Subject: Re: VS3100/M76 is FAST!
To: Allison J Parent <allisonp@world.std.com>
From: Brian D Chase <bdc@world.std.com>
List: port-vax
Date: 05/30/1998 15:52:39
On Sat, 30 May 1998, Allison J Parent wrote:

> <Machine        NetBSD   dhry/s    dhry/s
> <Type           version  (no otp)  (-O2)
> <------------   -------  --------  --------  vvvvvvvvvvvvvvvvvvvvvvvvvv
> <MicroVAX II    1.3       1319      2242      .9VUPs, slow MFM disks
> <VS3100/M38     1.3E      2325      3968    ~2.7 VUPS faster SCSI disks
> <VS3100/M76     1.3      12195     20833     7.6 VUPS faster internally.

I thought the VS3100/M38 was around 3.8 VUPS, and the VS3100/M30 was the
one with the 2.7 VUPS number.  Looks like I'll have to drag out the M30
from under my bed :-)

> However the MVII still makes a better system to service terminal lines as 
> it can hold more IO.

Without a doubt.  The amount of stuff you can cram in a BA123 enclosure is
mind-boggling.  I could probably store all my socks *and* underwear in the
drive bays alone.  I'm not sure what this would do to the overall air-flow
through the VAX though.  The intelligent Q-bus disk and I/O controllers
are nice too I suppose. 

Now that I've got two systems which are similar in dhrystone performance
(386DX/40 and VS3100/M76), I'd like to pit them against one another doing
something useful.  Povray rendering immediately comes to mind as a good
way to compare FP performance.  Unfortunately I've yet to be able to get
povray to run correctly under NetBSD/vax.  It works fine on a NetBSD/i386
system that I tested, but it seg faults under NetBSD/vax.  Shortly after
it starts running, and is just starting the ray-tracing sections of code,
the process attempts to dereference a pointer to an invalid memory
location.  (In one case I found it to be trying read data at location
0x00000000). 

I am however looking forward to running compiles on the M76, it should
move along quite nicely.

-brian.
---
Brian "JARAI" Chase | http://world.std.com/~bdc/ | VAXZilla LIVES!!!