T.R | Title | User | Personal Name | Date | Lines |
---|
2280.1 | It works fine, but... | LEDS::ACCIARDI | | Sun Feb 26 1989 22:57 | 54 |
|
I've had one in my A2000 for months now. The speed is nowhere near
doubled, but my benchmarks show 10-20% speed improvement, sometimes
hitting as high as 50%. The CMI board also has a socket for a math
chip, which I also have. Since the 68000 uses the 68881 chip as as
peripheral and not a true coprocessor, the speed improvement (under the
1.3 MathIEEE lib) is not as dramatic as it would be if you had a
68020/68881 pair.
As Wes Plouf has explained elsewhere, 68000 instructions that take
more clock cycles see the most speedup. Games DO see a noticable
speedup, and most demos run much faster at 14.3 MHz.
This board is a good way to add a math chip to your system. With
1.3 released, more software will be using the newer, faster math
libraries. Any software that uses the math libraries will
automatically use the math chip, if present. CMI includes software
to install and de-install the math chip, and several programs that
demonstrate dramatic (like 2000%) speedups. You can also software
toggle the speed of the 68000. If you like, you can jumper select
to hardwire the system at high speed. This is handy for self-booting
games with no startup-sequence or any way to software select
the high speed mode. I never had a single instance of software
that would not run at 14.3 MHz.
Now for the bad news... this board DOES NOT work with any DMA
accessories installed. If you have a SCSI/DMA controller, forget it.
I had no trouble with the PA when I had a C Ltd non-DMA controller
installed, but when I switched to a fast MicroBotics board, the system
would not even boot. (MicroBotics, in their installation docs, warns
that certain unnamed 'pop into the 68000 socket' accelelerators would
not work with a DMA device anywhere on the buss).
I sent the board back to CMI for a firmware upgrade, since I (you
guys know me by now) had serial number 00003 or something. They
were aware of the problem, and have been working with Commodore
and MicroBotics to resolve the problem. The board was returned
in about 10 days. They sent an entirely new board, which actually
looked like completely new artwork. It would not work at high speed.
I tried every combination of jumper settings, but it just doesn't
fly. You can software select the high speed mode from a floppy,
but the minute the hard disk makes a single seek the system has
a massive coronary.
So, I am currently running at a leisurely 7.14 MHz. The math chip
still works fine. CMI has been most helpful, and they do feel that
the problem will eventually be resolved, but I have large doubts.
They now have a MB HardFrame to experiment with themselves. They
told me to give them a few more weeks to ponder the problem.
Since I DO plan on upgrading to an A2620 32 bit board, I was planning
on selling the CMI PA with math chip.
Ed.
|
2280.2 | Getting interested :-) | WJG::GUINEAU | | Mon Feb 27 1989 09:23 | 12 |
| Ed,
Do they supply a new 68000 that works at 14Mhz? How about a 68010? I wonder
if it [68010] will get even more performance boost.
Also, what's this thing look like for an A500? Do you think it will coexist
with the Spirit Insider memory expansion?
Since I have a 500, I guess DMA is not a concern...
John
|
2280.3 | | LEDS::ACCIARDI | | Mon Feb 27 1989 10:59 | 19 |
|
Yes, they include a 16 MHz Thomson MC68000. Unfortunately, there
doesn't seem to be any source for a 16 or even a 14 MHz 68010.
Bill Coldwell at CMI knows a guy at Motorola who was able to get
a 12 MHz 68010 to work, but he had to try about 10 different parts
to get 1 good one.
There are actually two different versions of the board (there may
be three now); an A500/2000 and an A1000 model. The board itself
is about 5" long x 3.5" wide and plugs directly into the 68000 socket.
It sticks up about 1" above the motherboard. There doesn't seem
to be much hardware on the board; just the 68000, 68881 socket,
mathc chip clock socket, some PALS and a jumper block, plus some
misc widgets.
CMI's phone number is (503) 684-9300.
Ed.
|
2280.4 | | IGETIT::ELLISM | Purring on a straight six.... | Mon Feb 27 1989 12:18 | 4 |
| If you don't want the maths chip, is there anything to stop you
from just plugging a 14Mhz 68000 straight in
Martin
|
2280.5 | Difference between capability and performance | LEVERS::PLOUFF | Semipro Semiologist | Mon Feb 27 1989 12:39 | 10 |
| > If you don't want the maths chip, is there anything to stop you
> from just plugging a 14Mhz 68000 straight in?
Assuming you mean plug the 14 MHz chip straight into the Amiga's 68000
socket... No, there's no reason you can't do this, but you will also
get no performance gain. The "14 MHz" (actually 16) refers to the
chip's capability. But to actually get the increased performance, you
must feed it a 14 MHz clock. The other circuits on the accelerator
card are there partly to supply the clock and massage some of the
processor chip signals to be compatible with the Amiga 7.1 MHz bus.
|
2280.6 | Is it worth it? | WJG::GUINEAU | | Mon Feb 27 1989 12:50 | 9 |
| That brings up a good point, Wes - the 7.14 Mhz BUS. If the CPU is running at
14Mhz, then all internal operations happen faster (like multi cycle
instructions), but is the memory (RAM/ROM) accessed at 14Mhz?
I'd like a 2500 but can't afford one just yet, so a turbo 68000 sounds
interesting. I'm beginning to wonder if I'd see much more than the 68010
gets me.
John
|
2280.7 | | LEDS::ACCIARDI | | Mon Feb 27 1989 13:25 | 7 |
|
The system roms are read at 14.3 MHz, so you see a noticable speedup
when moving windows etc.
They include a jumper to read the roms at high or low speed.
Ed.
|
2280.8 | High speed = marginal operation | LEVERS::PLOUFF | Semipro Semiologist | Mon Feb 27 1989 15:53 | 7 |
| re: .7
The reason for the jumper is... the PA card reads the ROMs faster
than their worst case speed. So, in any random machine, there's
no guarantee that the ROMs can be read at high speed reliably.
Don't know if the PA card tries to read RAM faster than the bus
spec, but the same argument applies.
|
2280.9 | | LEDS::ACCIARDI | | Tue Mar 07 1989 09:17 | 7 |
|
Spoke with CMI again yesterday. The woman claimed that they were
re-designing the entire board, a process which will take six weeks.
(Sure). They have a list of all people who are having problems
and will notify them when the redesign is complete.
Ed.
|
2280.10 | Problem with CMI | EEMELI::LINDEN | | Fri Jun 30 1989 09:14 | 10 |
| Well I've bought an CMI PA(rev 5) too and I use A2000 model A and got quite
serious problem with it. When in 14 Mhz the keyboard just sometimes
went crazy. The crystal in the keyboard was 6 Mhz and when I replaced
it to 12 Mhz crystal and since then I haven't had any problem using it.
After all I'm quite pleased to it, I've found most speed inreases on
using the paint/cad/3d programs like Photonpaint, Pixmate, S-A 4D and
Aegis Draw, speed increase on games does show quite so up, but that
might just depend on my slow phosphore monitor.
|