There was a late version of 9900 that was done in I think bipolar or some strange combo process that was many times faster than the earlier machine be they NMOS or TTL. TI was not a MOS house for the most part and were ahead by doing like PDP-8 and PDP11 putting their earlier TTL machine on a chip. What they did wrong was to lag severely in marketing and advancing the technology. The 9900 was slow becuase at the time 2102s (fast ones were 400ns) were slow. By 4 years later ram would be down to 45NS (2147 and 2167 as examples). If the 9900 were to ramp up the clock as did the 8085 and Z80 by 1980 it would have gone from 2mhz to around 6-8mhz and that speed increase would have made it from a pure performance standpoint, fast. The speed would ahve been aciveable as the total transistor count on the die was far lower than most (few registers) and the complexity was fairly low. The sad story is they didn't. The TI99/4a was a sad detour that really didn't show off the CPU but did embody some interesing ideas. Grom was one. Registers were costly when a FlipFlop was an entire board. Then they started to get two in a 16legged chip the cost was way lower and the advantage of having more registers was nowhere near as costly. Also PDP-5 was mid 60s and and we had PDP-10, PDP-11, CDC monsters UNIVAC 1180s and VAX ahead of us at that date. Your thinking pdp6 and earlier. Allison