Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > alt.folklore.computers > #157126 > unrolled thread

360/85

Back to article view | Back to alt.folklore.computers

Contents

  360/85 Jon Elson <elson@pico-systems.com> - 2016-01-19 23:51 -0600
    Re: 360/85 Quadibloc <jsavard@ecn.ab.ca> - 2016-01-20 03:13 -0800
      Re: 360/85 Jon Elson <elson@pico-systems.com> - 2016-01-20 12:09 -0600
        Re: 360/85 Anne & Lynn Wheeler <lynn@garlic.com> - 2016-01-20 10:26 -0800
        Re: 360/85 Quadibloc <jsavard@ecn.ab.ca> - 2016-01-20 10:40 -0800
          Re: 360/85 scott@slp53.sl.home (Scott Lurndal) - 2016-01-20 19:43 +0000
          Re: 360/85 Jon Elson <jmelson@wustl.edu> - 2016-01-20 13:55 -0600

#157126 — 360/85

I think we had some discussion here about the 360/85 quite some time ago.

There seem to be a few bits of info on the net that are wrong, indicating 
that the /85 used ASLT components.  I finally ran across something more 
definitive.

"IBM's 360 and Early 370 Systems"  By Emerson W. Pugh, Lyle R. Johnson, John 
H. Palmer

This book indicates the model /85 was the first to use MST-4 chips, same as 
the 370/165.

As soon as I came across the incorrect technology reference, I suspected it 
couldn't be right, as the model /85 had 500 words of writable control store.
500 words of 108 bits is 54,000 bits.  With ASLT being made of discrete 
transistors on double-stacked SLT ceramic substrates, and assuming a storage 
bit could be squeezed into 4 transistors, that would be over 200,000 
transistors.  That would be a HUGE chunk of hardware, and just doesn't make 
sense.

With the info that the /85 was using MST-4 (and the earliest memories were 
made with 16-bit ECL chips) you could get this down to ~3400 chips.  If you 
put 2 chips into each MST package, that would fit in 1700 packages, and 
maybe could be packed onto about 100 small MST cards.  While that is a 
pretty big piece of hardware, it makes sense it could be fitted into such a 
machine.

Another reason I suspected that the /85 and /165 were VERY similar is if you 
look at the instruction timings, they appear almost to be the same pages, 
duplicated.  the timings for almost EVERY instruction is identical down to 
the same number of clocks.

Jon

[toc] | [next] | [standalone]

#157139

On Tuesday, January 19, 2016 at 10:51:11 PM UTC-7, Jon Elson wrote:

> Another reason I suspected that the /85 and /165 were VERY similar is if you 
> look at the instruction timings, they appear almost to be the same pages, 
> duplicated.  the timings for almost EVERY instruction is identical down to 
> the same number of clocks.

Indeed. As I've noted, the length of the microcode word is the same between the 
360/85 and the 370/165 - and the 3033: internal microcode is 108 bits wide, but 
writeable microcode is 126 bits wide.

In physical appearance, the console of the 85 is very much like that of the 
later 165; two large microfilm readers, one of which overlays front panel 
lights, and a CRT.

"IBM's 360 and Early 370 Systems" is definitely the appropriate definitive reference to use for the technology.

Of course, it says that the 360/85 was "the first product shipped with MST-4 
circuits"; possibly the writeable control store was the only part of it made 
with them, and the CPU logic was still made from ASLT. The reason I suggest 
this possibility is that the 360/85 took up a *lot* more space than the later 
370/165, which seems odd if they were implemented using the same type of chip.

John Savard

[toc] | [prev] | [next] | [standalone]

#157174

Quadibloc wrote:


> Of course, it says that the 360/85 was "the first product shipped with
> MST-4 circuits"; possibly the writeable control store was the only part of
> it made with them, and the CPU logic was still made from ASLT. The reason
> I suggest this possibility is that the 360/85 took up a *lot* more space
> than the later 370/165, which seems odd if they were implemented using the
> same type of chip.
Well, I'm not sure it was a *LOT* more space.  The /85 was also the FIRST 
indirectly water-cooled machine IBM made.  There could have been a lot of 
conservative design in that to make sure they didn't have a thermal 
catastrophe, and then later decided they could safely increase density.

Since even the early MST was WAYYY higher density than ASLT, I think the 
equivalent machine architecture would have been a LOT bigger in ASLT.
Each MST gate was at least 8 transistors.  You could get, MAYBE, 12 
transistors in a single ASLT can.  If you had 4 gates per MST can, that 
would be 32 MST transistors in the space of 12 ASLT.  Or, in gates, 4 MST 
vs. MAYBE 2 per ASLT.  When you get to flop-flops, expecially common-clocked 
D-flops, the difference gets bigger, like maybe 4:1.

Also, I think Pugh et al. at least IMPLIES the entire machine was MST-4.
The /85 was built with core memory one generation older than the /165, that 
could make the difference right there.  Core was getting shrunk periodically 
during that time.  Also, they might have moved to a slightly more dense 
memory chip, going from 16 bits/chip to 64 bits/chip or something, which 
would have shrunk the WCS and buffer memory substantially.  I think the /165 
used the more dense SRAM chips.

Jon

[toc] | [prev] | [next] | [standalone]

#157179

Jon Elson <elson@pico-systems.com> writes:
> Well, I'm not sure it was a *LOT* more space.  The /85 was also the FIRST 
> indirectly water-cooled machine IBM made.  There could have been a lot of 
> conservative design in that to make sure they didn't have a thermal 
> catastrophe, and then later decided they could safely increase density.

There was 370/165 that had thermal problem .... turns out that they
didn't have flow sensor on the external water ... by the time thermal
sensor started to show a rise ... it was too late to prevent meltdown.
After that, they added flow sensor to the external water.

-- 
virtualization experience starting Jan1968, online at home since Mar1970

[toc] | [prev] | [next] | [standalone]

#157183

On Wednesday, January 20, 2016 at 11:09:59 AM UTC-7, Jon Elson wrote:

> Well, I'm not sure it was a *LOT* more space.

I had to hunt for a picture - this one is of a 370/168, which has VM added -

http://www.jklp.org/profession/books/mix/c10.html

showing the processor cabinetry, not just the front panel;

while it's shrunk in appearance by foreshortening, it's still clear enough how 
big it is; it is larger than I expected. But it is still smaller than the 
360/85,

http://www.techrepublic.com/pictures/photos-looking-back-to-the-birth-of-the-mainframe/2/

John Savard

[toc] | [prev] | [next] | [standalone]

#157191

Quadibloc <jsavard@ecn.ab.ca> writes:
>On Wednesday, January 20, 2016 at 11:09:59 AM UTC-7, Jon Elson wrote:
>
>> Well, I'm not sure it was a *LOT* more space.
>
>I had to hunt for a picture - this one is of a 370/168, which has VM added -
>
>http://www.jklp.org/profession/books/mix/c10.html

Hmm. Some of the burroughs information this book is incorrect.

The b[234]x00 systems didn't follow the 'general architectural design'
of the B5000/B5500.  They're actually radically different machines.

[toc] | [prev] | [next] | [standalone]

#157193

Quadibloc wrote:

> On Wednesday, January 20, 2016 at 11:09:59 AM UTC-7, Jon Elson wrote:
> 
>> Well, I'm not sure it was a *LOT* more space.
> 
> I had to hunt for a picture - this one is of a 370/168, which has VM added
> -
> 
> http://www.jklp.org/profession/books/mix/c10.html
> 
> showing the processor cabinetry, not just the front panel;
OK, I was thrown off as the two things that looked like the microfiche 
reader that was part of the CE on-site gear for reading maintenance 
documents were turned so they were not visible in that photo.

<http://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=images&cd=&cad=rja&uact=8&ved=0ahUKEwj_kbzJlLnKAhXBJiYKHWIlCPgQjRwIBw&url=http%3A%2F%2Fhistory.cs.ncl.ac.uk%2Fanniversaries%2F40th%2Fimages%2Fibm370_1682%2Fprint02.html&psig=AFQjCNHlUUSLTwl-4USSjnGbcPbZ3ohpuQ&ust=1453405770694186>
> 
> while it's shrunk in appearance by foreshortening, it's still clear enough
> how big it is; it is larger than I expected. But it is still smaller than
> the 360/85,
> 
> http://www.techrepublic.com/pictures/photos-looking-back-to-the-birth-of-
the-mainframe/2/
Some docs I've looked at show the /85 set up just as your 2nd picture,


||      ||      ||
||      ||      ||
||      ||      ||
||      ||      ||
||------||------||
||      ||      ||
||      ||      ||
||      ||      ||
||      ||      ||


And, the 165/168 were set up like the following in the installation planning 
guide that is on bitsavers.  Your pic might be configures with less main 
storage, thus two of the wings are missing.


||         ||
||         ||
||         ||
||         ||
||---------||
||    |    ||
||    |    ||
||    |    ||
||    |    ||
      |
      |
      |
=============

So, the amount of backplane/cards seems to be about the same in both cases, 
6 "wings".  Not sure if these wings were the same size, but I'll bet they 
were.  SLT and MST used very similar backplane boards and component cards.

Jon

[toc] | [prev] | [standalone]

Back to top | Article view | alt.folklore.computers

csiph-web