[fa.info-vax] DEC-20 vs. VAX

info-vax (07/09/82)

>From Linnerooth@SANDIA Fri Jul  9 14:39:13 1982
This message solicits information pertaining to a comparison
between a DEC-2040 (or DEC-2060) running TOPS-20, and a
VAX 11/780 running VMS.

I have been able to get a reasonable comparison between these
systems for programs that are heavily weighted toward floating
point calculations, which hits the VAX in its strongest area.
How about tasks that involve lots if bit and byte manipulations,
sorting, and disk I/O?  Does anybody have comparison information
in this area, or in the area of "balanced" or "typical" time-
sharing?

Thanks for any tidbits you may pass back to me.

	- Tom Linnerooth -
-------

info-vax (07/10/82)

>From MILLER@SRI-NIC Fri Jul  9 19:00:07 1982
	When I was at DEC, I had the opportunity to perform
an extensive benchmark on floating point FORTRAN on a
2060 and a VAX.  Hardware config's were:

2060:	512 Kwords memory
	2 RP06 Public Structure

VAX:	2 MegaBytes memory
	1 RP06 file system

	There were five programs to be run.  They produced a
moderate amount of printouts, about 200 pages as I recall.
I cannot remember the exact timings, as this was close to
three years ago, but in four out of five cases, the 2060
was ahead of the VAX.  The best case was a 28% performance
improvement on the 2060 vrs the VAX, and the worst case was
12% slower.

	As I recall, the total benchmark ran 25% faster on the
2060.

	As for relative performance, the PDP-10 can queue up
multiple page transfers at a time.  The VAX is limited to
1 I/O operation at a time (I.e., queue, wait, queue, wait, etc.)


	I have also seen a relatively small user base bring
a moderate size VAX to it's knees.  8 users should not be
able to do this, especially since they were not doing
anything exotic.

	On the whole, if you want a batch oriented Fortran
machine, I'd say the VAX is a win.  However, if you want
good performance on batch and timesharing, plus a pretty
sexy Fortran machine to boot, I'd go with a PDP-10.

-HWM
-------