Problem running 2 x DKB3128 on A4000T

[Methuselas]

X-ray wrote:
Not just a nerd, a super nerd!

*laughs*


I have *NO* comment!



*cough*

...nerd...

*cough*

:roflmao:

[LoadWB]

Someone asked earlier about the output from ShowConfig.  I'm curious about the mapping as well.  Could you post the output?

[X-ray]

Sure, I'll get on it.

[X-ray]

Here is the PIV in slot one, and three DKBs all with 96mb RAM:



Here is the PIV in slot one, 1st DKB has 128mb and the other two have 96mb:



Here is the PIV in slot one, next two DKBs have 128mb and last one has 96mb. All of a sudden, you lose one DKB!



When I put the PIV in slot two, I can have two DKBs with 128mb:



But I can't have the last board fully loaded, because it disappears entirely, or in one case it turned into 48mb

[Kin-Hell]

Have you tried all 3 DKB's with 384Mb, but leave the Picasso out?  ;-)

[Ilwrath]

Hah!  Interesting...  It looks like you found a bug with Z3 addressing.  

I'm not much of a deep-level hardware guy, but it looks like any time you have a border condition, (like having the end of the RAM board at 48000000-4FFFFFFF run right up to the one starting at 50000000) it is causing problems.  When there is extra space there, (like the board at 48000000-4FFFFFFF and then empty space until 58000000-5FFFFFFF) everything is ok.  

Moving around the PIV seems to move around where the physical addresses of the memory ends up, allowing you to get those breaks between banks.  (Or, of course, removing some RAM from the board, also gives you that break between banks, allowing the system to work.)

[X-ray]

Ja, its weird, but that makes sense.

Kin-hell, removing the PIV makes no difference.

[LoadWB]

That is odd indeed.  Not what I would expect, but then I have not made in-depth analysis of the ZIII specs, either.

What is the closest you can get to a fully populated board?  Can you put a 16MB SIMM in the fourth socket for 112MB?

I wonder if this isn't a problem with DKB's AutoConfig logic?  I would assume they thought that people would want multiple 128MB boards in their system, but perhaps didn't test it out?

(Just deleted a whole section of conjecture because the last configuration wasn't what I thought, and it blew my whole theory out of the water hehehehe)

You know, X-Ray, it just seems like you're not going to be able to use more than two.  Darnit.  Why not just send the extra my way and I'll give it a few years' testing in my 4000D :crazy:

[Kin-Hell]

@ X-ray

Was just a wild guess, tho' methinks DMA & the memory addressing is a bit fubar'd here!  :-?

*edit*

I wonder if running Pre-Boing Bag updates with different Kick files of OS3.9 would make a difference/ Even plain old vanilla WB3.1?

Worth checking out, surely! :-)

[X-ray]

That's definitely worth checking out, and I'll try it when the Miggy is all back together with the other drives in place.

[Tron2k2]

Here's the answer from Dean Brown, creator of the DKB RAM board:  There's a bug in Commodore's expansion.library.  When he made the boards he tested them by sticking a bunch of 'em in a 3000, and found this problem.  He notified Commodore, but it was too low on their to-do list.  Remember, this was when 128MB RAM was SPENDY.  C= didn't see this situation happening enough to warrant fixing it, and of course they later went under before it could be fixed.

That's how Dean explained it to me, anyway :-)

[LoadWB]

Well, shyt.  If it's a bug in the expansion.library, we should sick Piru on it!  :lol:

This might be mentioned in another thread.  I wonder if anyone's considered a new ZIII RAM board which could hold more memory (like 1GB or so.)  Would be fun for kicks and giggles.

[pyrre]

This might be mentioned in another thread. I wonder if anyone's considered a new ZIII RAM board which could hold more memory (like 1GB or so.) Would be fun for kicks and giggles.

Just because we can :-D

[Kin-Hell]

*bump*

....but are we?

Wozzon Xray, - any developments! ??

[X-ray]

I've been busy putting the Miggy back together (it was completely stripped for testing. I'm almost done, have to solder some fan wires.)
Then this weekend I can make a few more checks, but at present the last SIMM slot of the last DKB must remain empty or I lose the whole board from the RAM pool.
I also got an ATX PSU for the Miggy and found an unwanted 36gb wide SCSI drive I want to try (if it works I'll ditch one of the standard SCSI ones, probably the 1gb)
Oh, and I also built proper retaining standoffs for the Cyberstorm, and replaced some mobo screws and LED panel screws. It is somewhat of a major MOT for the old girl.

 :-)

[T3000]

Photos damn it.  We want photos.  :-D