I am curious to discuss possibilities, theoretical and practical, about what could be changed/improved while staying with the 99/4A mainboard.
And second what are the possibilities of a new Tms based Computers.
We have seen the efforts that went into 99/8 and 99/2, the 99/4B, the 99/5, the 990 mini computers, the geneve, the 9995 breadboard by Stuart, the SGCPU by Snug, the TMS99105 fpga based computer by Erik.
I am wondering on what could be achieved, if we are free to redesign everything around the CPU family.
As far as I read it, TI did decide for many compromises when designing the 99/4A. We only have an 8 bit bus to the PEB. Most memory and devices are connected via 8 bit.
Many Cpu signals are not passed onto the I/O Bus, or the signal is inverted.
Afaik all the Cpus have an address range limitation of 64KByte. There are memory mapping possibilities that could bring us 16 (like the 99/8) or 64 MByte (like planned on the geneve 2).
What would the best TI computer have as specs and can we create it?
This also goes a bit into the direction when does it stop feeling special to all of us, when does it stop feeling like a TI.
So some questions I can think of to start:
Which Cpu should we pick? Will it be Fpga based to be able to overcome limitations that even the best Cpus have?
Can we do a full 16 bit data bus and design a PEB using that or could we even use the existing PEB for a 16 bit data bus by switching to another interface card?
How do we deal with the Grom approach used by most 99/4 software? We could only reuse the Grom0-2 chips from the 99/4A in higher numbers, but that would introduce dependencies to the grom system software part of the OS.
Would the F18A be the video chip we want? Afaik the only limitation in comparison to the 9938 and 9958 is the video ram size.
What would be the advantages/disadvantages or the VDP to be using Cpu Ram (like the 99/2).
How would the memory map look like? I always feel a bit disappointed seeing on the 99/4A so much memory space is permanently assigned to Cartridge space, Console Rom, DSR Rom and the user (or an user program) can not use all 64k memory for a certain program.
Can we copy the system rom from an Eprom chip into the RAM when the computer is turned on? Or is there a nicer approach?
Are CRU based DSR the thing to go for? Or is there something smarter already?
To also discuss the 99/4A topic:
I am thinking on things like replacing the System ROM by fixed versions. Providing some Grom 0-2 replacement pluggable circuit. I was amazed by Matthew on his F18A. He stayed compatible to the mainboard while suddenly providing the 16K of Video Ram from his FPGA board, providing additional 2K of ram and providing direct access to the registers and video ram for the virtual Co Processor, a Tms9900. Or think of the solution that Erik Piehls showed where we attached his device to the I/O port and simulating a 640KByte cartridge (i think it was the XB2.7 suite) with having zero at the cartridge port and with having zero mods done to the mainboard.
This intro might sound a bit fuzzy but so seems to by my knowledge on this hardware-dominated topic.