Researchers create laser light interconnects on silicon

[John Larkin]

<grin> Yep. Among other things. They are still manufactured
but under a different logo.

Really? Got a name or a link?

I program mostly 68K's nowadays, very like a 32-bit PDP-11.

John

 

[Homer J Simpson]

The first one I bought (a PDP-11/35 in a Tektronix SPS system) was
56K bytes of *core*. They didn't make 512byte planes, nor did they
make 512KB PDP-11s, AFAIK.

56K is a lot of core - wasn't it all hand knitted back then by ladies with
tweezers?















...
...

 

[krw]

I started timesharing with a PDP-11/20 with, I think, 32K words (64
kbytes) of core, moved up to an 11/45, and, by the time they fired me,
was running RSTS/E timesharing on a PDP 11/70 with 256k words of
cached dram, a bunch of big disk drives, magtape, DecTape, card
reader, paper tape reader/punch, line printers, modems, and a
full-time operator. We even interfaced an IBM 029 keypunch machine.
The RSTS os would run in native mode or would emulate the RSX-11 or
RT-11 operating systems perfectly.

I guess I'd forgotten (if I ever knew) that PDP-11s had an MMU.
The only ones I used were in the Tektronix systems. The last one I
bought had a /45 (or was it a /44?) in it. I later bought a VAX
11/780 via Fairchild to run our test floor.

 

[krw]

56K is a lot of core - wasn't it all hand knitted back then by ladies with
tweezers?

It was a lot of anything in '76/77. The core came in handy though,
I didn't have a hard disk at first and only had a dual 8" floppy
drive. One could halt the system and power it off. The next day
just power it back on and hit the run switch.

 

[David Brown]

Really? Got a name or a link?

I program mostly 68K's nowadays, very like a 32-bit PDP-11.

John

Which 68k's? I am doing more and more work with ColdFires these days,
which are 68k at heart.

 

[John Larkin]

Which 68k's? I am doing more and more work with ColdFires these days,
which are 68k at heart.

MC68332. They still make them, and tell me that demand is still strong
and that they are soliciting new business. We'll probably cut over to
Coldfire some day... it has 10x or so times the horsepower. But it
seems like, when we need some hard number crunching, we just stuff
that bit into a Xilinx chip. I worked out a DAC calibration equation,
a table lookup, a couple of adds, a multiply and a divide, and one of
my guys said "oh, I'll just do that for you in the FPGA."

The 68332's 32:64 mul/div operations are great.

John

 

[[email protected]]

56K is a lot of core - wasn't it all hand knitted back then by ladies with
tweezers?

I wouldn't call it knitting. It's more like macrame. I never
got to watch them work. They were quite protective of their
area. Males were not able to do this work. I think there's
a name for the tool used. It wasn't tweezers and field service
used it to fix hardware bugs.


/BAH

 

[David Brown]

MC68332. They still make them, and tell me that demand is still strong
and that they are soliciting new business. We'll probably cut over to
Coldfire some day... it has 10x or so times the horsepower. But it
seems like, when we need some hard number crunching, we just stuff
that bit into a Xilinx chip. I worked out a DAC calibration equation,
a table lookup, a couple of adds, a multiply and a divide, and one of
my guys said "oh, I'll just do that for you in the FPGA."

The 68332's 32:64 mul/div operations are great.

John

I'm very fond of the 68332 myself - we have used them as a high-end
microcontroller for over a decade. Motorola (then Freescale) tried for
years to persuade customers to move to more modern devices, but they
remain ever-popular. If you are looking for a step up, go for the
MCF5234 - it is 10-20 times faster, supports more memory, has a faster
and more powerful TPU, ethernet, CAN, and other useful goodies, but is
still very familiar to a 68332 expert.

 

[redbelly]

This isn't just waveguides on a chip, they've also added a light source
right on the chip. That part is different. Most light-emitters are
incompatible with silicon.

Mark

[/QUOTE]

They glued individual compound-semiconductor lasers on top of a
silicon chip. Cute, but it hardly deserves front-page headlines. And
all of the press articles I've seen so far have been loaded with
technical inaccuracies and absurd projections. Intel has a fondness
for this sort of thing, and "science writers" rarely know anything
about what they are writing about.

John

As I understand, even the attaching of non-silicon devices onto a
silicon substrate poses technical challenges and
materials-compatibility issues. Wish I knew more about it to explain
better, but that is what I recall from what I have read.

Agreed that it has yet to merit front-page headlines, but I'm surprised
by the disdain shown by some towards any technological news that does
not pertain to a final product that can be purchased right now.

Mark

 

[krw]

As I understand, even the attaching of non-silicon devices onto a
silicon substrate poses technical challenges and
materials-compatibility issues. Wish I knew more about it to explain
better, but that is what I recall from what I have read.

Thermal coefficient of expansion, for one.

Agreed that it has yet to merit front-page headlines, but I'm surprised
by the disdain shown by some towards any technological news that does
not pertain to a final product that can be purchased right now.

Perhaps because we've seen the same or similar claims far too
often? Pehaps because of the absurd extrapolations? Nah, it'll
make your laptop run cooler and your Internet connection faster.

 

[John Larkin]

As I understand, even the attaching of non-silicon devices onto a
silicon substrate poses technical challenges and
materials-compatibility issues. Wish I knew more about it to explain
better, but that is what I recall from what I have read.

Agreed that it has yet to merit front-page headlines, but I'm surprised
by the disdain shown by some towards any technological news that does
not pertain to a final product that can be purchased right now.

Many, many headlines marveled over how this Intel breakthrough is
going to change the world. Not one mentioned that Intel just dumped
their electro-optics business for about 5% of what it cost them. Not
even Intel has faith in this!

John

 

[[email protected]]

Really? Got a name or a link?

Sorry for the delay. I could not recall the name and had
to ask. Mentec. Note that my statement was second-hand
information. You need to verify.

I program mostly 68K's nowadays, very like a 32-bit PDP-11.

But do they suck?

/BAH

 

[[email protected]]

I ran the RSTS/E timesharing system on the big 11/70.

Ah, for some strange reason RSTS didn't get along with me.
It's possible that we had an awful combination of software
bits. I never had the time to figure that mystery out.

DEC also had...

DOS-11, like it sounds.

RT-11 junior version of DOS. I think DOS inherited a lot from RT-11

I doubt it although I no longer remember who worked on what and
when they did work on it. The similarity among DEC's OSes is
because the same people moved from one software (usually monitor)
project to another project. They also drank beer at the bar
across the street. So one good idea developed in one monitor
would either be carried to the next monitor project or told
to the guy working on another monitor. Then he would carry
the bits, if possible, or the idea to that monitor. That's
how TECO ended up on all systems. It was the most efficient
editor that we used.

In addition, the same group typed all the documentation for
all systems (I was there). Thus, by trail and error, a
common standard of documentation formats and usage were
evolved. DEC was known for its thorough and excellent
documentation.

They fired me in early '79, so it wasn't much later.

'79 is a lot later. I was thinking of when the 11/70 was
first shipped. I don't remember seeing any laying around
in '75.

/BAH

 

[John Larkin]

Ah, for some strange reason RSTS didn't get along with me.
It's possible that we had an awful combination of software
bits. I never had the time to figure that mystery out.

RSTS was very reliable, and very productive, for us. I wrote N/C
compilers and cross-assemblers and all sorts of engineering apps in
BASIC-PLUS. It was the associated big-Unibus systems that tended to be
flakey... the Unibus was a nasty heap of mismatched transmission lines
and wirewrap backplanes. RSTS just used it hard.

Sorry, I meant PC-DOS. Most of the early micro os's, like CPM and
such, pretty much resembled PDP-11 DOS and RT11.

John

 

[John Larkin]

Sorry for the delay. I could not recall the name and had
to ask. Mentec. Note that my statement was second-hand
information. You need to verify.


But do they suck?

Nope. The 68332 is a solid machine, a pleasure to design with and
program.

John

 

[[email protected]]

On Mon, 25 Sep 06 11:59:51 GMT, [email protected] wrote:

Sorry, I meant PC-DOS. Most of the early micro os's, like CPM and
such, pretty much resembled PDP-11 DOS and RT11.

And resembled OS-8. IAS was the best PDP-11 OS, from my limited
usage.

/BAH

 

[joseph2k]

I'd be guessing 512 bytes if it was core.

If you were talking 1965 it would be 512 16-bit words. In 1968 i got to
visit a core memory manufacturer and they had 4k 32-bit words about the
size of a paperback, the drive electronics was about 2 to 3 times as big.
By 1975 i was working on computers with 64K 32-bit words in a module not
much bigger than my fist, and the drive electronics was about 2 X the size
of the core module. Of course heat extraction is becoming a problem at
this point.

 

[joseph2k]

I started timesharing with a PDP-11/20 with, I think, 32K words (64
kbytes) of core, moved up to an 11/45, and, by the time they fired me,
was running RSTS/E timesharing on a PDP 11/70 with 256k words of
cached dram, a bunch of big disk drives, magtape, DecTape, card
reader, paper tape reader/punch, line printers, modems, and a
full-time operator. We even interfaced an IBM 029 keypunch machine.
The RSTS os would run in native mode or would emulate the RSX-11 or
RT-11 operating systems perfectly.

We bought scores of PDP-11s for pipeline control systems, mostly run
under the REX rtos authored by, um, me.

John

If you still have it you should consider making a copy for a computer
museum.

 

[joseph2k]

56K is a lot of core - wasn't it all hand knitted back then by ladies with
tweezers?

..
..

Actually by 1968 they had invented machines that could do it better than
humans. Of course with core sizes shrinking to reduce switching energy and
thus power, it was getting nigh impossible for humans to do anymore.