8086 uP Require

[Kryten]

I'm pretty sure the reason they used the 8088 instead of 8086 chip
was to keep what later became known as the ISA bus, more simple.

And cheap.

They were sitting on the fence between the 8-bit and 16-bit architecture, in
a time where the latter meant twice as many DRAM/ROM chips and sockets for
an entry level machine.

Hence they got stuck with the pain of the segmented 808X architecture.

There's an old* Sicilian saying that he who sits on the fence
gets splinters up his ass.


(* and entirely fictitious)

 

[Keith]

And cheap.
Yes.

They were sitting on the fence between the 8-bit and 16-bit architecture, in
a time where the latter meant twice as many DRAM/ROM chips and sockets for
an entry level machine.

Pretty much. IBM used the 8088 (later than the 8086) mainly to save
connectors.

Hence they got stuck with the pain of the segmented 808X architecture.

That was no issue at the time. The alternative was the 8085. I'm sure
that would have mane many happy! ;-)

There's an old* Sicilian saying that he who sits on the fence
gets splinters up his ass.

The fence isn't what you think it was. Motorola wasn't in the running.

 

[Jasen Betts]

["Followup-To:" header set to sci.electronics.misc.]

I did a 68000 design in the same time frame (while sneering at Intel). Used
64K DRAMs, but the first batch cost us about $100 each. That's why I was
pretty sure they were not used in the first PC. Got the thing to legally run
at 12 MHz with 0 wait states from the 128K onboard DRAM. Running C
benchmarks showed performance about equal to a VAX. I may still have the
same Byte magazine you allude to.

yeah, but which vax? I'd expect more performance from your setup than from an
than an 11/760.

 

[Kevin G. Rhoads]

The initial, "standard" speed of 8088/6 was 4.77 MHz.

Actually that was only the IBM PC, they used a 5 MHZ part and derived
the microprocessor clock from the same crystal that was used to generate
NTSC video clocking, thus the weird 4.77. Same way with the 640k vs 1M,
a PC-ism. Of course, 8088 and 8086 boxes that weren't really IBM compatible
didn't fare too well in the marketplace, so ...

 

[Rich Grise]

I knew if i kept that original '84 IBM PC it might be useful to someone or
worth something again one day now if i could get the original price for i
would be very happy

good luck fnding one if you exhaust your other sources then post back here
if my cpu will help you

An 8088 is perfectly useless in an app that calls for an 8086. The core is
the same, but the 8086 has an actual 16-bit data bus, where the 8088 has
only an 8 bit bus, and doubles up on memory/peripheral access. I've read
somewhere that the reason they did this is because existing peripheral
devices were all 8 bits, and a 16-bit wide RAM costs twice as much as
8-bit.

I wonder why the OP wants a "vintage" model. Something to do with rev.
level?

Thanks,
Rich

 

[WayneL]

My origrinal request states why I need an 8086.

WayneL






Hi

Could anyone help us find an old very early 8086 uP for our University's
SEM?
It is a very old unit be our Scanning Electron uScope will not operate
untill we find one.
I think the speed is either 1 or 4MHz. It is imprtant we get one of the
earliest version.

Thanks

WayneL

 

[Winfield Hill]

WayneL wrote...

My origrinal request states why I need an 8086.
WayneL

Could anyone help us find an old very early 8086 uP for our
University's SEM? It is a very old unit be our Scanning
Electron uScope will not operate until we find one. I think
the speed is either 1 or 4MHz. It is imprtant we get one
of the earliest version.

You need an NMOS 8086 in its standard 40-pin DIP package.
It's maximum speed capability should not be important; any
of them can be run more slowly, so an 8086-2 should work.

Alternate brands to Intel, such as Kryten's Fujitsu parts,
should also work well, they were meant to be replacements.

Yo'all should have easily gotten one by now, is your party
at the University still looking for one? Shall we bring
out one of s.e.d.'s big guns for you? {smile}

Or, is it possible the failure has been mis-diagnosed, and
it's not the microprocessor? A common problem in very old
instruments is that the PROMs fail, so the program crashes.
Many types of old PROMs only had 10-year lifetime ratings.

A maintenance safety procedure for an old valuable machine
is to read and store away copies of all the working PROMs,
before they fail.

A smart thing to add to a micro's code is a CRCC check of
the entire code very early in the bootup, so a "program
started, and ROM code OK" LED can be turned on. This can
save a lot grief in maintenance debugging, when the machine
is otherwise just sitting there, apparently doing nothing
after applying power.

 

[WayneL]

Hi Win


We have tried the 8086-2 and it works but the monitor rolls. When
we tried an 8086-1 (or there abouts) from another University it works fine.
So we are sure we need an 8086-1.
One of the people from this group sent me a 8086-2 but it did not work. So
if your offer is open to send a 8086 then we would greatly appreciate it.


WayneL

 

[Winfield Hill]

WayneL wrote...

Hi Win

We have tried the 8086-2 and it works but the monitor rolls. When
we tried an 8086-1 (or there abouts) from another University it
works fine. So we are sure we need an 8086-1. One of the people
from this group sent me a 8086-2 but it did not work.

I wonder if there's a difference in the execution time of some of
the instructions, and your instrument has critical software timing
associated with generating the display. Software timing is simple
and was popular for a time in the early days before programmers
realized what a mistake it was. The IBM PC had from the beginning
Timer0, running at 1.19MHz, which could be checked anytime, adding
hardware certainty to software timing tasks, despite changes in
CPU clock rates or instruction-execution time.

So if your offer is open to send a 8086 then we would greatly
appreciate it.

I don't have one to send, but I may be able to help you get one.

 

[Winfield Hill]

Winfield Hill wrote...

WayneL wrote...

I wonder if there's a difference in the execution time ...

OK, here's what I've found out so far. First, I looked at our
old ICs, and sorry, no 8086s.

I have in front of me my old Intel databooks. First, a huge
book called The 8086 Family User's Manual, dated October 1979.
This book covers the 8086 in great detail, but doesn't mention
a -1 version. Second, the Intel Component Data Catalog, dated
1981. Aha, here they cover the 8086, 8086-1 and 8086-2 on the
same datasheet. What this reveals is that they're the same IC
but with different speed grading during inspection. They all
have the same minimum clock speed, 2MHz. The 8086 has a 5MHz
maximum, the 8086-2 has an 8MHz max, and the 8086-1 is tested
for up to 10MHz maximum. That's right, the -1 version was
later than the -2 version.

My third book is the Intel iAPX 86/88 User's Manual, dated 1985.
This book shows the 8086, 8086-2 and 8086-1, but nothing faster.
Since by 1985 Intel was starting to make other more powerful ICs,
it seems they didn't go fooling with instruction execution times
or anything else for their basic 8086 processor.

So that's all there is to it, and it appears any 8086 processor
IC version that doesn't crash should work fine in your machine.
I'd guess that the 8080-1 parts should work fine as well, even
at 10MHz. I'd also say that newest 8086 ICs you find should
have the best chance of working at higher clock rates. Second-
source 8086 chips should be even better.

OK, that's it, over to you Wayne, for comments.

 

[Winfield Hill]

Winfield Hill wrote...

Winfield Hill wrote...

OK, here's what I've found out so far. First, I looked at our
old ICs, and sorry, no 8086s.

I have in front of me my old Intel databooks. First, a huge
book called The 8086 Family User's Manual, dated October 1979.
This book covers the 8086 in great detail, but doesn't mention
a -1 version. Second, the Intel Component Data Catalog, dated
1981. Aha, here they cover the 8086, 8086-1 and 8086-2 on the
same datasheet. What this reveals is that they're the same IC
but with different speed grading during inspection. They all
have the same minimum clock speed, 2MHz. The 8086 has a 5MHz
maximum, the 8086-2 has an 8MHz max, and the 8086-1 is tested
for up to 10MHz maximum. That's right, the -1 version was
later than the -2 version.

My third book is the Intel iAPX 86/88 User's Manual, dated 1985.
This book shows the 8086, 8086-2 and 8086-1, but nothing faster.
Since by 1985 Intel was starting to make other more powerful ICs,
it seems they didn't go fooling with instruction execution times
or anything else for their basic 8086 processor.

So that's all there is to it, and it appears any 8086 processor
IC version that doesn't crash should work fine in your machine.
I'd guess that the 8080-1 parts should work fine as well, even
at 10MHz. I'd also say that newest 8086 ICs you find should
have the best chance of working at higher clock rates. Second-
source 8086 chips should be even better.

OK, that's it, over to you Wayne, for comments.

Further to the story. I have now read the datasheets of three
second-source 8086-1 ICs, by AMD, Siemens and Fujitsu. I've also
read datasheets dated through 1995 for all four manufacturers.
The parts offered by the four are just the three original Intel
versions listed above, despite the years of fab development and
the likely obvious improvements in performance of the real parts.
I'd venture to say that any of the three versions, from any of
the four manufacturers, if made after say 1985 or so, should in
fact run just fine at 10MHz, or above.

I'd also include the CMOS versions, the 80C86, made by Intersil,
(also called Harris), OKI and Intel. These became quite popular
and there's extensive inventory available in the aftermarket.
OKI explicitly rated their msm80C86A-10 version at 10MHz.

As for finding exact ancient 8086-1 marked parts, I have located
some and asked for quotes. We'll see what they say. But in the
meantime, I'd continue to try in your machine whatever 8086 ICs
you get your hands on, whatever the label.

 

[Kryten]

WayneL wrote...
Hmm, if it is a speed issue - and it looks like it is - then you might be
able to compensate for the slower CPU by using faster RAM. I have lots of
32Kx8 SRAM chips from Pentium caches, rated at 15 ns. That ought to be fast
enough for an antique 8086.

 

[Joseph2k]

Winfield Hill wrote...

Further to the story. I have now read the datasheets of three
second-source 8086-1 ICs, by AMD, Siemens and Fujitsu. I've also
read datasheets dated through 1995 for all four manufacturers.
The parts offered by the four are just the three original Intel
versions listed above, despite the years of fab development and
the likely obvious improvements in performance of the real parts.
I'd venture to say that any of the three versions, from any of
the four manufacturers, if made after say 1985 or so, should in
fact run just fine at 10MHz, or above.

I'd also include the CMOS versions, the 80C86, made by Intersil,
(also called Harris), OKI and Intel. These became quite popular
and there's extensive inventory available in the aftermarket.
OKI explicitly rated their msm80C86A-10 version at 10MHz.

As for finding exact ancient 8086-1 marked parts, I have located
some and asked for quotes. We'll see what they say. But in the
meantime, I'd continue to try in your machine whatever 8086 ICs
you get your hands on, whatever the label.

I understand that most everyone is enamored to high clockosis, but OP made
it clear that is had to operate at original equipment clock speed. Just
the same it is nice to see that it is still available.

 

[Pooh Bear]

Hmm, if it is a speed issue - and it looks like it is - then you might be
able to compensate for the slower CPU by using faster RAM. I have lots of
32Kx8 SRAM chips from Pentium caches, rated at 15 ns. That ought to be fast
enough for an antique 8086.

I beg your pardon ! You reckon a 10MHz cpu needs 15ns RAM ?

What a funny idea.

Graham

 

[[email protected]]

I beg your pardon ! You reckon a 10MHz cpu needs 15ns RAM ?

What a funny idea.

Graham

Funnier yet is that he seems to think faster RAM will 'compensate' for
the slower processor. Not on any slow bug I'VE ever used. I think he
may think there are wait states thrown in for 'slow' RAM that can now
be avoided. Good luck on that one.
GG

 

[[email protected]]

We have tried the 8086-2 and it works but the monitor rolls. When
we tried an 8086-1 (or there abouts) from another University it works fine.

Would be interesting to know more about the rolling. Horizontal or
vertical? How fast?
Can you put a scope on the synch signals and see if they are off
frequency, and if so by how much?

Any idea what the monitor driver architecture is?

Can a hold control be adjusted to get the monitor to lock, even at the
wrong frequency?

It just seems like an odd failure mode...

 

[Kryten]

I beg your pardon ! You reckon a 10 MHz CPU needs 15 ns RAM ?

No, I just said I'd got some that fast, and that it might be something he
could try.

It sounds like he is trying to run an 8 MHz rated part at 10 MHz.
If the CPU is slower to set up address lines then this shortens the address
valid time, perhaps to less than the memory permits. Hence if the memory was
faster, it might be able to deliver the data in the shorter valid address
time.

Of course if the CPU cannot run at 10 MHz then the memory speed don't enter
into it.

 

[Winfield Hill]

Kryten wrote...

Pooh Bear wrote ...

No, I just said I'd got some that fast, and that it might be
something he could try.

It sounds like he is trying to run an 8 MHz rated part at 10 MHz.
If the CPU is slower to set up address lines then this shortens the
address valid time, perhaps to less than the memory permits. Hence
if the memory was faster, it might be able to deliver the data in
the shorter valid address time.

Of course if the CPU cannot run at 10 MHz then the memory speed
don't enter into it.

The CPUs he tries seem to be running programs fine, it's the display
screen he's complaining about. But I think your intuition might be
right on target, with respect to the shortened timing requirements.
It's not necessarily access speed, since that was pretty relaxed for
the 8086 family. My concern was with the data bus hold-time issue.

.. CLK ___ ___ ___ ___ ___ ___
.. ___/ \____/ \____/ \____/ \____/ \____/ \__
.. ________ ___________________________________
.. ---X________X___________________________________X---
.. ADDRESS DATA OUT |
.. TWHDX
.. ________________ |___________
.. WR* \________________________________/

I'm thinking that the TWHDX data-hold time, which can be as short
as 20 to 40ns for the 8086-1, can get designs into trouble. Look
at the time scales here. Faster memory might get them out of this
trouble. Of course, the new fast memory may not be pin compatible.
Life is not easy.

 

[Winfield Hill]

Winfield Hill wrote...

Kryten wrote...

The CPUs he tries seem to be running programs fine, it's the display
screen he's complaining about. But I think your intuition might be
right on target, with respect to the shortened timing requirements.
It's not necessarily access speed, since that was pretty relaxed for
the 8086 family. My concern was with the data bus hold-time issue.

. CLK ___ ___ ___ ___ ___ ___
. ___/ \____/ \____/ \____/ \____/ \____/ \__
. ________ ___________________________________
. ---X________X___________________________________X---
. ADDRESS DATA OUT |
. TWHDX
. ________________ |___________
. WR* \________________________________/

I'm thinking that the TWHDX data-hold time, which can be as short
as 20 to 40ns for the 8086-1, can get designs into trouble. Look
at the time scales here. Faster memory might get them out of this
trouble. Of course, the new fast memory may not be pin compatible.
Life is not easy.

Unlike some of the right-wing advocates here on s.e.d., I'm not one
to see things in black and white. While faster ram memory might help
solve WayneL's problem, it's also possible slower memory might be the
solution. That's because some slow memory ICs will store the data-bus
contents 50's of nanoseconds before the end of the WR* strobe pulse.
Slow memory tends to have a longer data-change setup time. In some
designs, slow RAM available when the design was tested might have led
to an unworthy confidence. Later on, new, faster RAM can destroy the
design's safety margins, and render late production a fragile product.
As in so many issues, the precise details determine the exact answer.

Modern designs could potentially suffer from these same issues, but
many employ a power solution, that wasn't available in the old days.

 

[Winfield Hill]

Kryten wrote...

I've packed up an 8086-2 for Wayne L, that will start
its journey to him tomorrow.

It would be great if it would solve his problem!