drew@cgofs.dec.com (Steve Drew 26-Nov-1989 2158) (11/27/89)
I bought a Supra 8up board with 2 meg installed, and stuck this
into slot 2 of my Amiga 2000. In Slot one exists my a 2090A,
the rest of the bus is empty.
I happened to of benchmarked one of my programs before
installing the supra memory, and noticed that the benchmark
ran 50% SLOWER after intalling the supra memory.
I am doing, FASTMEMFIRST, in fact if I did not do a fast
memfirst and the benchmark loaded into the internal 1 meg of
the A2000 then it ran at it's normal speed.
I wrote a small 'for' loop program todo 1 million loops. This
ran at exactly 10 secs. on the internal 1 meg, and 15 secs
on the supra external memory.
Observations:
-------------
1. I removed all drive cabling from 2090A .. Still 15 Secs.
2. If I removed the 2090A, the supra memory ran at 10 secs.
3. If I put the Supra board in Slot 1 and 2090A in slot 2
then after one warm reboot, then the supramem runs fast.
(***) 4. If I put the supraboard in slot 3 and 2090A in slot 1 then
the supra works fast (10secs). Thus leaving one empty
space between the modules.
Summary,
--------
Of course the solution was '4.' above. I called Supra and
they we're interested in my findings and said they will get
back to me. (still waiting >1 week).
If I had'nt of benchmarked this I would of never noticed this
problem. I wonder if anyone else has this problem and does'nt
know it????
Anybody got any ideas?
/Steve
----------------------------------------------------------------------------
Steve Drew at ENET: CGFSV1::DREW
ARPA: drew%cgofs.dec.com@decwrl.dec.com
USENET: {decvax!decwrl}!cgofs.dec.com!drew
---------------------------------------------------------------------------rsingh1@dahlia.waterloo.edu (11/28/89)
Not much to do with your original message, but I too have a question
about benchmarks.
I'm running 2 megs, 1.5 on an insider board on an A1000. With a 68010.
I have run 2 benchmark utilities. One is called 'checkspeed', and one
is called 'si' (a cramden utility).
Well, here's what one reports:
Amiga System Information 1.0
Gerald Brandt 21-Jul-88
EXEC reports a 68010 CPU
Available memory is 746 KBytes
CPU Timings
Copy Blocks in CHIP memory is 412 ticks vs 621 150 %
Copy Blocks in FAST memory is 388 ticks vs 621 160 %
Move from register to CHIP memory is 503 ticks vs 855 169 %
Move from register to FAST memory is 457 ticks vs 855 187 %
Move from data reg to data reg is 457 ticks vs 476 104 %
Unsigned divide in data register is 1566 ticks vs 1812 115 %
Unsigned multiply in data register is 823 ticks vs 856 104 %
Unsigned multiply in CHIP memory is 992 ticks vs 911 91 %
Unsigned multiply in FAST memory is 964 ticks vs 911 94 %
Stack manipulation ( MOVEM.L ) is 620 ticks vs 598 96 %
Total percentage speed: 127 %
(The checkspeed output). The si program gives my rating as 1.2 times
a standard A1000.
MY QUESTION IS THIS:
How much of this rather 'special' increase comes from my fast memory,
and how much from the 68010.
If someone has an insider, or just a 68010, I could mail you the
utilities, if you could perform the benchmark, and get back to me.
Later