"A" Artifact- IBM intended originally to bring out a "low-end" Mod. 90 with DBA-ESDI and a 386DX-25 / DX-33 processorboard - but luckily dropped these plans. (Ed. Someone, either the US Army or a big insurance company (different rumors) had 386DX-20 complexes made for it- the infamous "Type 0"). So these DBA-connectors are "design artifacts" which can be found regularly on IBM boards and cards ... the planar descriptor file for the Mod. 90 however contains informations on board features, but these are no longer  supported from the machines firmware and BIOS. No need to bother with these ports, they are non-functional. The later series Mod. 90 no longer had the DBA-ports - only the soldering spots remained until the end of the production. 

Video RamThe VRAM chips are Toshiba TC524256BZ-10, mfd 9309 or NEC D42274V-10, mfd 9214 should work too.
   If you have a different Video ZIP make that works, send me a note
The 8590 systems have 8 sockets for video memory available. The 9590 systems come with 4 VRAM chips (512HB) soldered to the planar.

NOTE: Even though IBM Canada site sez the 9590 has XGA-2 integrated into the planar, IT DOES NOT!!! It has 512K soldered on the planar, plus 4 sockets for the 512K video memory upgrade. The 95xx just means ISO compliant. That's why the 9590 came with XGA-2 cards in them.

Differences between 8590 and 9590 Planars
 The 9590 lacks the DBA artifacts, has 512K video memory soldered on the planar, and is a pretty green. The 9590 planar is identified as an XP 90 system board under setup. It does not support Synchrostream. The parallel port is a standard one, no Expressprint.

Three Floppy Experience
  Just because people said it isn't done, I threw three 1.44MB floppies into my 9590. All three showed up under setup as 1.44MB drives. Under DOS, they are accessable as A:, B:, and D:. Under W95, it blows the mind of the IOS driver and Win95 says you must shut the system down and restart Windoze. But under safe mode, I was able to access and read off the D: floppy.
  I will eventually try it under NT 4.0 just for snorts 'n grins...

* Marked Floppiy Drives on 8590s
   Older 8590s may have their floppy controllers FRIED if you use a 2.88MB floppy that has an asterisk ( * ) on the upper surface of the eject button. The 9590 is not affected by this charming quirk. 

Loading SIMMs Onto Memory Risers
   Memory must be loaded in matched pairs (size and speed) into sockets J1+J3 and J2+J4 for interleaved configurations. (Type 1, 3, and 4 complexes). Type 2 complexes allow you to stuff SIMMs in the sockets in any order or combination, but if not in matched pairs (J1+J3, J2+J4) there will be a performance hit.
   Don't stuff one Riser with modules (especially double-sided) and leave the other blank. It *hates* imbalance on the memory drivers. Try to organize them the way to achieve a balanced load on *both* memory risers by having equal number of chips per pair, then on both risers.
   Certain releases of the Mod. 90 had problems with the double-sided SIMMs - especially with the 8MB ... I vaguely remember a sort of pamphlet from IBM dealing with this issue - but cannot remember if I'd copied it into a file or kept in in paper form and if: where I have left it. At least the problem is not listed in the ECA-database. 

Error 201
    Error code 201 says "Reseat system board memory" and can afflict the systemboard as well as the memory only. I would suggest to remove the memory risers, reseat all modules, plug them back and see if they are seated properly.
    I would also suggest that you start with one single pair of matching memory modules in the connectors J1 + J3 (Peter, pay more attention when you post!!!) on riser J11 - the one closer to the processorboard. This is just to test out if your problem is memory- or systemboard related.
    If the machine comes up fine (counts memory) - install the next pair in sockets J1 + J3 in Riser J14 - the on closer to the power supply to keep balanced load of the memory decoder lines. As I wrote: the Mod. 90 has a sensible feeling for imbalanced memory modules and may "spin out" with somewhat strange and unexplainable errors by no obvious reason. There once was a recommendation from IBM on that topic and they explicitely mentioned it for the Mod.90 - particularly for those cases where double-sided memory modules are used (which put a higher load on the decoder lines).

SIMM Holder Clips
There also was a problem with local power-drops on the early Mod. 90 memory riser cards (the ones with all-plastic SIMM-sockets). Improved versions had metal holder clips. And - logically - you should not mix the two versions.

Memory Riser Card Support Bracket
There *must* be a plastic Support Bracket clipped over the 2 (two !) memory riser boards to properly fix them. This part is called "memory riser card support bracket" and is FRU 57F3029. It also has a "bay" to guide the SCSI-cable surface wave filter (that large heavy ferro-oxide block). 

Memory Expansion Boards
You can't. Sort of. The 90 will accept an expanded memory card's memory IF there is less than 16MB on the risers. But IF the 90 has 16MB or more, then setup disables the expansion board memory. If you can cram 64MB on the risers, who the heck needs another 16MB on a card?