In message <43E90992.7000801@2005.smokva.net>, Petar Bogdanovic writes:

>Jonathan Stone wrote:
>> 2. Single-core CPU only. I expect you'll have insurmountable problems
>> (lost interrupts, hangs requiring hard powercycle) if you try to use a
>> dual-core CPU with nForce4 and NetBSD.
>
>Now I'm a little bit confused. A few months ago, I was very certain, 
>that the S2865 is Athlon & Athlon X2 (dual-core) -ready. Now I read on 
>the Tyan-page, that S2865 is only designed for single-core Opterons.
>
>However, there is the S2866, which seems to be that what I am looking for.
>
>
>Alan: AFAIR my last 'Tyan-search' on mail-index, I found more than just 
>this thread. Maybe no recommendations but a lot of Tyan-praising..

Peter,  let me try to rephrase that.

With any of a half-dozen multi-core amd64 nforce4 motherboards, from 3
or 4 different vendors, and NetBSD-3.0_RC3 through 3-stable, I haven't
been able t get a stable system with the hardware (nforce4 or add-in)
I need.  My experience is that NetBSD just plain doesn't work properly
with nforce4 and multiple CPUs.

To be fair, if one carefully uses only a subset of the nforce4
functionality with multiple CPU cores (either dual-core CPU or two
single-core CPUs in a dual-socket-940 board), you *can* find stable
subsets; but in my experience it's not worth trying.

In particular, anytime I start an X11 server, I get a double-fault, or
with ddb configured into my kernel, a spontaenous reboot (triple
fault?).  I've only gotten X servers to work with NetBSD+nforce4 if I
disable "APIC mode" in the BIOS; and the socket-939 BIOSes I've seen
do not allow one to disable APIC mode in SMP configs.

Ditto for various builtin nforce4 hardware, or other add-in cards.
OTOH, I'd be very glad to hear others' success stories with
nforce4+SMP+NetBSD; even more so if I could reproduce them.