[SunRescue] Solaris >2.5.1 on 4/600MP & 4/330 Memory Boards

[James Lockwood]

On Thu, 2 Sep 1999, Bjrn Ramqvist wrote:

> What do you mean by the old Ross' CPU modules? From the webpage on
> false.net it says clearly:
> "It can be made to work but only if it is an MP system (not the older
> CPU's)."

Solaris 2.6 and up removed support for the Cypress/Ross/Fujitsu SM100's
(Cypress sold Ross technology to Fujitsu during that time).  If you have
SuperSPARC or HyperSPARC CPU's then you can run 2.6.

> My memory reminds me of a discussion here some months ago about older
> and newer revisions of the SM100 Ross CPU modules. The older ones (rev
> 7?) didn't support MP, but the newer ones (rev 8?) do. We currently have
> a 4/600MP system with dual-SM100 rev8 CPU modules, runs SMP fine, so
> should I rely on the text in that matter?
> Although, the example on the webpage was a 4/600MP with dual-SM51's, so
> that should be a good reason to rely on the fact that SM100's isn't
> supported at all.

The web page is inaccurate.  The distinction is that 2.6 and above don't
support the SM100 of any revision.

> But then again, why should Sun drop support for CPU-modules?

Because the SM100 is a hack and wasn't shipped with any other system (also
has an MMU which isn't quite like either sun4c or other sun4m, IIRC). 
It's also incredibly slow, it makes zero sense for anyone to be using them
now.  A single SM41 will outperform both CPU's of an SM100 with ease. 

> I guess that makes sense in what people say that "the SS600 is
> considered a rackmounted SS10"...

A rackmounted SS10 with a VME interface, no DBRI, most with slower SCSI
(5MB/sec SCSI-II) controllers, and a more restrictive mbus/sbus slot
layout (second mbus slot chews up sbus space).  The memory bus is nearly
identical, down to the access timings (both use an 80ns cycle), the only
real difference is RAM form-factor.

I always wondered why Sun didn't write a driver to run IP over the VME bus
and release the SparcCluster 1 as a slew of 4/600 boards in one box.
Using SS10's and a 10mbps switch wasn't nearly as nice of an approach, and
a shared 25MB/sec VME bus would have been faster than ethernet even with
overhead (a token-passing arrangement might have been the best here to
saturate the bus).

> BTW, is there anyone out there with 4/330 Memory Boards leftover? You
> know those small, 6U VME boards that couldn't be fit anywhere else?
> Since the 4/330 is lacking VME-slots, I'd be happy to expand my 4/330
> past the 32MB RAM.

I believe you mean 3U.  I don't have any but they turn up regularly.

-James