X-pop3-spooler:	POP3MAIL 2.1.0 b 4 980420 -bs-
Message-ID:	<19980709000751.A6903@math.fu-berlin.de>
Date:	Thu, 9 Jul 1998 00:07:51 +0200
From:	Felix von Leitner <leitner AT math DOT fu-berlin DOT de>
To:	beastium <beastium-list AT Desk DOT nl>
Subject:	Re: please benchmark / MMX #2
Mail-Followup-To:	beastium <beastium-list AT Desk DOT nl>
References:	<Pine DOT LNX DOT 3 DOT 96 DOT 980708220822 DOT 884A-100000 AT localhost>
Mime-Version:	1.0
X-Mailer:	Mutt 0.92.8i
In-Reply-To:	<Pine.LNX.3.96.980708220822.884A-100000@localhost>; from Jukka Vuokko at Juke's own Linux machine on Wed, Jul 08, 1998 at 10:16:27PM +0300
Sender:	Marc Lehmann <pcg AT goof DOT com>
Status:	RO
Lines:	34

Thus spake Jukka Vuokko at Juke's own Linux machine (jvuokko AT cc DOT helsinki DOT fi):
> AMD K6-2 300MHz ( 3x 100MHz) results:
> File: "povsrc.tar"
> Size: 3717120      Filetype: Regular File        

>         usr    sys  elapsed
> mmx     10.43  0.11 0:10.53
> nommx    9.76  0.11 0:09.86
> mmxonly 10.45  0.08 0:10.53

I just benchmarked the same binary on the same file, and I got about
17.2 secs user for nommx and 20.1 secs for mmx.  I have a K6-166.  This
might serve as an interesting number for people considering to upgrade.

Extrapolating linearly, I would have expected a running time for the
K6-2 of

  17.2/3*1.66 = 9.52, it delivered 9.76

The K6-2 scales almost linearly on bzip2.  Now we should find a more CPU
intensive benchmark.  How about the SSLeay "speed" benchmark?  Or maybe
a ray tracer?  Maybe we should make a real-world benchmark!  For
example, the three things where _I_ need CPU speed most are

  * mp3 encoding
  * compiling stuff
  * quake ;)

BTW: bzip2-nommx on a Pentium II 266 takes about 8.3 seconds, so it
clearly outperforms the K6-2 300.  And for the MP3 and Quake benchmarks
are FPU intensive, so the K6 would obviously not be a good choice for me
(if the saved money is not the issue).

Felix

- Raw text -