I think it is probably impossible to benchmark machines of different
architectures and come up with anything useful.
Even back in the z80 dayz, machines were incomperable. A C=64, an Atari
8-bit, and an Apple II might all score about 52 drystones, but both the
Commodore and the Atari were much faster at graphical tasks than the Apple
because they had dedicated coprocessors for the purpose. PC hardware is
not nearly so intimately connected as those old machines were. The AMIGA
still does pretty well at video today. More than ten years after the Amiga
hit the market, the PC is still struggling to realize desktop video
editing.
The only way to get any cross-platform numbers is to go for the lowest
common denominator because it is impossible to predict how the
hypothetical target machine is architected. Something like
Drystones/whetstones, Flops, mbits_at_sec for hard disks, mpixels_at_sec for
video.
This has been a particularly thorny issue for me as I'm a cross-platform
guy who mainly supports PC systems - most run Windos of some variety.
Customers go out and get the latest benchmark and expect it to tell them
something useful. - Oh, and install 80mb of highly intrusive 'utilities'
guaranteed to trash your registry.
There actually are very good tools out there in use by processor
manufacturers and chipset makers. They use metrics based on good science
and observation which is more than I can say for most tests available to
us. For them though, the tests and results are mainly bounced off a
theoretical model of the product being worked on. I am sure Intel does
not use SpecInt or whatever internally.....
A machine testing its self is pretty fishy to me. Operating systems being
involved. The only time I ever trusted a benchmark was "CHECKIT" for DOS.
You'd boot to a prompt and run the thing. It was pretty trusty but I
couldn't share my results with an Alpha owner.
Oops, rambling. TTYL
Jeff
In <5.0.0.25.2.20001115154035.026a5008_at_208.226.86.10>, on 11/15/00
at 08:26 PM, Chuck McManis <cmcmanis_at_mcmanis.com> said:
A PIII with 256MB of ram, 128K of cache and 40GB of disk does >not seem
"balanced." In this case the processor seems under powered.
>I thought at one time that it related to bus bandwidth and processor
>bandwidth (how long a processor would take to execute ram filled with
>NOPS) that kind of stuff but I've never been able to quantify it.
>--Chuck
--
-----------------------------------------------------------
Jeffrey S. Worley
President
Complete Computer Services, Inc.
30 Greenwood Rd.
Asheville, NC 28803
828-277-5959
Visit our website at HTTP://www.Real-Techs.com
THETechnoid_at_home.com
-----------------------------------------------------------
Received on Wed Nov 15 2000 - 19:26:12 GMT