In his June 1985 foreword to Programming the 65816 by David Eyes and Ron Lichty, Bill Mensch expresses his hopes for a 6502-derived 32-bit microprocessor: the 65832. WDC is still thriving, but the ‘832 was apparently never released. What happened to these plans?
3 Answers
In the 1988 Report on the 65c832, Mensch described the 65c832 as a back-burner project with an uncertain timeline:
Since WDC is not a gigantic conglomerate, it has limited resources. If all your manpower, time, and money are going towards the development of the 65c265, you don't have any left for the '832. That's exactly what was happening with the 65c832 as of June, 1988. It's a concept that WDC employees kick around all the time, but on which active work has yet to begin. On the positive side, there's still time to influence WDC's design. On the down side, it will be a couple of years, at least, before the 65c832 is real.
A 51 page preliminary datasheet was then produced in 1990 and updated in 1991, but as far as anyone knows, no silicon ever came of it. As to why, one can only speculate (and a quick google will get you plenty such speculation -- either how great a chip it would have been, or what a failure).
Personally, I think it was the relative failure of the 65816-based Apple IIgs that convinced Mensch to shelve the project. Its key selling point had been that it was both a shiny, new 16 bit machine, and fully compatible with the Apple II line at the same time. But this didn't translate into sales in the way Mensch and Apple had hoped (Apple doing its share here by hobbling the IIgs and promoting the Macintosh).
The report quoted earlier frames the 65c832 as an upgrade to the IIgs. With the IIgs failing, and the 32bit market already largely divided up between Intel and Motorola, it must have been becoming clear in 1991 that a 6502-compatible 32 bit processor, with its performance hampered by compatibility requirements, would not be able to gain a significant market share.
-
1Many parts of the '832 datasheet (both versions) seem to just be copies of the '816 datasheet. In some places they search-and-replaced 816 and 802 with 832 and never proofread the result, so there are strange sentences like "However, some instructions and addressing modes are intended for W65C832 24-bit addressing and are therefore less useful for the W65C832." (page 49). I get the impression that someone knocked it together very quickly and there was never really a project as such.– benrgCommented Sep 12, 2022 at 6:59
I suppose you have heard of the draft datasheet of the 65832?
https://downloads.reactivemicro.com/Electronics/CPU/W65C832%20CPU%20Datasheet%20v2.0.pdf
Apple was the main customer of the 65816 (for Apple //gs) which needed to be compatible with 6502. They had no need of a 32bits version as Macintosh had already chosen the MC68000 family which is a sounder base as a 32bits CPU.
(Nintendo later also used a 65816 derivative in the SNES)
At some point before 1990 the idea seems to have first surfaced as a 'report' about 6502 chronology and future processors might suggest. It includes not only information about the ghostly 6516, but also information about a 65832 which should have more features than the datasheet showed. There was also no substantial follow up.
The project might have gotten serious in 1989/90 as 65C832. There is a 'preliminary' data sheet of March 1990 and a follow up data sheet dated March 1991. AFAICT the only difference is an added disclaimer.
The CPU described is a 65816 with A/X/Y extensible to 32 bit and a new emulation bit for 16 bit (65816) workings called E16. Like XCE which exchanges E8 and Carry a new XFE instruction should now exchange E8/Carry and E16 with Overflow (*1). Except there is no free code space for a new instruction, so it seems as if this might have been intended as a renaming of XCE - or simply not thought thru. In 32 bit 'mode' A can be 8/16 or 32 bit, X/Y only 8 or 32 bit.
Everything else was kept the same, this includes the hardware interface, so 8 bit data and 24 bit address. Any imagined speed gain could only be realized for a few instructions, often accompanied by quite some overhead for mode and register size switching.
Bottom line: Without substantial advantage there was no real resonance with professional (embedded) users, WDC's main customer base
The project got resurrected by WDC as Terbium in 2006, but this time with even less information and no resonance at all. The Terbium name got reused for a 65xx IDE (halfway down the page).
*1 - The missing SEV instruction in 6502 mode would be no hurdle, as switching to 32 bit mode only needs CLV.
-
Please check that "'report' about 6502 chronology and future processors" -- it's from 1988, and the same one I quote. It frames the 65c832 as an upgrade to the IIgs; refers to the time shortly after the release of the Motorola 6800 in 1974 as "over 12 years ago"; and describes the design status as of June, 1988. With its constant mention of how everything is still up in the air, it clearly predates any datasheet. Commented May 21, 2020 at 0:16
-
Another thing -- how certain are you that embedded users were WDC's main customer base in the early 1990s? I'd have thought that the main part of their revenue cam from royalty payments from Nintendo (or Ricoh, who made the 5A22 for Nintendo). The focus shifted to embedded uses a little later, I think, when it became clear that the SNES had run its course. Commented May 21, 2020 at 0:20
-
@MichaelGraf possible, but there is no real date given. Regarding customers, Game consols are embedded devices, but even when counting them, they are dwarfed by total numbers. Embedded has been the main use for 6502 from the very first days (think Rockwell, Mitsubishi, Renesas). WDC claims 5-10 billions made, but adding up all game consoles and home computers will end up way below half a billion, so where else have the other cores been used? Heck, even the famed IIgs was only ca. 1m units Also Rico did not pay any royalties for the 2A07 (not sure about the 5A22).– RaffzahnCommented May 21, 2020 at 0:31
-
As I understand it, the 5-10 billion number claimed by WDC includes all vaguely 65xx-based chips, including the original 6502, patent-skirting implementations like the 2A03/2A07, and third-party implementations after the original patents expired, summed up over the design's 40+ year lifetime. I completely agree that these are mostly embedded. But the relevant question here is where WDC's money was coming from in the early 90s. I'd hope that Mensch learnt from the 2A03/2A07 patent fiasco, and got paid by Nintendo. Commented May 21, 2020 at 10:42
-
2Some interesting reading on 2A03/2A07 here: linkedin.com/pulse/… Commented May 21, 2020 at 10:43
lda (zp),y
would go from taking 5-6 cycles to taking 7-8, and such speed penalties would affect a huge number of instructions. Something analogous to the 68HC16's instruction set might have made sense with a 16-bit data bus (a certain prefix byte would be placed before instructions...