MegaSTE 68881/2 not being seen

[Cyprian]

Thanks

[rubber_jonnie]

Just a quick note, on my MSTE Sysinfo.prg v8.34 shows my 6882, or at least an FPU:

IMG20250407094239.jpg (611.04 KiB) Viewed 1019 times

And Gembench 6.28 also reports it:

IMG20250407094542.jpg (369.06 KiB) Viewed 1019 times

[Chris O.]

Atari SFP-004 FPU

[Darklord]

Okay, scratch that...

I tried to run the fractal program above and it just bombs out to the desktop
if I pick FPU.

So I went back to DML's FPU testing software and it crashes out as well, even
though I thought I had run it successfully before.

Gembench show's a 68881 and so does SysInfo.

This is on my Mega STe, with a 68881. Oh well...

[Chris O.]

Darklord or anyone with I/O FPU 68881 please try this FPU test.
I need to replace my 68882 for 68881 as it seems to be stuck doing nothing?

README!.TXT

Code: Select all

MC68881 Test Utilities

(c) 1991 Quidnunc Software. All Rights Reserved.

YOU MAY USE, COPY, AND FREELY DISTRIBUTE THESE PROGRAMS.  JUST KEEP
ALL THREE FILES TOGETHER AND DON'T CHARGE ANYTHING FOR THEM.

These two programs perform a simple test to determine if a MC68881
FPU (Math Coprocessor) chip exists, and if so, if it is working.

These programs have only been tested on the MC68881 mapped into 
memory as a peripheral device (e.g. installed on the ICD AdSpeedST
board or on the Atari SFP004 Floating Point Coprocessor board).

To use the utilities:

Both programs perform floating point calculations for ten seconds
and then display a "speed factor" which is proportional to the number
of calculations performed.

First, run FPU_SOFT.PRG.  This program does NOT use the MC68881
regardless of whether it's installed or not.  Make a note of the
reported speed factor.

Now, run FPU_HARD.PRG.  This program WILL use the MC68881 if it is
installed.  Compare the resulting speed factor to that obtained for
FPU_SOFT.PRG:

*  If the two numbers are about the same, you either do not have the
   FPU installed or it is not being recognized.

*  If the "Hard" number is significantly larger than the "Soft" 
   number, the FPU is installed and recognized.

The ratio of the two numbers is an indicator of the speed "benefit."  This
ratio should NOT be used as an overall indicator of the FPU benefit, since
the actual benefit will depend on the efficiency of the software floating
point operations being replaced by the FPU hardware operations and the
actual operations themselves (FYI, these utilities use the sine trigonometric
function).

For your reference, the values we obtained for various configurations
were:

System: 1040STe 4MB with ICD ADSpeed STe.
        MC68881 FPU with 16 MHz oscillator.

                    Speed Factor
                   --------------
        CPU Speed   Soft    Hard
        ---------  ------  ------
          8 MHz      100     646
         16 MHz      186    1053

The hard and soft values obtained for a 520ST/1MB with no ADSpeedST
board installed (and no FPU) were both about 100 (as you would expect).

We'd be curious to know what the results are for a MC68881 FPU with
an 8 MHz oscillator...

[Chris O.]

All right I just replaced the FPU but due to other mods issues I need some time.

EDIT: Since 1989 - Upgrades people UPGRADES
https://www.youtube.com/watch?v=99DeRHisQtI

PXL_20250409_021936689.jpg (394.81 KiB) Viewed 963 times

PXL_20250403_171218465.jpg (459.23 KiB) Viewed 963 times

[dml]

This is on my Mega STe, with a 68881. Oh well...

Is this with or without running the LineF/IO-mapped FPU emulator patch mentioned earlier?

It should crash without the patch, but can work with the patch loaded. The patch is needed to catch & translate 020/030 FPU opcodes to 68000 I/O-mapped opcodes.


If it worked previously and now crashes - with the same software config - it may suggest the FPU needs re-seated in its socket. I very much doubt the IC would have failed but PLCC sockets are terrible for pins losing connections with age.

If you can get any other tests to pass though for plain I/O-mapped FPU, it suggests the emulator patch is not loaded or not working properly. Auto folder ordering or something like that.

[Darklord]

I did try the test from a "clean" boot - no ACC's, no AUTO folder programs, etc.

I did not run the FPU emulator first. I'll do that later today and report back.

Thanks.

[Darklord]

My apologies folks, I completely forgot about going through this already:

"Okay, I ran the emulator then ran FPUTEST on my Mega STe with the 68881.

Everything passed with flying colors."

(from another thread).

This was when I found out I had a bad/weak 68882 in the Pak 68/3 board on
my STacy.

Just move on...nothing to see here.

[Chris O.]

So I finally managed to run I/O mapped FPU TEST on 68881 ( Note: Doesn't work with MC68882 ? )

MC68881 Test Utilities
(c) 1991 Quidnunc Software. All Rights Reserved.

README!.TXT
For your reference, the values we obtained for various configurations
were:
System: 1040STe 4MB with ICD ADSpeed STe.
MC68881 FPU with 16 MHz oscillator.

Speed Factor
--------------
CPU Speed Soft Hard
--------- ------ ------
8 MHz 100 646
16 MHz 186 1053

My System:
System: 1040STe 4MB with ICD ADSpeed STe.
CPU @ 16 MHz -- MC68881 FPU with 20 MHz oscillator.