@matthey
Changing only the CPU oscillator is still overclocking the programmable logic, which seems to be a really bad idea on the MK-III/CSPPC cards - I've heard it from a few people as well, one guy had his clocked at a mere 66 MHz before it failed. The Cypress chips on the card are rated at a maximum of 125 MHz, so (big conjecture here) if stock they are already running at an internal clock of 100 MHz, a 75 MHz CPU/bus clock would put them at 150 MHz internal clock - obviously a huge gamble.
The Apollo is definitely a step down from the MK-III, I know I would want a Z3 SCSI card if I had one. It performs well though, memory benchmarks at 40 MHz being roughly equal to an MK-III at 50 MHz. If something goes wrong, chances are it's only the CPU and not the irreplaceable logic that's fried. It would be slower, but enabling the divider on the Cyberstorm cards would be much safer (I looked at the details but technically it's simply over my head, no idea if it's even a possibility outside of the MK-II).
Anyway, I'm starting to wonder if there's something other than just the clockrate that kills the logic when the cards are overclocked. I checked the datasheets for the chips on my TekMagic before bumping it up to 64 MHz, installed a Rev 6 CPU, put heatsinks on the FPGAs "just in case" and lots of extra cooling around the card. I have had ZERO issues whatsoever that could be traced back to the overclock. However, I have to admit Castellen has me a little spooked - I think I'm going to put it back to 50 MHz, simply enjoy it as it is and save the overclocking for the (easily replaced) PC hardware.
