#Slow Amigas
6 messages in this thread
We just got a new shipment of Amigas into our shop, and while setting up and
testing them we noticed something very peculiar. As part of the test
procedure, we ran Jez San's "Speed" program, a basic benchmark program. When we
ran it on the new machines, we got this result:
SPEED 2.00 – Amiga Performance Analyser by Jez San (c)1989 Argonaut Software.
The Amiga EXEC reckons that a 68000 processor is installed in this system.
Analysing clockspeed… Took 483 ticks, which is 3.58 Mhz +-2% (fig A)
Computing 10,000 SlightlyMoistSmallRocks in… 1391 ticks.
0.50 times (fig B) the speed of a normal Amiga at 7.159 Mhz +-2%. 0.16 times
the speed of an A2620 at 14.318 Mhz +-2%.
I've seen some slowdown in screen updates after putting in the 1 Meg Agnus,
I'd be interested to see the benchmarks with fast ram (didn't the 2620 have
ram? oh well!) Other than that, try a different monitor/ monitor cable to
see if the Agnus is being tricked into genlock mode. If this is the case,
it should run full speed after resetting the display to interlace in
preferences. I vaguely remember that the machine slows to 1/2 speed if not
in interlace mode when the Agnus thinks it is in genlock mode.
Yes, the A2620 comes with 2 meg of fast ram. We tried both a regular video
cable and a FlickerFixer (with its own cable). The results were identical.
The only thing in common is the monitor, which is a Mitsubishi DiamondScan,
AUM-1341. If you think this is it, we'd even be able to swap this. I'll
try resetting the WB to interlace from prefs, and run the speed test
again.
If you are willing to take the bad machine apart, check the mask of the
68000 chip. If it is Moto 0B26, then change it. I've heard that the 8520s
are sensitive to the glitches out of that mask 68000. 'Speed' relys on the
8520s to time operations, so that could also be a problem. My dealer opened
up a new A500 and lo and behold _IT_ had a 0B26 mask 68000 in it. I would
worry that someday, they might have xpansion peripherals that might be
sensitive to the extra noise! PS I ran speed on my A2000 and it crashed. I
located the problem to bad 8520s (I had suspected them bad for some time as
my system clock didn't increment properly {the battery backed clock kept
time, but not the system!})
We checked both machines that were acting "slowly", and sure enough both
had the 0B26 mask on their 68000's. Do you know if this is a bug with the
68000 or is it an 8520 problem? Something I don't understand is that if
this is a 68000 difficulty, why do we continue to see half speed operation
when the A2620 is installed? Could you please explain how this could
happen?
The problem is with 2 volt spikes on each transition of most signals of the
68000; The 'E' clock runs all the time, including when in 68020 mode; it is
the only 68000 signal not to be tristated or be an input. The 8520s are
clocking on each spike, twice as fast as they normally would. The timers
inside the 8520s are seeing twice as many clocks per second. This means
that the 'Speed2' program would think the Amiga is twice as slow; it thinks
twice as much time has expired as what actually did!
Solution: replace the 0B26 mask 68000 with another mask version 68000, a
Signetics 68000, or a Moto 68010 and the problem will go away.
An alternate method is to use a small (47pf?) capacator on the 'E' line to
ground to knock down the amplitude of the spike to acceptable limits.
However, other lines may/will cause expansion bus problems. Chuck the 0B26
68000!