»

[k4lmp]

I recently picked up an a2000 with an A2091 scsi controller. It came up as 'defective' when the machine booted, and had problems. So I tried my GVP Impact a2000-HC, for which I opened a different thread. I wasn't able to get the GVP to work, so I started messing with the A2091 again. I figured out is was an on-board ram problem, and then discovered that the 1 meg of ram was in the wrong sockets. It was in the outermost 2 rows instead of the inner 2 rows. I put 512K in, and ran the ram test, and all was good. I stuck the additional 512K in, and the chips in U18-U21, which is the last 512K of the 1 meg, brings up many failures. I swapped the chips with the ones in the first row, which tested OK, and the problem stayed with the sockets U18-U21. I have read that the A2091 has problems unless the full 2 meg is installed, and I have some ram on the way from Mech to fill it up. It was the same errors on all of the sockets U18-U21, so I am wondering if it is just a bug. The card works fine now. Any ideas? Thanks in advance.

[danbeaver]

I guess the answer is how you define "bug?" To Grace Hopper it was a moth stuck in a relay. You describe an addressing error that sound hardware-like in origin such as a bad trace. If it recurs with Mech's RAM, then check your traces for any disrupted lines by using a multimeter. There is a program out there for checking which RAM socket, but I can't recall the name.

In the end there can be only one, Highlander! I meant to say in the end you only need a RAM buffer for data transfers on the card, and 512k is plenty. Use another source of RAM for the system if need be.

[k4lmp]

Here is where I found the info on the bug http://www.amiga.org/forums/showthread.php?t=34711 it is in around the 6th post down.