🚀 go-pugleaf

In article <TMH.92Nov7193313@bigfoot.first.gmd.de> tmh@bigfoot.first.gmd.de (Thomas Hoberg) writes:
>In article <1992Nov3.041557.20712@ntuix.ntu.ac.sg> eoahmad@ntuix.ntu.ac.sg (Othman Ahmad) writes:
>   SPEC benchmark is no longer a typical user program. My main load now is
>   the running of X-window applications. Anyone has figures for X11 running on
>   Intel Microprocessors?
>
>   I only try it on our 386/25(no cache)  and 486/33. The figure for
>    X11perf -bigtilerect500
>   is about 2 versus 4.4 per second. The speedup is only 120% which shows that
>   the internal cache effect becomes negligible.
>	   Even then x11perf is a repetition benchmark which can make available
>   cache look more impressive than it should be.
>Be very careful when using Xperf as a benchmark on an ISA based system.
>You might end up measureing ISA throughput and unsurprising there is
>little difference there. To get some useful numbers you should use the
>pixmap/pixmap values and no window/window or pixmap/window
>window/pixmap values (although they might be interesting in an other
>context). Actually I'd be very interested to see those values for
>otherwise identical systems using the various DX, DX2 processors from
>25 to 60MHz.

I would suspect that the QUALITY of the GRAPHICS board would have even 
a larger impact on performance than the ISA bus bandwidth.  Can you
describe the GRAPHIC subsystem in the two machines that you used to 
get your CPU ratios?  The CPU only can speed up that portion of the
execution time it controls.  When you configure a system with a
bottleneck that consumes large portions of the execution time the 
processor cannot scale that fixed time with CPU performance improvement.

The Intel definition of a 80486DX processor includes the performance
improvement of 2x of a DX-25 over a 80386DX-33.  This holds true
for Intel processors on all benchmarks when the computer system disk,
memory, and graphics systems are properly configured.  Slow down anyone
of those three to where they begin to limit user performance and there 
is no reason to buy faster CPUs.  

rod

>   --
>   Othman bin Ahmad, School of EEE,
>   Nanyang Technological University, Singapore 2263.
>   Internet Email: eoahmad@ntuix.ntu.ac.sg
>   Bitnet Email: eoahmad@ntuvax.bitnet
>---
>Thomas M. Hoberg   | Internet: tmh@first.gmd.de
>1000 Berlin 41     |           tmh@cs.tu-berlin.de
>Wielandstr. 4      |
>Germany            | BITNET:   tmh@tub.bitnet 
>+49-30-851-50-21   |


Rod Skinner				I speak for myself only.
Intel Corp
3065 Bowers Avenue   MS NW1-20            PH (408) 765-4474
Santa Clara, CA 95052                     FAX (408) 765-4920