From: Karl Guttag Subject: First Set of Documents
I have the documents I sent previously in 3 sets that I was scanning while watching TV. This first set was in response to a bunch of questions Matthew had about my work on the 9900 family processors. Literally within a day of your email, I got an email from Matthew and then he responded with a list of questions about my involvement in the 9900 family. In answering those questions, it hit me that I had a lot of old documents on the 9918/9995/99000 that were sitting in boxes that some others might enjoy for historical purposes.
If you read the 99-4 History by KG MS-Word document first, it will put into context this first set of documents. Also this document gives a good bit of information on the 99/4 development from my point of view (that of one designing the 9918 and later architecting/designing 9900 compatible processors that were slated to go in future 99/x products). At the start of the 99/4A, I was just a young new hire at TI. In a year and a half, I went from a new hire out of college to the head architect of TI's next 16-bit microprocessor. I had a tremendous amount of responsibility as head architect, but not a lot of clout on the direction of the product line as a very young engineer.
Related to your hobby, at TI I hired another person Joe Zbiciak who's first computer was a 99/4A when he was about 8 years old. Joe is extremely bright and his hobby is making software games for Mattel's Intellivision emulators (see: http://spatula-city.org/~im14u2c/intv/ ); he also had one of those as a kid. Mattel was interested in the 9918 for Intellivision, but we couldn't sell it to them because the Home Computer Group was suppose to need so many 9918 of them for the 99/4 (pre A), but as it turned out the Home Computer was delayed and Mattel went with an older less capable chip.
We were about the cancel the whole 9918 program when Coleco came in with Colecovision and the TI Home Computer took off while at the same time the MSX computer in Japan (the "MS" standing for Microsoft which had a short lived joint venture in Japan) started selling big time using the 9918.
Not many people know this but Nintendo's original game system was based on a register level compatible superset clone of the 9918 done by Yamaha. Nintendo had developed games for the Colecovision, including Donkey Kong, and they moved this to their new game system.
In the early 1980's, I was a big fan of the show "Connections" by James Burke was a great program that showed (albeit loosely sometimes) how there are a series of connections between inventions and often times the person that becomes famous may have just been the one person that added one key improvement on top of a bunch of other ideas/inventions. I am a bit of a computer history buff (but less so as I go married and had children).
One document has the 9985 instruction cycle times. This is very rare and unpublished since that part never went into production. Note that we ran with divide by 2, 3, or 4 of the "crystal frequency" to get the clock cycle time on various CPUs that TI did at the time and this is a source of confusion. In these documents you have to look for the "state rate" or "machine cycle time" but sometimes documents will talk about the crystal frequency. This document will give you an idea of the performance of the 9985 relative to the 9995, for example a 16-bit register to register Add on the 9985 took 10 cycles were it took 4 on the 9995 (and the 9995 clock rate was higher). In particularly compare this document to those on the 9995.
I also came across some miscellaneous communication between the Home Computer group on the 9995 and the 9918 modifications that became the 9118. On the message I sent the "TO:" list were all in the Home Computer group.
Lastly there are some performance comparisons between the 99000 (internal TI code name was Alpha), the 9995 (which was first call the 9985A but was later changed to the 9995 which other than for byte operation took the same number of cycles for the most part), the 9900, the 8086, 68000, and Z8000. Note that we didn't even bother comparing to the 9985 as it was so poorly designed (but keep the code names straight because the 9985A became the 9995 and was totally different than the 9985 with no "A").
Also in your last message you said you hoped I got a good raise off the 99000. Actually I was pretty poorly paid through the 99000, but soon after the 99000 was finished they made me the youngest "Senior Member of Technical Staff" and more than doubled my salary in one pay raise (I also skipped over several job gades). In my first 4 years at TI, I had gone from one of 6 engineers on the 9918 (although I got do work on much of the architecture because the engineers other than Pete Macourek primarily did circuit design) to being the lead architect of two 16-bit microprocessors.