<The VAXstation 3100/M76 runs at about 7.6VUPs which is fairly decent for
<such a small package, but it's not likely to impress anyone.  You'll

For shear size and performance the m76 is it all in a pizza box. Generally 
they are cheap to free as Ultrix is not available/supported for it.

For someone with a bit more space a 6000 series can be 1-3 40" cabs and 
offers good performance and modest power requirements for the smaller 
versions.  These are starting to turn up cheap.

Other common and generally low cost system are 3400s and 3500s and while 
only in the 2.7-3.1 vups range they offer better mass storage options than 
the older and very popular BA123 boxed MicroVAX-IIs (ka630).  All of these
can be found cheap to free as well and power needs are modest and under
600w for the biggest.

None of thosse are really physically big systems though they are all 
workhorse.  

Of the physically large systems the 11/780, 782, 785, 8200, 86xx can 
really fill a room and eat power but they don't offer a lot of VUPs
though their ability to do a lot of IOs is good.

One performace figure not noted and often hard to find (if at all) is the 
systems IO throughput.  That figure is related to the disk system and 
controller and can be independant of the VUPs available over a range.
For example the VS2000 is the same CPU as a MV-II but the MV-II will
out run the VS2k in most cases.  The reason is the RQDX3 offers just 
a bit more IO performance over the rather raw controller in the VS2k.
Moreso for using SCSI or massbus devices.  One microvax-II I have
uses two RQDX3s one per disk as the buffering offered by one controller 
per spindle offers a measurable improvement in performance expecially for 
swapping as both can concurrently be doing transfers.  With one that 
would be a largely serial operation. This reduces the "cpu waiting for 
the disk" that can make any system appear slow.

So with all of the above in mind some of the Unibus systems like the 780
at one VUP though slighly better than KA630(.9vup) offers far greater IO
and overall performance due to differences in the controllers and disks 
commonly used.  CPU speed on vaxen tends to show more when the operation
is more CPU intensive(math) rather than disk intensive (file transfers).

That characteristic is why some of the smaller and slower vaxen are still
retained in some environments.  A 750/MV-II though slow can still boot and 
file serve a bunch of machines/servers with ease if not asked to do any 
major number crunching.

Hope this helps people understand why older vaxen are still very useful.


Allison