Motherboard and Video Card selection

[Ken Gunderson]

On Wed,  1 Feb 2006 08:39:57 -0500
"Ian D. Leroux" <idleroux at MIT.EDU> wrote:

> Apologies for the threading errors, I'm stuck on webmail until I get a
> new machine and can't set the Reply-To headers right.
> 
> > I opted for a Tyan K8E board:
> > [...]
> > Sun has been basing some impressive 1U offerings on this board that
> > have been getting notice lately.  I've got mine paired with an Opteron
> > 180.  've not done any benchmarking but can subjectively report that
> > "it's just wickedly fast".
> 
> Any particular reason you went for a socket-939 opteron rather than the
> equivalent Athlon64 X2?  Nothing wrong with that, I'm just wondering if
> you know something I don't.
> 

Well, it seems to be debatable on fbsd-amd64 but I read some posts on
amd forums indicating that the Opterons were subjected to more
stringent QA than their Athalon brethern.  Per AMD's marketroid info
(for whatever it might be worth...):

<marketroid-info>
* AMD Opteron 100 Series processors with ECC unbuffered memory all
have 1MB of L2 cache.

* AMD Opteron 100 Series processors with ECC unbuffered memory are
produced on AMD Opteron processor die material and follow the same AMD
Opteron processor manufacturing process as do the 800 Series and 200
Series.

* AMD Opteron 100 Series processors with ECC unbuffered memory
undergo the same AMD Opteron processor-level testing and validation as
do the 800 Series and 200 Series.
</marketroid-info>

At the time the price diff was negligible so nugded me towards the
Opteron.  Granted some of the higher Athalons also do have 128K L1 and
1MB L2 per core as well.  Whether these units are otherwise the same as
their Opteron counterparts, as some assert, I can neither confirm nor
deny.

You may well be better off w/Athalon.  Get a much wider selection of
mainboard offerings and they'll tend to be a bit cheaper as well.  The
Tyan K8E is also in a couple of my recent servers, however, so having a
workstation based on it is one less thing I have to track for BIOS
updates, etc.

> > [...]
> > > I don't much like NVIDIA's binary-drivers-only policy, but is
[snip]
> Sadly, it looks like the ATI NDA program stops with the Radeon 9250
> chips <http://dri.freedesktop.org/wiki/ATI>.  All support for later ATI
> cards (that I know of) is reverse-engineered.  Not that I mind getting
> an older Radeon, provided I can be reasonably confident of it working
> with the mainboard.
>

Ah yes, I'd forgotten about the NDA limits.  Hate it when that
happens...  But I think bottom line is that ATI still has better
reputation for playing a bit more nicely with OS developers than
NVidia.
 
> > I think you can always opt for the i386 version of FBSD.  Might even
> > be preferred for workstation useage.
> 
> A good point, and one that I'd neglected.
> 
> Thanks for the input.  It sounds like an nForce4 mainboard is the way to
> go, if only because their ubiquity guarantees some kind of support.  As
> for a video card, I'll get an old one if I can, or else get a modern
> one and wait for nvidia to release 64-bit drivers or for X.org to
> finish reverse-engineering the ATI cards.  Hobson's Choice.  Here's to
> open-spec hardware, and a prayer for its return someday.

An older, well supported PCI gpu may be the best fallback option for 3D
at the moment.  Myself, I can live w/o 3D... 

ATI does provide AMD64 linux drivers for their recent PCI-E cards.
Don't have any clue about using them under FBSD AMD64.  Maybe someone
else can provide more enlightenment...

-- 
Best regards,

Ken Gunderson

Q: Because it reverses the logical flow of conversation.
A: Why is putting a reply at the top of the message frowned upon?