From phil@ultimate.com Thu Jul 26 11:43:37 2001
From: phil@ultimate.com
To: test-drb@ccmp.vtda.org
Subject: heck, are 11/785s extinct?
Date: Thu, 26 Jul 2001 11:43:37 +0000
Message-ID: <200107261643.f6QGhbQN003291@ultimate.com>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============1137295937583668097=="

--===============1137295937583668097==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable

> Date: Thu, 26 Jul 2001 07:02:22 -0700 (PDT)
> From: Ethan Dicks <ethan_dicks(a)yahoo.com>
>
> > The 11/782 was dual cpu 11/780's with shared memory bus run=20
> > master-slave for compute bound tasks with the i/o run off of one
> > cpu.
>=20
> Right.  Developed by George Goble (of the LOX-on-the-BBQ fame), et. al, at
> Purdue, IIRC.

The 782 was the DEC solution, using a multiported memory (MA780).
Each CPU had it's own SBI, with private UBA's and MBA's (UNIBUS and
MASSBUSS adapters);

<a href=3D"http://www.montagar.com/dfwcug/VMS_HTML/timeline/1982-5.HTM">VAX/V=
MS V3</a>

   VMS V3 supported three new processors: the VAX-11/750,
   VAX-11/725, VAX-11/782. V3 features included asymmetric multiprocessing
   (ASMP) for VAX-11/782, support for new architectures, protocols and
   busses, system communication architecture (SCS), mass storage control
   protocol (MSCP), lock management system services, and MONITOR utility
   for performance monitoring.

<a href=3D"http://www.netbsd.org/Documentation/Hardware/Machines/DEC/vax/vax7=
00.html">VAX hardware reference - VAX 700 series</a>;

   VAX 11/782
   Nickname: Atlas Cpu: Dual KA780
   Vup: 1.8
   Bus: 2 SBIs with 15 nexus slots each. Each slot can take a UBA (UNIBUS
   adapter), MBA (MASSBUS Adapter), or CI. (?)
   Introduced: 1982 NetBSD: (?) Probably

The Purdue work connected the two CPU's by replacing the SBI bus
terminator with the second CPU.  There aparently were also a few
11/784s, with 4 CPU's, also using the MA780.

http://www.cam.anglia.ac.uk/~systimk/History/Vaxes.Txt, discusses it all,
an includes a 1993 message from George.

> My understanding was that the 11/785 was DEC's official implementation
> of what George and crew cobbled together as the 11/782.  It only saw
> the light of day because, as you say, there was a gap in DEC's product
> line with the delays in the 8600 and customers wanted more than what was
> presently out there.

The 785 was a late life kicker to the 780;
It was not (by itself) a multiprocessor system.

<a href=3D"http://www.montagar.com/dfwcug/VMS_HTML/timeline/1984-2.htm">VAX-1=
1/785</a>
   CPU cycle time in the VAX-11/785 was 133ns, 50% faster
   than the 200ns cycle time of the VAX-11/780. The accelerated cycle time
   allowed all CPU operations to run up to 50% faster, resulting in higher
   throughput, faster response time and the ability to support more users

The NetBSD VAX hardware reference says that a 782 built from two 785's
would have been called a "787"

Disclaimer;
	I worked at DEC in the 80's, but on 36-bit products.
	(and briefly on the 64-bit RISC machine, codenamed "SAFE",
	which, after Cutler grabbed the project, was renamed PRISM).

--===============1137295937583668097==--


From pechter@bg-tc-ppp1603.monmouth.com Thu Jul 26 12:44:00 2001
From: pechter@bg-tc-ppp1603.monmouth.com
To: test-drb@ccmp.vtda.org
Subject: heck, are 11/785s extinct?
Date: Thu, 26 Jul 2001 12:44:00 +0000
Message-ID: <200107261744.f6QHi0n03537@bg-tc-ppp1603.monmouth.com>
In-Reply-To: <200107261643.f6QGhbQN003291@ultimate.com>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============1081188720077613831=="

--===============1081188720077613831==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

> > Date: Thu, 26 Jul 2001 07:02:22 -0700 (PDT)
> > From: Ethan Dicks <ethan_dicks(a)yahoo.com>
> >
> > > The 11/782 was dual cpu 11/780's with shared memory bus run 
> > > master-slave for compute bound tasks with the i/o run off of one
> > > cpu.
> > 
> > Right.  Developed by George Goble (of the LOX-on-the-BBQ fame), et. al, at
> > Purdue, IIRC.
> 
> The 782 was the DEC solution, using a multiported memory (MA780).
> Each CPU had it's own SBI, with private UBA's and MBA's (UNIBUS and
> MASSBUSS adapters);

Actually, DEC didn't (back then) support any massbusses on the attached
cpu...  They dual-ported the memory in between the sbi's.
The attached cpu usually ran no I/O devices and just worked on compute
bound tasks... this was in the vms 3.2 or so range.

Perhaps later they let the attached box have private disks.

> Disclaimer;
> 	I worked at DEC in the 80's, but on 36-bit products.
> 	(and briefly on the 64-bit RISC machine, codenamed "SAFE",
> 	which, after Cutler grabbed the project, was renamed PRISM).

Hmm... how close was it to Alpha AXP -- almost exactly prism?

Bill

---
  Bill Gates is a Persian cat and a monocle away from being a 
  villain in a James Bond movie              -- Dennis Miller 
  bpechter(a)shell.monmouth.com|pechter(a)pechter.dyndns.org

--===============1081188720077613831==--