A3000 and A4000 Code Testers Wanted!

[SpeedGeek]

Yes, I need some A3000 and A4000 code testers. Now, here's why.

The attached archive contains a small executable file called SDMACtest. This is the bug fixed code which didn't make it into RamseyConfig. The reason is because of this potential bug I found in the A4000 service manual:

Note that A5 must be high for all accesses. This is because contention with Ramsey results if not done.

I assume this is an issue with the A4000 IDE port register accesses and there is no problem with normal SCSI (SDMAC) register accesses.

If I run SDMACtest on my A3000 it reports "SDMAC found!" and safely exits. The desired result on the A4000 is to report "SDMAC not found!" and safely exit.

EDIT: SDMACtest has now been replaced by LISAtest.

The attached archive contains a small executable file called LISAtest.
If I run LISAtest on my A3000 it reports "Lisa not found!" and safely exits. The desired result on the A4000 is to report "Lisa found!" and safely exit.

The current release of RamseyConfig uses software model detection code which is reliable as long as A3000 and A4000 owners have properly built Kickstart ROMs and don't use Loadmodule (or similar tools) to install modules which don't belong on their machine.

So hardware model detection can be more reliable but also more difficult to debug (as this thread demonstrates).

Thanks to all the A3000 and A4000 code testers for your help!

[Wrangler]

@Speedgeek, I just tried this out on a few Amigas I had lying around and got mixed results:

SDMAC reported found?

AA3000+ Y
A4000D N
A3000T N
A4000T Y

As you can see, it gets the right answer for the desktop versions but the wrong ones for the tower versions. Why that should be the case for the A3000T (that I thought was electronically indistinguishable from an A3000D) I don't know.

FWIW, WhichAmiga reported Ramsey rev 13 in the two A3000s and 15 in the two A4000s

Hope that helps

[klx300r]

on my A4000D got not found.

I'm runnign OS3.2.1 with OS3.2 roms so I'm assuming LoadModule is involved & that's fine with me since I won't be changing roms

[Wrangler]

My results were on machines that LoadModules from flash or from disk. Not sure if that invalidates them. Let me know if you need more info

[SpeedGeek]

@Wrangler

Thanks again for your help. I'm glad the bug fixed code can safely exit on the A4000. The strange results from the A3000T makes no sense at all. Do you know what version of SDMAC you have in this machine?

On the matter of the A4000T the NCR SCSI registers are specified to start at a higher address than the first SDMAC register but the HD decode GAL doesn't fully decode the address and is probably terminating the cycle via DSACK1.


@klx300r

Thanks again for your help. The more reports we can get the better.


@Thread

I have some updated info. The A4000 Ramsey contention issue is with the ACR (This register only used for A3000 DMA) and certainly should be avoided when accessing the IDE port registers.

I'm sorry about the misunderstanding. SDMACtest doesn't know or care what version of Ramsey is installed, what modules have been loaded or what version of AmigaOS you are using.

I was simply trying explain how software model detection could be less reliable than hardware model detection. But the unexpected results could mean that vice-versa is true in this case.

[Wrangler]

It will be SDMAC rev 2 in the A3000T - I don't have a rev 4.

[SpeedGeek]

@Wrangler

I also have an SDMAC rev 2 in my A3000D.

@Thread

I decided to give up on SDMACtest and try LISAtest instead. So this thread has been updated accordingly.

[Wrangler]

LISAtest works correctly on my test machines:
Not found on the A3000T
Found on the A4000, A4000T and AA3000+

The AA3000+ is a bit of a hybrid, with AGA in it so does have a Lisa but it also has an SDMAC, so I'm not sure if Lisatest is giving you the answer want on that particular machine?

[SpeedGeek]

@Wrangler

Thanks again for reporting your results. I can see that you have not been following the development of RamseyConfig. I'm not really concerned with the AA3000+. I'm simply looking for a reliable hardware test which can identify either an A3000D/T or an A4000D/T.

I think LISAtest will do that quite well. If any hybrids or remakes fall though the cracks of model detection... well then the chips will fall wherever they fall.

[macce2]

LISATest found Lisa on my A4000D, mobo rev B !

[klx300r]

got Lisa found on my 4000D