Are we TI'er's living our second childhood?

[+mizapf]

Sorry, I did not intend to sound touchy. Still, I'm arguing against statements like "the TI-99/4A is not a real 16 bit computer" or even saying "it is an 8 bit platform". We've had that since the time the TI rivalled against the other home computers which were 8 bit indeed. The architecture width does not imply anything about performance either, nor about adressible memory.

 

Since the CPU has a 16 bit architecture we gain some powerful commands compared to the 8 bit platgorms, but we also saw that by the bus multiplexer we have to face heavy performance penalties.

 

Yes, there is also a tiny glimpse of pride that we may have for using a 16 bit home computer platform.

 

(And it's pain to write long texts with a tablet computer ... still being on vacation. Next week I'm back at the keyboard and can write longer texts that make more sense)

[sometimes99er]

See I thought that it was originally designed for an 8bit processor but when that processor wasn't ready for release they shoehorned in a 16bit processor. Is this not the case?

Yes, I think it is.

 

The TMS9900 is a 16-bit microprocessor.

Yes, looks and feels like one.

 

By all classic and contemporary computing conventions the TI99/4A is a 16-bit computer. To say otherwise is simply twisting semantics around - like ...claiming a gaming PC is 1024 bits because it has 2 graphics cards that have a memory bus width of 512-bits each. Or calling an I7 a 256-bit processor because it has 4 cores that are 64-bits.

 

And it's 16-bit RAM is 128 words (256 bytes with 32 taken by the CPU). Most TI-99/4A's sold had that and never more. And many super-cars have speed limiters not necessarily not changing the fundamental car and engine design performance. Not to mention read before write. Wait-states. Multiplexing. List continues ... Not the fault of the 9900, I know. - Overall performance for the TI Basic user and programmer ? Yark ! (still in love though - somehow)

 

Funny how the Z80A (MSX) can shake the 9918A and we barely can. You know, 8-bit vs. 16-bit. Just saying.

 

Edited August 20, 2014 by sometimes99er

[sometimes99er]

And it's pain to write long texts with a tablet computer ... still being on vacation.

You can buy some really cheap and good Bluetooth keyboards (9-10") that make a difference when you're on vacation.

Edited August 20, 2014 by sometimes99er

[Willsy]

Funny how the Z80A (MSX) can shake the 9918A and we barely can. You know, 8-bit vs. 16-bit. Just saying.

 

Right. I guess if the 9900 was running in a true 16 bit environment, with no wait-states etc, it would give the Z80 a run for its money. An even better choice would be a 9995, which could probably kick the Z80s ass, and still give all the advantages/convenience of a true 16-bit processor. (Just 'cos the 9995 has a 8 bit *data bus* it's still a true 16-bit processor).

 

An interesting argument is the Motorola 68xxx range. I've seen them described as 16/32 bit processors. Having programmed various 68k flavours (many years ago) they were equally comfortable in 16 bit and 32 bit worlds. But let's not talk about the 68008 (though I still remain in love with the Sinclair QL ).

[Gary from OPA]

It's not? See I thought that it was originally designed for an 8bit processor but when that processor wasn't ready for release they shoehorned in a 16bit processor. Is this not the case?

It was always going to be 16bit processor code just with 8bit bus instead similar to early 16bit intel chips.

 

The only thing that happen is the original silicon die that was similar to the 9995, failed, they could not get it to be stable enough in time for release, so they quickly had to figure out how to get 256 bytes of ram on the pcboard along with multiplex/demultiplex circuit build out of '74 logic and make space for larger 64pin tms9900, which greatly raised the production/tooling cost, leaving marketing stuck selling the system higher then originally planned on launch.

 

The original design was full 16bit 9900 core with onboard highspeed ram on die, along with mulitplex circuit bringing the pin count down to 40pins, and I have one of those unstable cpus in my collection.

 

Years later they figured out the cross-talk signal problems on the die and were able to release the 9995 which the geneve used.

[idflyfish]

Sorry, I did not intend to sound touchy. Still, I'm arguing against statements like "the TI-99/4A is not a real 16 bit computer" or even saying "it is an 8 bit platform". We've had that since the time the TI rivalled against the other home computers which were 8 bit indeed. The architecture width does not imply anything about performance either, nor about adressible memory.

 

Since the CPU has a 16 bit architecture we gain some powerful commands compared to the 8 bit platgorms, but we also saw that by the bus multiplexer we have to face heavy performance penalties.

 

Yes, there is also a tiny glimpse of pride that we may have for using a 16 bit home computer platform.

 

(And it's pain to write long texts with a tablet computer ... still being on vacation. Next week I'm back at the keyboard and can write longer texts that make more sense)

 

It's ok...I understand. The main thrust of my question wasn't to touch off a debate but to see what makes the TI-99/4a so compelling to everyone.

[Gary from OPA]

It's ok...I understand. The main thrust of my question wasn't to touch off a debate but to see what makes the TI-99/4a so compelling to everyone.

For me it was two factors, the 16bit assembly, even tho you needed to expand the console greatly to due so, unless you could live with the 4k mini-memory and loading line-by-line assembler in by tape.

 

And the music chip and speech addon, there was alot of nicest in it, compare to other systems at the time, even tho the TI99 was pricely at launch.

 

Sadly alot of people didn't jump onto the TI99 bandwagon until it was too late when it was being cleared out way below cost at $99 or even less in '83..

[idflyfish]

For me it was two factors, the 16bit assembly, even tho you needed to expand the console greatly to due so, unless you could live with the 4k mini-memory and loading line-by-line assembler in by tape.

 

And the music chip and speech addon, there was alot of nicest in it, compare to other systems at the time, even tho the TI99 was pricely at launch.

 

Sadly alot of people didn't jump onto the TI99 bandwagon until it was too late when it was being cleared out way below cost at $99 or even less in '83..

 

I believe that is when my Dad got my brother and I one. Unfortunately we didn't really understand what we had and really only used it to play Demon Attack and TI Invaders.

[Keatah]

I just felt the system was full of mud so to speak. When I had mine I didn't have any games that were superfast or colorful. But make no mistake, the little bit I did use it I liked it.

[Willsy]

Years later they figured out the cross-talk signal problems on the die and were able to release the 9995 which the geneve used.

 

That's not entirely accurate. The TMS9995 was designed, from scratch, by Karl Guttag (an American) in Bedford, England. The work started in 1979. You are correct in that the 9995 *did* replace the 9940, but the 9995 was not a "corrected" 9940. What I mean is, they didn't pull the design of the 9940 out and de-bug it. They (or rather, Karl) started from scratch, without reference to the 9940 design.

 

There's some nice background to the story here.

Edited August 21, 2014 by Willsy

[Willsy]

He also worked on the TMS99000, too. Cool

 

 

 

He headed the logic and design architecture of the TMS 9995 (1979) and TMS 99000 (1980-1981) 16-bit microprocessors. Both architectures went to production with only slight modifications to first silicon. His work on these architectures at a time when TI had had a series of problems building previous processors plus his work on the VRAM, led to him being selected as the youngest Senior Member of Technical Staff (SMTS) at TI in 1982.

[jens-eike]

Some good reading about Karl Guttag's work with the TMS99xx family:

 

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

 

About the 8/16-bit bottleneck: In his notes, he mentions the 9985 originally planned as CPU for the 99/4. The 9985 was derived from the (flawed) 9940, it was an 8-bit CPU made to work on the 16-bit instruction set....:

Quote: "The 99/4 was originally slated to use the 9985 CPU and not need the 9901 at all. The 9985 CPU was going to be a variation of the 9940 “microcomputer” with embedded EEPROM. The 9985 replaced the EEPROM with 256 Bytes of on-chip RAM. The 256 bytes of on-chip RAM was for the GPL and Basic interpretation, and had to be shared between instructions and data.

 

The 9940 was a terrible design, it was big, slow and had a huge number of design bugs and to top it all off, there was huge pressure from management. So the 9985 was waiting behind getting the 9940 debugged and the 9940 was very hard to debug the way they designed it.

[...]

Now the 9900 had a 16-bit CPU with a 16-bit ALU and a 16-bit external interface but it was pretty inefficient and took 14 clock cycles to do a register to register add. The 9985 had only an 8-bit ALU and 8-bit I/O and took about 10 cycles to do a register to register add. Inside the 9985 the interface to the 256 bytes of RAM was also only 8-bits wide. To emulate the 9985, the 99/4 had to add hardware to covert the 16-bit bus coming out of the 9900 to act like the 8-bit bus of the 9900."

 

source:

 

http://spatula-city.org/~im14u2c/vdp-99xx/e1/99-4_History_By_KG_in_answers_to_Matthews_Questions.doc

[Gary from OPA]

Some good reading about Karl Guttag's work with the TMS99xx family:

 

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

 

About the 8/16-bit bottleneck: In his notes, he mentions the 9985 originally planned as CPU for the 99/4. The 9985 was derived from the (flawed) 9940, it was an 8-bit CPU made to work on the 16-bit instruction set....:

Quote: "The 99/4 was originally slated to use the 9985 CPU and not need the 9901 at all. The 9985 CPU was going to be a variation of the 9940 “microcomputer” with embedded EEPROM. The 9985 replaced the EEPROM with 256 Bytes of on-chip RAM. The 256 bytes of on-chip RAM was for the GPL and Basic interpretation, and had to be shared between instructions and data.

 

The 9940 was a terrible design, it was big, slow and had a huge number of design bugs and to top it all off, there was huge pressure from management. So the 9985 was waiting behind getting the 9940 debugged and the 9940 was very hard to debug the way they designed it.

[...]

Now the 9900 had a 16-bit CPU with a 16-bit ALU and a 16-bit external interface but it was pretty inefficient and took 14 clock cycles to do a register to register add. The 9985 had only an 8-bit ALU and 8-bit I/O and took about 10 cycles to do a register to register add. Inside the 9985 the interface to the 256 bytes of RAM was also only 8-bits wide. To emulate the 9985, the 99/4 had to add hardware to covert the 16-bit bus coming out of the 9900 to act like the 8-bit bus of the 9900."

 

source:

 

http://spatula-city.org/~im14u2c/vdp-99xx/e1/99-4_History_By_KG_in_answers_to_Matthews_Questions.doc

yeah, I got a couple of 9985 cpu's myself also, always wondered about hacking up a ti99/4a board and wiring one in to see how it would work speed wise.

 

The 9995 had a totally redesigned microcode so the opcodes ran alot faster.

[+OLD CS1]

The TMS9900 is a 16-bit microprocessor. By all classic and contemporary computing conventions the TI99/4A is a 16-bit computer.

 

Yup. The original Amiga is a 32-bit computer because the core of the 68000 is 32-bit, though its data bus is 16-bit. This is where I got lost in the Mac world's idea of "32-bit pure" programs and similar conventions. The only "pure" with which I have to contend in the Amiga world was whether an executable is re-entrant or not.

 

Speaking of, a little "off-topic" in an off-topic thread, the latest UAE beta includes PPC support, and it apparently works awesomely, so UAE should soon be capable of running the "modern" AmigaOS 4.

[+OLD CS1]

Funny how the Z80A (MSX) can shake the 9918A and we barely can. You know, 8-bit vs. 16-bit. Just saying.

 

Could that be because of the read-before-write that 9900 does? I read somewhere the 6502 does the same thing but not certain about the 6510 (used in the Commodore 64) and neither the Z80 nor 8502 do (used in the Commodore 128.)

 

Please correct me, as I just cannot be arsed to look it up right now. (Tends to happen around this time of night.)

 

 

About the 8/16-bit bottleneck: In his notes, he mentions the 9985 originally planned as CPU for the 99/4. The 9985 was derived from the (flawed) 9940, it was an 8-bit CPU made to work on the 16-bit instruction set....:

 

"...To emulate the 9985, the 99/4 had to add hardware to covert the 16-bit bus coming out of the 9900 to act like the 8-bit bus of the 9900."

 

A "variant" is not necessarily a "derivation." Then, I suppose unless peripherals had already been designed, why emulate the 9985?