that much memory on it.  Not to mention that when you
get that many ram chips going, you really need to start
doing 3-Detect/2-Correct memory ECC, rather than the
2-Detect/1-Correct that many (but certainly not all!!!!)
systems do.  I believe the large IBM mainframes do 3D/2C
already, and the folks with Big Bucks anxiously wanting to
buy such machines will tell you quickly they agree with
the need to do that, lest your
	Mean-time-between-uncorrectable-memory-error
be come unacceptable.

And as for the price of memory making this attractive, the
price of the ram chips doesn't begin to represent the entire
memory subsystem costs, even if it is substantial.
	
	-Mike