The 9900 address up to 65,536 bytes of memory

[+mizapf]

And there definitely wasn't a USR() function call.

 

Well, something similar: You can load the start address of your machine language program into the interrupt hook (83C4). That way you can even implement a QUIT from a program (CALL LOAD(-31804,0,36)).

 

I used that for some quick'n-dirty machine language programming, loading the opcodes by CALL LOAD, then starting it in the above way.

[Tursi]

That's pretty clever, Mizapf!

 

James: the analog to USR() was CALL LINK, but yeah, you only had that with a module that added CALL PEEK and CALL LOAD.

 

As for 'how much would it cost' to launch with 32k? Well, according to http://www.jcmit.com/memoryprice.htm, 32K expansion would have been between $200-$900, though this is a wide compilation of sources. I remember paying $25 per 32k chip in 1993 for my 16-bit memory expansion, so that's 14-15 years later, seems fair to me. Let's say TI used their muscle and halved the cost above, that's still adding about $100 per machine. That's why NOBODY shipped with that much RAM.

 

I can't find the street price of the 32k card when it was released (sidecar or otherwise), though the timeline (https://www.99er.net/hist2.html) notes that it was still delayed in Aug 1981, with only prototypes released. The PEB was announced with 32k card in Jan 1982, although was in short supply. That's still 3-4 years later.

 

Remember that the console was /way/ overpriced already. They had hoped to release a console for the bargain home computer market, but at launch they were priced up in the 'personal' computer market like the Apple 2. That's not where they wanted to be, but they landed there due to the design challenges (like the one that ended up dropping a 9900 and glue logic in the machine). In Jan 1983 (according to the timeline again) TI cut the console's cost to $125 to match the Vic20, and that was believed to be a break-even point. In 1983 32k of RAM would have added about $60 to the cost of the machine (per the above table). They were producing 150,000 machines a month, so that's a cost of $9M / MONTH to add the RAM, money that would have been completely lost if they still needed to price match Commodore. (And that's only about 30-60% of what it was costing at launch).

 

I think, given the other challenges, shipping the machine with more CPU RAM would only have killed it sooner. Really, it needed a proper redesign when they failed on the 8-bit design, rather than a quick hack. They went to market rather than try to fix /everything/, and in fairness it almost worked. But TI demanded a high quality machine despite the lower end market, and in the end that strategy couldn't compete with Commodore's much lower cost of manufacture. Someone else once posted an article (was it here?) about why the 99/8 was doomed even if it came out, just due to TI's massive overhead. I remember thinking that article made a lot of great points, even if the technology in the 99/8 kicked absolute butt for the day.

[Sinphaltimus]

I can't find the street price of the 32k card when it was released (sidecar or otherwise), .....

 

This might be a good hint. Taken from the assembly book I'm reading written just about when the TI was to be discontinued I believe.

 

Edited September 6, 2016 by Sinphaltimus

[JamesD]

How much more would it have cost TI, and in the end the consumer, if the machine had been equipped with 32K in the first place, and the ability to poke and peek? .... I remember my cousin telling me at the time the reason people were buying Spectrums and even Vic-20's was down to the fact that you had to buy a RAM upgrade to do anything like that.

RAM was still expensive in 79 and the first truly budget machines like the VIC20 and CoCo hadn't been introduced.

So they would probably introduce the machine in the mid Apple II price range.

I think the first magic number for sales was below $500. That probably wasn't possible until 82 with that much RAM, at least not with a profit.

It wasn't until cheaper RAM chips became available that prices really dropped and that required a board redesign for TI.

.

Even if the TI had come with only 4K of CPU RAM, they had to charge for 16K more than the competition as other machines shared RAM with graphics off of the CPU buss.

When the CoCo is at $450-$500, the TI is probably at $600- $750.

16K of CPU RAM probably would have been the the best compromise for capability vs price with an optional upgrade for more.

You can have some pretty complex BASIC programs in 16K if your graphics and text have separate RAM.

 

If TI BASIC had the ability to store some variables in video RAM under program direction, then 16K is quite a bit of RAM for code, and it's much simpler to manage than putting code in video RAM.

Variables needed for loops and calculations that require speed can be on the 16 bit buss, while string arrays or seldom used variables could lie in video RAM.

That seems reasonable to me anyway.

 

If TI had pitched it as having 80K of RAM when fully expanded, they could have kept themselves positioned above the cheapest budget machines.

64K? Hah! We have 80K and at least 64K available for software even when using the best graphics, they only have 56K, or whatever.

They may not hit the sub $200 price point that gained huge sales, but staying profitable was a heck of a lot more important.

But then I think TI also would have had to make the design a little more appealing to professionals instead of as a game/home computer.

 

[JamesD]

On price, keep in mind that TI made their own chips. They aren't paying what it would cost you for a RAM board.

[Stuart]

JamesD, on 06 Sept 2016 - 9:35 PM, said:

On price, keep in mind that TI made their own chips. They aren't paying what it would cost you for a RAM board.

 

Bear in mind though that the TI Home Computer Division would have to buy the chips from the TI Semiconductor Division, who would want a good profit out of the deal ...

[JamesD]

 

Bear in mind though that the TI Home Computer Division would have to buy the chips from the TI Semiconductor Division, who would want a good profit out of the deal ...

Because clearly, they will want to charge themselves as much as someone else so the company can make more money.

And you ignored this part of my comment "They aren't paying what it would cost you for a RAM board."

[+Ksarul]

Actually, the cost thing was a major issue when purchasing between Divisions at TI. Profit was stove piped within each Division, so even other Divisions had to buy their parts at the listed wholesale price. The Division HAD to make its profit, and if they sold the parts to another Division at actual cost, that affected their bottom line negatively (because they didn't show a profit on those sales) while inflating the bottom line for the Division receiving the parts (because they would pay less for the parts and were thus more profitable). It actually made the machines more expensive, because each Division had to make a profit on the transaction.

[Willsy]

Exactly right. ABB nearly went broke 15 or so years ago by doing that. They were booking inter-division trade as profit. It didn't occur to anyone that all they were doing was moving the same money around different internal accounts. No new money was coming into the company.

[+OLD CS1]

Ugh, I have no idea what happened to my post about this. In ninth grade economics class I did a ten-year study on Texas Instruments, considering my love and interest thereof, and found plenty of articles and information which detailed this exact practice within TI. I did not really understand how it worked or why, it just seemed non-intuitive to me at the time and I suppose it still does.