@Orjan
The mk3 was based around the mk2, used pretty much the same sized PCB, same SIMM sockets, same 3.3V regulator design, etc.
The drive plate physically touches the top of the SIMMs when it's all screwed correctly into place, so it is a tight squeeze.
You can get around the problem by using reduced height SIMMs, but these can be hard to get.
It's not good practice having metalwork so close to the edge of PCBs (like the SIMMs) as if there are tracks close to the edge of the board, there is a risk of the grounded metalwork touching signal lines.
Other considerations are the heat dissipated by the board. For example the 060 itself has around 4W in heat dissipation (specified in the Motorola datasheet), and you'll notice it uses a linear 3.3V regulator, as opposed to a switchmode one, so the regulator generates another ((5V-3.3V)*(4W/3.3V))=2W of heat at least, which has to go somewhere. That's not taking into account the additional heat generated by the CPLD devices, which I don't have exact data on, but obviously adds to the problem.
Dave Haynie also stated the A3640 had heat problems in the A3000D, further indicating the lack of cooling in the case. If you study the airflow pattern around where the CPU board fits, you can see why this is. In many cases the CPU boards can fit, but there are physical constraints and heat dissipation issues which make it a bit boaderline. Things are getting very close to manufacturers' recommended operating limits, hence why I think it JUST fits.
