Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!philabs!cmcl2!harvard!bu-cs!bzs
From: bzs@bu-cs.UUCP (Barry Shein)
Newsgroups: net.arch,net.micro.att
Subject: Re: AT&T MIPS claim
Message-ID: <891@bu-cs.UUCP>
Date: Fri, 4-Jul-86 18:09:06 EDT
Article-I.D.: bu-cs.891
Posted: Fri Jul  4 18:09:06 1986
Date-Received: Sat, 5-Jul-86 07:14:03 EDT
Organization: Boston Univ Comp. Sci.
Lines: 31
Xref: linus net.arch:3340 net.micro.att:1366


>I think you're both right. We I benchmarked the Sun3 for engineering
>type loads it was only a bit faster than the 11/780, certainly less
>than 2:1. When I tested for text processing numbers were in the 2.5-3.2
>range, and software development values were mostly between 2 and 3. My
>impression is that the Dhrystone is more representative of an
>engineering load.
>	-bill davidsen

In the first place, the Dhrystone I have shows the VAX/780
running 4.3bsd as 1662 and a SUN3/160 as 3764 so I'm not sure what
the last sentence means in relation to the second sentence. The
uVax-II got 1612, so 2-3 times is a roughly accurate statement
(2.26 and 2.33 is more accurate, for this benchmark.)

The "roughly equivalent to a 780" is probably due to a floating point
mix. Perhaps you had an FPA on your 780 and only a 68881 on the SUN3
(standard, except the 3/50.)

I think if you added the SUN/FPA you would see it go back up to where
you expect it (roughly 2-3 times a 780/FPA.)

Also, for "engineering workloads" make sure the disk (flavor) and
memory (amount) are roughly equivalent with the 780 you are comparing
to unless you have good reason to believe this isn't accounting for
the difference. Wouldn't be fair to thrash a SUN3 to a SCSI in 4MB
against an 8MB/RA81 VAX, would it? I suspect tho that the FPA explains it.

I think your results should be better.

	-Barry Shein, Boston University