The argument was always about the hardware, never about the software.
I mean, you will agree that the fact that a certain system can run 16-bit software does in no way imply that the system itself has to be 16-bit.
Heck, the world still revolves around x86 processors, which are still capable of running 16-bit 8086 software. Most of them aren't actually are 16-bit though.
Okay, so that brings up another point: is a Pentium-based computer a "64 bit system"? The CPU has a 64 bit wide data bus so going by your logic that should be the deciding factor, not the length of its registers. Or, how about some of those Macintosh systems that mixed bus sizes like crazy, like the LCII and Mac Classic II: they have a 68030 CPU, a "32 bit" CPU that transparently supports dynamic bus resizing, and thanks to a diabolical piece of VLSI logic those systems have 32 bit access to ROM, 16 bit access to RAM, and most of the peripherals are 8 bit. How many bits is that "System"?
Personally I sympathize with your position in that I acknowledge that everything downstream of the bus unit in an 8088-based computer is on a bus 8 bits wide, but I can't really 100% agree that it makes the IBM PC an "8 bit computer" and I totally remember them being advertised as 16 bit machines outside of the specific instances where the narrow bus was being pointed out as a performance bottleneck.
My argument was rather the opposite: Since even the marketers didn't really dare to market these machines as fully 16-bit or 32-bit respectively, apparently it was quite a controversial issue.
In fact, the "ST" in Atari ST is said to stand for "Sixteen/Thirty-two", referring to the 68000.
So... let's acknowledge that with both the 8088 and 68000 we're somewhere on slippery slopes between 8/16 and 16/32 bit respectively. Slopes do have a dimension, however, and I'd argue that the 68000 is closer to being a 16 bit processor *despite its 32 bit ISA* than the 8088 is to being an 8 bit one. The 8088 has exactly the same execution unit/ALU/register file/etc as the 8086 and is therefore genuinely 16 bit internally; only the BIU has been swapped out. When an 8088 is actually crunching numbers it's doing it a full 16 bits at a time, exactly the same as its
older brother. The 68000, on the other hand, is more of a mixed bag; it has 32 bit long registers and the microcode describes a 32 bit ISA but it processes everything with a 16 bit ALU. Therefore the 68000 never actually does anything 32 bits at a time (other than some address calculations carried out by their own specialized circuitry), so it *is* actually more of a stretch to call it a 32 bit CPU than it is to call the 8088 a 16 bit one. Maybe *just* enough more of a stretch to scare the marketing people?
In any case as previously noted it's exceedingly silly to rate systems by how many "bits" they are anyway because, as noted, that number alone can have only the most tenuous relationship to the actual capabilities of the system. It's understandable how it became so popular in video game marketing given the adolescent male culture that buys game consoles is all about size matters (MOAR == BETTER, always!), but in serious conversations its value is pretty much nil.
It is sort of amusing that even that January 1982 Byte review, which is generally enthusiastic about the IBM PC, points out that its raw performance seems a bit unimpressive. (IE, they note that the BASIC seems hardly faster than Applesoft on an Apple ][, and also mentions it seemingly losing several off-the-cuff tests against a 4Mhz Z-80.) Their enthusiasm seemed to be entirely predicated on the idea that the larger RAM capacity would be the enabling factor the PC had over the competition, and I think they were dead on about that. As capable as a C64 might be compared to a 5150 when it comes to throwing graphics up on the screen you can still hold a much larger spreadsheet in RAM on the PC.
Over here the PC was only popular at the office in the 80s. C64s and Amigas were cheaper and the games were much better, so why would you want a PC at home?
The IBM PC simply wasn't a competitor in the home market. Far more expensive than the other options, and far less capable.
Well, America
is the land of working yourself to death, and when PCs started sprouting like mushrooms in offices the allure of being able to take work home became a deciding factor in most home computer purchases made after 1985 or so. And... maybe my memory is getting fuzzy, but I do remember it was somewhere around '85-'86 or so when suddenly the math just stopped adding up for the 8 bit micros. Sure, you could get a C64+1541 for only about $300 or so, but by that time rotgut PC clones with 256k and dual disk drives were starting to break the $1000 mark and for the difference in price you were getting four times the RAM (and trivially expandable), four times the disk space (and the ability to relatively cheaply add a hard disk), an 80 column screen, etc. Sure, it's an apples-to-oranges comparison, the C64 doesn't "need" as much hardware to be useful, but it still comes off looking like not that great of a bargain when you factor in the impossibility of lugging your office work home to it. Being cheap and "good at games" just make it look even more like a toy instead of a
real computer.
(Likewise the math ruthlessly undercut the chances of the Atari ST and Amiga in the US. Sure, objectively they were pretty attractively priced for the level of capability you got, but they both cost "about as much" as a complete PC once you factored in a monitor and similar amount of disk storage, but they weren't compatible with the "gold standard", and as the market ruthlessly moved forward through '86 and '87 they actually both started looking a little pricey for what you got, particularly if you wanted a hard disk. When the Amiga 500 came out in October '87 at $699 for just the system unit it was competing against off-brand Turbo XT clones that'd give you dual floppies and a mono monitor for that price; add a color monitor to the Amiga and you can just about have a hard disk in your XT.)
I totally envy the European home computer market for not having the crushing boot heel of PC compatibility pressed on its throat through the second half of the 1980's but that's how it was on the west side of the pond. Pity that.
That is not a logical conclusion, but the logical fallacy known as a "False dilemma".
Well, isn't this entire argument something of a false dilemma? You support the "not 16 bit" conclusion by claiming that they were not advertised as sixteen bit, but in the process you've summarily excluded from evidence all the editorial content that refers to them as such.
(The Byte archive is convenient so... here's a review of the TI Professional that says "Like IBM, TI abandoned its own proprietary microprocessors and based this computer on Intel's 16-bit chip instead." And, in fact, if you search for "16-bit" in this issue you will find multiple press releases and ads that refer to 8088 machines as 16 bit.)
(And that really is just one; there's another for Sanyo MBC-550's earlier, there's more later...) There are indeed ads and editorial content that call them "16/8 bit" computers instead of just "16" because of the bus size, but there *is* a good reason they're not advertised as simply "eight bit" (by people who actually want to sell them, anyway) and that is because they are not according to the most accepted measures of a system's "bitted-ness". (IE, register/ALU size and programming model.) But in any case, it seems sort of illogical/dirty pool to site a lack of advertising as support for your position while simultaneously denying that it's valid to apply the same argument in reverse.
Anyway, seriously, can't everyone win here? The PC is a 16 bit computer with an eight bit bus and, by extension, an 8 bit system architecture. I see no harm in someone informally referring to it by either number but apparently that must *really matter* in some way I'm missing.