Everhart, Glenn
From:	Terry C. Shannon [shannon@world.std.com]
Sent:	Saturday, June 20, 1998 9:55 AM
To:	Info-VAX@Mvb.Saic.Com
Subject:	RISCy Business of TITANic Proportions
Jerry Leichter wrote:
> 
> | Yup. I think they booted TITAN somewhere around 1983. Built about 50
> | of these 12-MIPS UNIX-only boxes. TITAN never saw the light of day
> | because it wouldn't run VMS and because it was composed of
> | nonstandard-sized boards that would have been tres expensive to
> | produce.
> 
> Revisionist history.
> 
> Titan was an experiment, developed at DEC SRC.  It was a very small,
> very fast (for the time - 1983 seems early, I think it was more like 86)
> super-RISC chip.  One interesting feature I recall was that it had no
> hardware refresh for its dynamic RAM:  The OS was expected to generate
> reads to all chips often enough to make sure the memory got refreshed.
> (It did this by executing a specially crafted bit of code in a high-
> priority interrupt.)  TITAN was a true exercise in hardware minimalism,
> and had a performance way beyond anything else at the time.  I know
> there was at least one paper presented at a chip meeting describing the
> thing, so most of the details can be verified.
> 
> Titan was never intended to be a product, any more than the original IBM
> 801 was intended to be a product.  I very much doubt it had any
> provisions for floating point, which would have doomed it in the
> workstation market; and with no software, it had no hope in any other
> market.  Except ... embedded systems.  In fact, the Titan core got used
> in a number of DEC chips:  It was fast enough and small enough that you
> could build a specialized chip using a Titan plus some ROM and RAM to do
> what you'd otherwise have to do with a custom chip.  I think DEC built
> some disc controllers this way.  Of course, the Titan core itself would
> be completely invisible from the outside when used this way.
> 
> I have no idea whether Titan ever ran Unix; my guess, though, would be
> that the answer it no.  At the time it was built, DEC SRC's OS of choice
> was Topaz, which was built in Modula 2+ (a predecessor of Modula 3).
> Topaz was an OS in the spirit of experimental OS's like Pilot built at
> Xerox PARC a number of years earlier.
>                                                         -- Jerry

As we discussed offline, the TITAN "definition" I cited above came from
very well informed folks within DEC. 

Topaz (and TAOS) were associated with the Firefly SMP project, which was
publicly discussed at MIT in Summer 1986. 

Perhaps I have the codename wrong, but there was in fact a RISC box
built in the early 80s that ran Ultrix. Prototypes were built and used
at DECwest. An article on DECwest in one of the early issues of Digital
News (late summer 1986) showed a DECwest developer standing next to one
of these boxes.

Can anyone provide more information on this early RISC project?

Terry Shannon