>  >The 486/66 16M would blow the pants off a sun4/110.

>  my 486/33 16M (with EISA BusTek SCSI) is significantly
>  faster then my SparcStation IPC with 36M and two SCSI buses.

>Right. but my 486/33 8M IDE is considerably slower than my IPX NetBSD 0.9
>and probably than an IPC or a sun4/110.
>memory and disk i.o. is the limiting factor, memory first.
>If you can avoid swapping, its not hard to be faster than early/small SPARC.

>I doubt if this holds true for DS3100/DS5000, the MIPS cpu and TURBOchannel
>seem a better combination than SPARC/SBUS.

My experience is that a 486 of average speed is roughly in the same
speed category as a small DS5000 (like a Personal DECstation 5000/25
(25MHz R3x00), and faster than a DS3100 in raw integer number
crunching.  Throughput on a heavily loaded system seems a lot smoother
on the Turbochannel DECstations than on the PC.  And even the DS3100
will pull away from a 486DX/2 when it comes to doing really intense
floating point crunching.  On the other hand, a good S3 or ATI
accelerated video board will make X just as satisfying on the PC as on
the PDS5000.

So, a medium to high-speed 486 is a good match for a low end
DECstation.

(For your info, a DECstation 5000/260 [60MHz R4400] is roughly 9
*times* the speed of my 33MHz 486DLC in a small suite of benchmarks I
ran.  A small Alpha [3000/300] was about 10 *times* faster.  Whew...
talk about smokin'... ;-)

				--Michael

-----------------------------------------------------------------------------
 Michael L. VanLoon                 Iowa State University Computation Center
    michaelv@iastate.edu                    Project Vincent Systems Staff
  Free your mind and your machine -- NetBSD free Un*x for PC/Mac/Amiga/etc.
-----------------------------------------------------------------------------

------------------------------------------------------------------------------