Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 23 Jun 2005 18:18:56 -0400
From:      Martin Cracauer <cracauer@cons.org>
To:        Jose M Rodriguez <josemi@freebsd.jazztel.es>
Cc:        freebsd-amd64@freebsd.org
Subject:   Re: Athlon64 board with ECC support?
Message-ID:  <20050623181856.A67269@cons.org>
In-Reply-To: <200506132102.56346.josemi@redesjm.local>; from josemi@freebsd.jazztel.es on Mon, Jun 13, 2005 at 09:02:55PM %2B0200
References:  <200506131616.j5DGGDfr067534@lurza.secnetix.de> <200506132038.25975.josemi@redesjm.local> <200506131147.50300.peter@wemm.org> <200506132102.56346.josemi@redesjm.local>

next in thread | previous in thread | raw e-mail | index | archive | help
Jose M Rodriguez wrote on Mon, Jun 13, 2005 at 09:02:55PM +0200: 
> El Lunes, 13 de Junio de 2005 20:47, Peter Wemm escribió:
> > On Monday 13 June 2005 11:38 am, Jose M Rodriguez wrote:
> > > El Lunes, 13 de Junio de 2005 18:16, Oliver Fromme escribió:
> > > > Hi,
> > > >
> > > > I'm currently evaluating possibilities to upgrade to a
> > > > 64bit system (preferably AMD).  I would like to get a
> > > > single-processor Athlon64 system, no Opteron, because of
> > > > heat, noise and power consumption (and price).
> > > >
> > > > Furthermore, I would like to have ECC RAM.  However, it
> > > > seems that this requirement is not easy to meet.
> > > >
> > > > So far, Google told me that the Athlon64 and the socket939
> > > > basically support ECC.  However, it also requires support
> > > > in the chipset and in the BIOS.  I've looked at a few
> > > > random socket939 board specs, and all of them allow the
> > > > use of ECC memory, _but_ they don't support using it for
> > > > actual error correction, i.e. they treat 72bit DIMMs like
> > > > 64bit DIMMs and ignore the ECC part.  This is not what I
> > > > want, of course.
> > > >
> > > > Now my question is:  Are there any Athlon64 (s939) boards
> > > > that really support ECC RAM?  Any recommendations?
> > >
> > > As far I know, only nvidia nforce4 have support for this.  But this
> > > may get you into problems with lan and disk (SATA).
> >
> > The system chipset has nothing to do with ECC support.  Unlike on
> > Intel systems, memory is connected to the CPU, not the chipset.  The
> > chipset (nforce vs via vs whatever) has no say in the matter.
> 
> Well, I only see this in new nvidia nforce4 based boards.  I also think 
> this is more a bios problem.

My NVidia NForce 3 250GB board (socket 754) did support ECC
(unbuffered).

I had way too much trouble with that board and got a via-based socket
754 board - which doesn't appear to support ECC.

While the chipset might not be able to mess with it the mainboard can
certainly omit lanes required.

And I suppose the BIOS needs to support it, too, although it is not
clear to me how ECC exceptions are supposed to be routed anyway.
Clearly some chip not being CPU or RAM needs to have a say in the
exception delivery? Anybody understands how this works?

I am switching to socket 940 with AMD chipset now, I'm fed up with
that NVidia mess and toy chipsets not supporting ECC (that Via-based
board that I have, Abit K8V-pro, I have also seems to have omitted any
hardware for temperature/fan monitoring).

Martin
-- 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martin Cracauer <cracauer@cons.org>   http://www.cons.org/cracauer/
 No warranty.    This email is probably produced by one of my cats 
 stepping on the keys. No, I don't have an infinite number of cats.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050623181856.A67269>