Folks, guess what:(,

this could be coincidence but the ss5 appears to have
fried the cg6.  It was a sunGX frame buffer : 501 1672.

It ran most of the day no problem. Then I went out
of the room, came back and the whole machine was
down. Took me 15 minutes of pulling things apart
to try and work out why it wouldn't even boot any
more!   When the cg6 was removed it booted. The
centre of the main SunGX  chip is a bit browned
and trying to boot with it I noticed it was very hot.
/var/log/messages reported:  (i assume relevant)
Sep  1 19:20:18 lewy /netbsd: NMI: system interrupts:
40000000<VME=0,SBUS=0,ME>
Sep  1 19:20:18 lewy /netbsd: trap type 0x29: pc=0x196b0
npc=0x196b4 psr=4400086<S>
Sep  1 20:24:52 lewy syslogd: restart



I've got a cg3 (my last card!) in it at present and all
is going again.

The cg6 sat in my old ss2 for years with no problem
and then 1 day in the new ss5 and kaboom.

1. It could be just a coincidence that it was about to
go. At the end of it's life etc. What do you think?

2. Or it was the : "arcane incompatibility".

3. Or there is something wrong with my new ss5
that caused this to happen. The previous owner
only ever ran a serial console, I wonder if this was
through necessity, I don't know but I'll have to ask
him.

Can you guys let me know your thoughts on if you
think 1,2,3 or other please. I don't know if it's going
to happen again.

Thanks,
Paul.








"Kurt J. Lidl" wrote:

> On Fri, Aug 30, 2002 at 12:17:01PM -0400, Brian Hechinger wrote:
> > On Fri, Aug 30, 2002 at 10:39:40AM +0200, der Mouse wrote:
> > >
> > > Or rather, "almost certainly"; there may be some arcane incompatibility
> > > lurking, though I have no idea what it could be.  I've never seen
> > > anything of the sort myself, and I've slung cg6s around between at
> > > least some six or eight machines.
> >
> > to my knowledge the cg6 will work in any SBus machine, end of sentence.
>
> No.
>
> Now, wasn't that simple?
>
> Any cg6 will work in any pre-ultrasparc sbus machine.(*)  "Standard"
> cg6 cards will not work on ultrasparc based machines.  The newer
> TurboGX cards will work on both ultrasparc and older machines.
>
> This problem bit us when using the (now defunct) sparc port of
> BSD/OS 4.2 on mbus based sparcs and ultrasparcs.  Here's what the
> BSD/OS man page on "cgsix" has to say about the matter.  This
> information is operating system independent, as it merely documents
> electrical incompatibility:
>
> -- snip, snip --
>
>      There are two different kinds of cgsix cards, known variously
>      as either the ``LEGO'' or the ``GX'' and ``TGX'' (Turbo GX)
>      display cards, come in both plain and ``plus'' variants.  The
>      GX and TGX cards support a maximum display size of 1152x900.
>      The GX+ and TGX+ have additional on-board mem- ory and can
>      support displays up to 1600x1280.
>
>      The Turbo GX cards are somewhat faster than GX cards.  In
>      addition, the GX cards are electrically incompatible with the
>      Ultrasparc Sbus; only the TGX cards can be used on these
>      machines.  The following part numbers can be used to distinguish
>      GX and TGX cards:
>
>          501-1481   Double width GX, revision 1
>          501-1645   Double width GX, revision 2
>          501-1672   GX
>          501-1996   GX
>          501-2325   TGX
>          501-2922   TGX
>          501-2253   TGX+
>          501-2955   TGX+
>
> -- snip, snip --
>
> -Kurt
>
> * - subject to slot availability and power issues.  The very first
> cg6 cards were double wide and drew more power.  You cannot physically
> insert the double-wide cards in some pre-ultrasparc sparc machines.
> I don't know if the power was also an issue on any machines or not.
> There are certainly sbus card combinations that are prohibited due to
> excessive power loading from the cards.

--