This topic brings me back to a very old concept that is likely more “possible “ than burning new masks and fabbing modern dies.
I was always told that folks would remove all the dram and use SRAM (cache) in place of dram on old PC XT systems to gain a bit of performance, obviously you can’t put sram in dram dipp sockets.
Even though I historically knew a person with a system setup that way I cannot find any evidence of how it was done.
Fast forward, I own a Tandy 1000RLX, it’s purposely crippled to block more than 1mb of ram and only has an 8 bit slot, despite being a 286.
Megabytes of obsolete but faster than needed Cache memory is easily available in this modern era, sometimes as a single chip.
How would I build a little board that includes 4mb of “memory “ (sram) and a socket for my 286 that would fit in place of my existing 286 in the processor socket?
Can I just wire the cache to the address and data pins off the cpu?
Any voltage/ground/chip select would need to be sorted but what am I missing?
There are several systems I have encountered that have very limited motherboard ram address limits that could be bypassed if such a setup were possible.
Given the motherboard can’t set the memory size properly I’m not sure how to get dos to see the extra memory even if I can manually address it via special utility.
My final thought is that I could potentially use a faster crystal on the daughter board to run an identical but higher clocked CPU on a little board like this up to the zero wait state limit of the cache while running async to the system clock.
Let me know your thoughts