RAM address space

[Asmusr]

What was the reason for splitting the expansion RAM address space in two? It seems easier and more logical to have only one 32K block starting at >8000. Was the console originally planned to have 16K ROM?

[+mizapf]

I guess this is a consequence of the memory layout. The decoding only works in 8K blocks.

 

0000-1FFF 8K ROM = two circuits @ 4K

 

4000-5FFF 8K DSR ROM

 

6000-7FFF 8K cartridge ROM. Doesn't need the full address bus at the cartridge connector (staying below 8000).

 

Then you still need an area for memory-mapped devices, which leaves two positions: 2000 or 8000. Also, the scratch pad RAM must be put somewhere, preferably in the memory-mapped area.

 

Maybe yes, seems as if they hesitated to use the 2000 area.

 

It seems as if this layout was generally considered sub-optimal, hence the 99/8 has two modes, like the Geneve, with one mode showing the 99/4A layout, and another one where this addressing mess is cleaned up.

[+OLD CS1]

Can address space between >2000 - >3FFF be decoded anywhere?

[Asmusr]

To rephrase my question: why didn't they put all the fixed stuff at the bottom of the address space and leave the top for expansions?

[+mizapf]

I did understand your question, but I'm also wildly guessing. As it seems to me, the design was that they decided for an 8K block decoding, and this leaves 4 such blocks below 8000. One is the 8K Monitor, one is the cartridge space, another one is the peripheral block. This leaves only one 8K block for the memory-mapped devices and scratch pad RAM, but maybe, as you already suggested, they originally planned for another 8K ROM.

[Tursi]

From accidental experimentation with Extended BASIC, we learned that at some point in time there was thought about a 16k memory expansion, and I believe Karsul confirmed having read about it. It's possible the original intent was 16k and they slapped the extra 8k in there when it became cheap enough. But it's all speculation.

 

A lot of things under the hood are sloppy or opportunistic, I think this is just another one of those things.

[kl99]

There is a Peripheral Specification called RAM Expansion Specification which is referred from the Product Specification for the Home Computer System.

That referred document might add nice input to this thread. Sadly I didn't find this specification anywhere so far.

Since the Product Specification for the Home Computer System is from May 1979, it must have existed already in 1979.

 

From Timeline99:

October 1980:

"In a letter to dealers in the United Kingdom, TI announces that an Extended Basic GROM is underdevelopment for the Home Computer, but that memory expansion for the 99/4 is not planned."

January 1981:

"January 5. TI releases prototype models of Extended Basic and 32K Memory Expansion to selectedusers for testing. A year and a half after the TI-99 is announced to the world, TI finally gets aroundto beta testing optional products that some of the competition comes with as standard equipment!To make matters worse, the actual availability of these two items for the masses is still more than6 months away."

"Extended Basic and a new 32K RAM unit are announced for the 99/4."

 

I will add your question to the memory layout as interview question for the one 99/4 engineer I got in contact with.

[jens-eike]

The TI-99/4 was to have a TMS-9985 CPU, according to it's specifications, the internal RAM was at >8300. Later TI designed the TMS-9995, again with internal RAM - but why at >F000 and not at >FF00 to be contiguous with the NMI/LOAD vector at >FFFC...???

Internals about TI CPU and VDP design:

http://spatula-city.org/~im14u2c/vdp-99xx/