Jump to content

Are there accelerator cards for the TI-99/4A?


Recommended Posts

Well after I get down releasing RXB 2024 that has more assembly replacing almost all VDP assess my next goal is RXB with no Floating Point in interpeture.

Yea this mean many XB programs that require Floating Point will not work, but the overall speed of XB will be 3 times faster at least in execution of programs.

There will still be access to Floating Point but using it will slow the XB programs.

  • Like 1
Link to comment
Share on other sites

On 11/25/2023 at 7:25 AM, Vorticon said:

that is where humans come in as we are better able to extrapolate based on limited data and non-specific clinical findings

Translation: we have imagination and the ability to "think outside the box."  Remember, kids, there are two kinds of people in this world: those who can make an appropriate conclusion based upon incomplete data.  V'ger approves this message.

  • Haha 7
Link to comment
Share on other sites

On 11/26/2023 at 5:04 AM, senior_falcon said:

I think ChatGPT is trying to say this: The TMS9900 was clocked at 3MHz but the internal architecture of the processor requires about 10 clock cycles to process each instruction. (I think the actual number is somewhat higher)

Of course, this was not done intentionally. It is just the nature of the 9900 and so it would not be accurate to say it was crippled.

Well, it was done intentionally. It's one way of designing a CPU. If you can make more things happen in parallel, then the number of cycles can be reduced. That's a method used in the TMS 9995, which doesn't use as many internal states to achieve the same thing.

10 clocks is about the least an instruction will need with the TMS 9900. More common is 12 to 14 and if you use more complex addressing modes and such, the number increases.

Still, a good habit with the TMS 9900 is to use all advanced addressing modes as much as you can, since a single instruction with advanced addressing modes is faster than a couple of simpler instructions to do the same thing. Contemporary 8-bit architectures may for example require you to read a value from memory to a register, increment the register and store the value back. With the TMS 9900 you just incremented at the memory address in one instruction and suddenly you were done in about the same time. Especially if you actually needed to increment a 16-bit value.


The fact that the TMS 9995 is about four times as fast is the reason why some people claim it's clocked at 12 MHz vs. 3 MHz for the TMS 9900. But that's not true. The clock cycle is 3 MHz for both. The TMS 9995 just gets more things done in one cycle. Since the TMS 9900 uses a four-phase clock, the crystal is typically 48 MHz, but that is then divided down to generate accurate timing for the four different time-displaced 3 MHz clock cycles it uses internally.


As for ChatGPT in general, I sometimes get the feeling that it's repeating the principle described by Sefton Delmer in his book Black Boomerang: Cover, cover, dirt, cover, dirt.

  • Like 2
Link to comment
Share on other sites

On 11/26/2023 at 5:57 PM, jedimatt42 said:

I will assume from here on out that thread titles are meaningless. 

Titles often do not necessarily imply content (much to my chagrin.)  By implicating ChatGPT, as with any other source, the topic itself was quickly handled and put to bed while the source continues to be roundly thrashed.  We would do the same if it were Computer Shopper, Wikipedia, or Leo down at the Best Buy.

  • Like 2
Link to comment
Share on other sites

On 11/25/2023 at 12:05 PM, Ksarul said:

On the 99105 accelerator, I think that @Gary from OPA may be able to shed some light on the subject. ISTR that the major issues were related to bus timing, especially as related to the other peripherals in the system.

I was able to get a 99105 accelerator working on TI99/4a console, it is not easy, as you have to demux the 99105 bus, and then of course the ti99 converts it back to 8bit, timing is critical, what made it work on the prototypes I designed was "wire length" the distance between the CPU itself, and the 74 logic to demux the bus and the connection to the old 9900 cpu pins, length of the wires is very critical, too long timing is off, and you end up missing cycles and adding more wait states then its needed, or worse lose data.


The other issue is not all 99105's chips are the same, there is large number which are TMX (not TMS) which are buggy prototypes with errors in the microcode, so you got to make sure you have final production model, this is more of issue with the 99115 chips as they never reached production status.


Now, that I am back in the land of freedom, and basically have my permanent home now since 1st of nov. and almost everything setup here that I need, i will be looking again into this project and see if I can re-create the design or improve on it, now that we have better tech today in making PCBoards and faster fpga's, etc. will be about another month to get all my things out of cold storage and a workbench setup with soldering irons, etc. so this will be on my list of todo things starting early in the new year.


As for the 99105 this project from 2016 by Erik, is very nice as well: https://hackaday.io/project/15430-rc201699-ti-994a-clone-using-tms99105-cpu

  • Like 5
Link to comment
Share on other sites

  • 3 weeks later...

Wow, the OP dropped a bomb and left (save for a single additional post).  Was the question legit, or a joke and everyone took the bait?  Do people really use ChatGTP for search and information now?  Not sure I want to know the answer, actually.  I'm glad to see the quality of human responses out weigh the question by a massive amount.


Also, the question is too vague.  "Are there accelerator cards for the 99/4A?".  Yes, and that would cover anything you can plug into a 99/4A that is an accelerator, i.e. the TIPI, Geneve, F18A, Finalgrom, strangecart, etc., etc..  Or no, there is nothing you can "plug in" that directly makes the 99/4A's CPU or the computer itself faster.  So it depends on how you want to interpret the question.


The only qualifiers for accelerating the base system directly are probably the faster crystal to overclock the CPU, and the wait-state circuit modification, and neither of those are "plug in" solutions.  Even if you had a 9900 that took zero-time to execute an instruction (infinitely fast), you would still ultimately be limited by the system architecture speed, which is mostly defined by the memory (ROM and DRAM) used.  IIRC, that is somewhere around 200ns for ROM, and 150ns for fast DRAM of the era (so 5MHz to 6MHz best case).


  • Like 1
  • Haha 1
Link to comment
Share on other sites

5 hours ago, GDMike said:

Not REAL people....googly employees is it's best customer

I keep getting calls from the Google Business Listing people.  They are going to help me claim my Google listing because it shows I can permanently closed.  Oh, voice search services the same, they will update my business registration with Samsung, Apple, Google, and Microsoft.  Sometimes they just take it, sometimes they get indignant when I tell them that I do not want the kinds of customers those systems will bring to me.  The opener does nothing but verify my information, ask me if I am aware my listing is not up-to-date, and if I know what those listings are.  Then, boom, hands me off to the consultant who gets to debate with me.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Create New...