Quote:
Does "Anything other that a 41C body" include the "DIY5"?
The DIY4X already a high-performance 41-equivalent calculator. There would be no point to trying to put a 41CL board into a DIY4X or DIY5; it would take a huge amount of new engineering, and the result would be slower than the DIY5.
Currently the DIY4X at 14 MHz is 6-14x faster than the original 41, depending on the specific functions used, and at 28 MHz is about 10-20x faster. Doubling the clock frequency doesn't double the performance because it introduces a wait state on the flash memory, and because some overhead such as display updating does not get faster.
I haven't yet done any code optimization, and I think there will be at least 2x performance gain from that, and probably much more than that.
The DIY5 hardware will be able to run at 48 MHz, with two wait states on the flash memory. I have some plans for reducing the performance degradation caused by the wait states.
If someone wants to support engineering a *really* fast 41 replacement, I have a design for an FPGA datapath that runs more than 5000x faster than that of the 41. The idea is similar to Monte's NEWT processor in the 41CL, but is implemented as a 56-bit parallel data path rather than bit-serial. Some work is needed to finish the control portion. The actual speed would not be 5000x, just as the 41CL in Turbo50 mode isn't actually 50x faster than the original 41. It would also drain batteries much faster than the 41CL.
I designed that FPGA data path about ten years ago to satisfy my own curiosity as to how fast a Nut processor implemented in hardware could be. In an ASIC, it could easily be another factor of ten faster than the FPGA. I had no intention of using it in a product. However, once I have tooling for a DIY5 case, there is no reason why I couldn't at relatively low additional expense make a limited edition 41CMMMMM with the FPGA.
