TMS9900 designer bags our beloved CPU

[jbourke]

http://spectrum.ieee.org/geek-life/history/the-inside-story-of-texas-instruments-biggest-blunder-the-tms9900-microprocessor

 

Quite harsh and undeserved I think :-)

 

Cheers,

 

John

[carlsson]

I'll be the first to admit when I saw the subject, I was thinking of something along these lines. Perhaps a Christmas gift for the beloved one?

 

[apersson850]

Snyggt. Is there any that will take the console?

[MueThor]

Hello folks,

 

This article is from a former manager of the TI microprocessor group in charge of the TMS9900. The latter doesn't mean that he also was (one of) the designer of this microprocessor. So be careful, because he can possibly have euphemistically depicted in his article the events at TI concerning the development of the TMS9900. As far as I know, the performance of the TMS9900 had also been deliberately crippled by one or more design decisions to avoid that microcomputer equipped with the TMS9900 chip would become serious competitors of TI's own 8-bit microcomputers of that time.

 

 

Regards

Edited July 25, 2017 by MueThor

[apersson850]

I don't know what you mean about deliberately crippling something? What is crippled? Which microcomputer? The 99/4A? Yes, that's "crippled", partially because it was actually a 16-bit CPU with an 8-bit data bus that was supposed to be there, but it didn't get ready in time. Then if GPL is "crippling" or "saving ROM space" can be argued.

[MueThor]

Hello apersson850,

 

The performance of the TMS9900 had been deliberately crippled by one or more design decisions. Hereby the statement is referred to only the TMS9900 (=the microprocessor of the TI 99 / 4a) with regard to its internal, microelectronic design.

 

 

Regards

Edited July 26, 2017 by MueThor

[JamesD]

This is just semantics, but I don't think it's correct to say that the 9900 was intentionally crippled.
Unintentionally perhaps, but you have to look at the market before you really say that.
I think there were some bad design choices for sure.

 

Consider this...

The 9900 was introduced the same year as the Altair which is arguably the first personal computer, at least the first one that sold any significant number of machines.
Development on the 9900 probably began at least 2 years before that, and it just duplicated the functionality of their existing multi-board CPUs, so it was aimed at the mini-computer market.
I believe several other mini's had their multi-board CPUs turned into single chips at that time and TI was keeping up with the competition.
The 9900 was just a step on a roadmap for their mini-computer line.

When you think about it, the 8080 and 6800 were probably introduced after the design of the 9900 had started, and those were expensive.
The 6502 hadn't even been designed yet when 9900 development started, so no 8 bit price war had begun yet.
Nobody thought of them as something that would end up in a general purpose computer.

The thing that I wonder about, is why they wouldn't have have had a redesign aimed at personal computers finished in time for the TI-99, and why there wasn't at least a faster version they could have used by then. They had about 3 years to do it.
As long as the new CPU was opcode & register compatible, software development could have started before a prototype of the new CPU was available.
It's obvious that TI management had no idea of what to think of the personal computer market.

[JamesD]

FWIW, I'd guess memory registers were chosen because of how much board space would be required to implement the CPU using discreet logic.
Memory registers are pretty much a form of indexed addressing that uses 4 bits to identify the offset.
I would think this simplifies the ALU while giving software the use of a large number of general purpose registers..
Speed wasn't a huge deal because they weren't up against fast and cheap 8 bits.