Hi !

On Fri, 12 Jan 2001, Carlini, Antonio wrote:
> 	The KA675 runs the NVAX at 16ns, the KA680 at 14ns.
> 	Both have a backup cache of 128KB. The KA675 disables the
> 	NVAX on-chip VIC (but I'm fairly sure the other on-chip cache
> 	is left enabled).

So the speed might be not that different? According to the HW reference on
www.netbsd.org, KA675 (4000/400) is Mariah based and rated 16 VUPs, while the
KA680 is rated 32 VUPs. But ok, i'll see. It just came up in networked single
user mode. I'll compile a kernel now, put my KFQSA in it, and have fun with
a cool new machine... at least i hope so :)

> 	Which tech manual are you looking at? I've just looked at the
> KA675/KA680/KA690
> 	maintenance guide and it agrees with my recollection. I don't think
> any VAX 4000
> 	used a Mariah: The 100-, 400, 500- & > 500 all used NVAX. The
> 4000-300 was
> 	Rigel (IIRC), the 4000-200 was CVAX.

I only have KA660 and KA680 here. KA660 helped me a lot for this port, and
KA680 still waits for support... I think, i'm _extremely_ near. Maybe Stefan
will get something for the TM the next days, if everything works like i want
it to...

> 	One of the uVAX 3100 series used a mariah ( uV 3100-80 ??) and the
> VS4000-60
> 	also used a Mariah. The only other one that springs to mind is the
> 6000-500 (for which
> 	the chipset was designed).

Ok, the 4k60 and 3100m80 is correct in this case. And things math in this
case as i got the same strange errors, and the same interrupt vectors for
memory (0x54 and 0x60).

> 	I don't think there is much you need to do differently to support
> 	the KA675 if you already support the KA680 (except leave the VIC
> 	off obviously - I believe the console turns it off somewhat
> 	abruptly, possibly patching the NVAX ucode on the fly to
> 	do so).

680 is still missing. I encountered strange problems about... half a year ago,
but now i see new light, if the 680 is that near to the 675. I was missing
the experimental hardware for this, and the KA53 i have here is too much
different from the KA680 due to the dz at ibus instead of PR-console.

> 	I have no idea about interrupt problems.

Lots of memory failures during cold start. I think... access to nonexistent
memory locations is not what this CPU really likes like the others.

> 	Dunno - I had (possibly still have) two of those but I only ever
> 	saw one DEQRA and one LNV21. I never actually saw anyone use
> 	any of these though!

This ISDN card was in use in a MicroVAX II. Support for it would be nice, but
in my first tests, Ultrix doesn't seem to like it. Using a MicroVAX II as ISDN
dialup-router would be a nice thing to try :)

...Michael

-- 
In TV, there are bluescreens to put a faked reality behind a real played scene,
in Windows, you sometimes see the real scene, when the fakes go out for lunch.