Kan's site contains much wonderful information on the 6100. However, I'm pretty certain that the report he had of someone using 256MB SIMMs (520 MB total) in the 6100 was false. I.e., someone lied to Kan.
I've traced out the SIMM socket to memory controller connections in the 6100 and I just don't see any way that 256 MB SIMMs could be supported.
Properly wired 128 MB SIMMs do work, however, they must have their RAS pins paired properly. The 72 pin SIMM socket has four RAS lines. The SIMM socket in the 6100 ties these together in pairs. In other words, one set of two RAS lines is connected together. The remaining two RAS lines are connected together. If your 128 MB SIMM pairs the RAS lines in the fashion opposite to that which the 6100 logic board uses then they won't work.
So, e.g., if the 6100 pairs RAS0 with RAS2 (don't remember the actual scheme) and RAS1 with RAS3, and your SIMMs pair RAS0 with RAS1 and RAS2 with RAS3, then those SIMMs aren't going to work.
RAS0 = pin44
RAS1 = pin 45
RAS2 = pin 34
RAS3 = pin 33
The RAS pins are used to control which bank of the memory module is active for any transaction. This is how a 72 pin SIMM can have multiple banks on board despite the fact that it doesn't have any BANK Address pins.
In theory, each RAS pin could go to a separate bank of memory giving four banks on a 72 pin SIMM. However, in the 6100 pairs of RAS lines are tied together. Ipso facto, it is impossible for the 6100 to distinguish between more than two banks of memory on a 72 pin SIMM. Since each bank is limited to 64MB of capacity by the address lines and width of the module (12 X 2 address lines => 24 address bits => 16M addresses X 4 bytes wide = 64MB capacity) and there are only two banks possible, the maximum supportable SIMM in the 6100 must be 128 MB.