Subject: Re: VIA VP2 chipset
To: None <current-users@NetBSD.ORG>
From: Greg A. Woods <>
List: current-users
Date: 02/05/1998 00:01:08
[ On , February 2, 1998 at 09:46:32 (-0800), Wolfgang Rupprecht wrote: ]
> Subject: Re: VIA VP2 chipset
> Not all of the required operations to do ECC are in parallel with
> other operations.  Turning on ECC on the Intel chipsets, for example,
> will slow things down a bit.

Yet another reason to avoid Intel!  ;-)

> The question is, is the warm fuzzy worth it to you?

I not only want the warm fuzzy I'm willing to pay for not only the extra
clock tick, but the premium price for the chipsets, CPUs, and the extra
bits of storage for the ECC bits.

If you had not had parity detection on your Sparc-SLC what do you thing
you would have blamed the occassional strange behaviour or bad results
on?  With parity error detection you knew instantly that you had a
marginal chip on one SIMM and that replacing the offending device fixed
the error and probably gave you extra warm fuzzies about the health of
your hardware.

In fact I want not only ECC (even if it never reports a blip in years of
operation), but I also want to have my OS spend a few extra cycles
performing integrity checks on some of the things it's doing every once
in a while.

Ruling out hardware and OS problems has been a major pain in the ass for
me between the time I first started using reliable computers and the
time I finally got back to depending on them.

(Of course my "trusty" old sun3 has turned out to be not so trusty now
that it's got 128MB RAM in it and I'm extremely frustrated with the
strange errors and crashes it now suffers despite having full ECC (I
think it's a CPU motherboard design fault).  When I finally get around
to replacing it again with something faster I also demand the very
highest degree of reliability I can afford without going to something
like a fully fault tolerant system.)

							Greg A. Woods

+1 416 443-1734			VE3TCP			robohack!woods
Planix, Inc. <>; Secrets Of The Weird <>