--AhhlLboLdkugWU4S
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Sun, Feb 13, 2005 at 12:44:53AM -0500, Dave McGuire wrote:
> On Feb 12, 2005, at 4:21 PM, Sean Davis wrote:
> >>-mcpu=3Dultrasparc schedules the code such that the=20
> >>even-more-superscalar
> >>nature of the ultrasparc can be used.  it doesn't really allow the use
> >>of raw 64 bit arith...
> >
> >See, this is one of the oddities of sparc64 that makes me want to put
> >SpamAssassin back on it - with 2MB cache, 333mhz CPU, it would eat=20
> >through
> >messages about half as fast as a 256KB cache 2.167ghz CPU (athlon xp) -
> >shouldn't the performance margin be a LOT bigger there?
>=20
>   Well, it should be if we were comparing apples with apples, but we're=
=20
> not.  There's nothing particularly speedy about the x86 architecture. =20
> The only reason those guys get any performance at all out of their=20
> systems is the amazing process technology which results in huge clock=20
> rates.  It's still a mid-1970s processor architecture.

Aye. That's exactly why I want the sparc64 to be my SA box - I just don't
have enough RAM to handle the newer versions of SpamAssassin.

> >It does spamassassin at over twice the speed of a p3-650 with the same=
=20
> >ram,
> >for example.
>=20
>   Not entirely surprising. :-)  This is precisely why UltraSPARCs do=20
> all the heavy lifting on my networks.  Well, that and the reliability.

Agreed. That's why I'm eager to find a task for that machine, if I can't do
SA on it.

- Sean

--
 _
( ) ASCII Ribbon Campaign
 X
/ \ For Plain Text Email

--AhhlLboLdkugWU4S
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (NetBSD)

iD8DBQFCDz8MncZv+931Zb8RAp7dAKDaPf4ojc6Hjxqo+2G1lEjf4S9nIQCdHrZt
7x/3f/fdOGeFS4G80DZ+DyE=
=Bmlr
-----END PGP SIGNATURE-----

--AhhlLboLdkugWU4S--