Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!mailrus!tut.cis.ohio-state.edu!rutgers!uwvax!tank!nic.MR.NET!eta!rpeglar
From: rpeglar@eta.unix.ETA.COM (Rob Peglar)
Newsgroups: comp.arch
Subject: Re: ETA-10
Message-ID: <638@woods.unix.ETA.COM>
Date: 1 Nov 88 15:23:10 GMT
Reply-To: rpeglar@woods.UUCP ()
Distribution: na
Organization: The Final Frontier
Lines: 48


It's been interesting to read the various (mis-) information on the
newsgroup about the ETA-10.  I have a mind to just kick back and enjoy,
but some people here have been urging me to respond, so here goes.

Chris Lewis believes the 10 is built with 6k gate ASIC CMOS.

Actually, it's 20k CMOS from a Very Large Company ( starts with H,
HQ in Minneapolis ).  It's ASIC in the sense that Chris describes.  
(re posting # 3472 )

Chris also mentioned cycle times (cpu).  Michael Tighe corrected him,
almost. (re 3484)

The cycle times are dependent on the method of cooling.  The CPU boards
are in fact identical in terms of arrays used;  a board is a board is a
board.  The liquid-nitrogen cooled models (E and G) are 10.5 and 7.0 ns,
respectively.  The air-cooled models (P and Q) are 24 and 19 ns,
respectively.  Michael is correct when he says the Cyber 205 is a 20 ns
cycle.

Don Lindsay, in posting 3453, reports that there is a "Great White Hope"
at ETA, concerning the ability to "...roll these things out like cookies".
Puns on boxing or weather or the Mackenzie brothers aside, Don, the fact
is that the air-cooled systems (not just cpus) are indeed coming out of
manufacturing quickly;  about 7 days per system.  Liquid-cooled systems
take longer (30-40 days for a typical 4-cpu system) due to long lead times
for cryogenic equipment and such.  As production really gears up in 1989,
these times are expected to shrink as more efficiencies appear.

Then, a brief flurry about Cray "shared memory" from Eugene Brooks out
at LLNL and Michael.  It's highly amusing to "read between the lines"
and see the effect of Real Users (more correctly, Real Users' perceptions)
on architectural entities.  BTW, Eugene, which Cray were you referring to?
I'll let the net determine if the "honest to God shared memory" is
actually of divine ordinance.

Cheers


Rob



-- 
Rob Peglar			internet:  rpeglar@bigfoot.unix.eta.com
ETA Systems, Inc. ETC03J 	uucp:      {amdahl,rutgers}!bungia!eta!rpeglar
1450 Energy Park Drive		fax:	   (612) 642-3448
St. Paul, MN  55108 		voice:     (612) 642-3197