[sdiy] Chip ID

[Magnus Danielson]

From: "Batz Goodfortune" <batzman at all-electric.com>
Subject: Re: [sdiy] Chip ID
Date: Sun, 05 May 2002 00:54:21 +0930

> Y-ellow Magnus 'n' all.

Batz,

>          Yes indeed the 8250 is the nat-semi beast. A dual UART in fact. 

No... if my memory serves me right is the 8250 a single UART device.

> Didn't they call them DARTs?

Dual UARTS have been called DARTs, but I think you'll have to go into
the 6800 series to find them.

> I thought it was that intel also had an 8250 which caused the
> confusion but upon checking this is not so.

I was into the chips in the 80thies, so I keep supprising me by still
remember these things accurate enought.

> What I was thinking of was the 8253/54. In that although they're a
> different beast entirely there were two chips with basically the
> same functioning but two numbers apart.

Yes, that's Intel's variant of 3 programable counters in one chip -
Programmable Interval Timer (PIT). The 8254 is the developed out of
the 8253.

> For anyone actually interested in this hysterical ranting. 
> Err Sorry. HISTORICAL ranting, they are counter timer chips. Basically 3 
> counter/timers per chip. The original PC XT/AT's had all of the above.

The original PC had one 8253, the AT had the 8254.

> Err not the 8255. Damnit!
> 
> These were all 8 bit chips designed for 8bit systems. And just before all 
> the Moto-heads out there start laughing, Motorola did exactly the same with 
> the 68K at the time too. Tell me again how they're all 16 bit machines again?

For some stuff, Motorola didn't map over to 16 bit, not did Intel. It
would just be a waste of pins since you where not in need of high
transfer rate for much of them anyway.

> I never really thought about how the legacy actually goes all the way back 
> to the 4004. That's funny when I think about it because it's so true. Intel 
> would never officially admit this but when they were developing the 8086, 
> it was simply meant to be a faster 8085. (8 Bit).

But they did, they did! This is Intel-tradition actually, just push it
slightly in the right direction for each re-take. The 80286 is
actually an odd beast, bringing only protection but no virtual memory
handling. The 80186 is also an odd beast. Hell, the 8086 is an odd
beast too. While many will easilly see me as an Intel-head, I'd say
that the 68k was a marvelous chip since it broke with the 6800
tradition and where a new clear cut. I still want to do 68k projects!

> But Motorola were due to release a 16 bit processor so they kinda
> changed their plans mid stream. My guess is that they were too far
> down the track to an 8 bit and had no time to assemble a new team to
> throw together anything that would rival motorola.

Well, maybe...

> So when you look at the architecture of the 8088/86 it looks strangely like 
> an 8 bit processor with tacked on bits. Which is essentially what it is. 
> The instruction set is remarkably similar. Everything. This is why the 
> stupid thing can only see a 64K window at a time and has that God awful 
> segmented addressing.

Exactly!

> All legacies which we are still dealing with to this 
> day. Motorola had a contiguous address space and kicked ass from a machine 
> code point of view. Although it had other issues which were just as insane.

Right.

> 8089 I/O co processor:
> I actually had a machine which used an 8089 co-pro as well as an 8087 math 
> co-pro. In fact I had 3 of them. Made by AWA and were designed to be used 
> as multi-drop servers. Essentially diskless work stations. Though there was 
> provision for 2 drives. Kicked ass in their day I guess although it wasn't 
> "In their day" when I acquired them. But interesting to see the 8089 in 
> action. Actually they had a slot for a second one as well. Dunno what they 
> used that for. Probably disk I/O if they had disks? Anyway. A great many 
> people would often lament the fact that provision for the 8089 was not 
> included in the architecture of the PC. And "Bob" knows it could have done 
> with one.

Right. Also, the 8089 assembler had no resemblance to the 8086
assembler. Another set of tools. Lovely! ;O)

> At 04:13 PM 5/4/02 +0200, Magnus Danielson wrote:
> 
> >The 6502 was two steps forward and one step back. The 8008 has more
> >registers, six data registers and one accumulator. The 8008 doesn't
> >tick as quick as the 6502, which formidably stands still by todays
> >preferences.
> 
> Well, I was only trying to put a date on it by mentioning the 6502 but 
> since we're here... The 6502 was supposed to be a souped up version of 
> Motorola's 6800. The 6802 was motorola's own slight improvement on that. 

Right. The 6809 seems like about the peak of that
architecture... (this is where I expect a bunch of 68HC11 people to
jump over me... I just haven't looked at 68HC11 in comparision with
the rest of the 6800 crew).

> The zilog Z80 was born from a bunch of disgruntled intel engineers who 
> thought they could do better than intel's 8080. And they did. While the 
> 6502 was stripped down and fast, the Z80 was scaled up and fast.

Right.

> And right there we have the birth of both RISC and CISC, and of the
> concept of "The Motorola architecture" VS "The Intel architecture."

Actually, much of the hurdles have been taken place elsewhere and on
much more important architectures. Over in the Mainframe and
Minicomputer market things have been happening over time. Just diving
into the PDP-11, DEC-10 stuff gives insight, then seeing the VAX stuff
makes you understand what CISC means. Meanwhile had Cray cooked up a
much simple thing while the MIPS and SPARC projects where ongoing at
Stanford and Berkeley (I keep forgetting which project where at which Uni).

> Processors that strayed too far from either path tended to have
> limited life spans. I think because if your processor didn't have
> adequate support chips available to it, then it's use was limited.

True.

> The irony of all this is that these legacies dog us to this day. From the 
> Itanium to the latest G4 PPCs.

The PPC builds on IBM tradition, which is actually older than Motorola
tradition. It was actually the IBM 360 which where the first machine
with (8-bit) byte addressing. Today's IBM S-390 series has inherit
from the 360 architecture. If we should thank/blame someone I think
Fredrick Brooks is the only single individual that I can name who is
to blame/thank for it.

Whatever happend in the micro-processor developement had allready
happend at least 10 years before. Intel tried to patent upgradeable
micro-code. They failed. In the 70thies, DataSaab (computer side of
the SAAB auto/air manufacture) had both implemented, sold and written
an article about it. They did fail to patent it back then since they
wrote the article on it. Digital implemented the feature in their
DEC-10 machines, even the ECL-based KL-10 had the microcode loaded
from tape as part of the boot procedure. It was just a natural
development then, since up till then the microcode had been first
wired logic, then combined logic and PROMs, small and fast RAMS
allowed for simpler development. Intel saw their patent vanish into
thin air when they where taken to court. It *was* well known a
published.

This is the fun thing about digging into old computers like DECs
etc. You learn the real history. This was the hardware the
microprocessor guys looked at when the got
inspiration. Microprocessors did not develop out of thin air from the
4004 and onward, it happend when the processing technology allowed it.

> Even though none of them use any of the 
> legacy peripheral chips. Along the way there have been some truly awesome 
> architectures but no-one's putting them on the desk top because the world 
> has been so neatly divided into two camps. Where there was once a fair 
> diversity in processors, you really only have a choice of 2. Especially now 
> that MIPS is being side-lined and the StrongARM has found a niche in the 
> PDA market. And of course, they held a funeral for the DEC ALPHA a few 
> years ago. Burial at sea I think?

I personally want to burry the IA-64 and let the Alpha live. The IA-64
is a badly done architecture compared to the Alpha. Intel should just
shut their buissness down in shame of what they have done!

> Of course, Linux runs on everything. Well almost everything. I had trouble 
> running it on my abacus but the Babbage engine runs fine. Though I had to 
> convert the boiler to gas recently. Wood and coal are getting expensive. 
> Hey. It might be slow and noisy but it makes a good cappuccino.

Hehehe... Linux rules! ;O) I tuned in, turned on and flipped out
somewhere in late authum of 93, before Linux kernel 1.0 release. I
naturally have a signed copy of "Just for fun".

Penguin on ya'!

Cheers,
Magnus