On Wed, 8 Aug 2001, Heinz Wolter wrote:
> I don't unfortunately have one... but the IBM ISA card
> available for the XT allowed that lowly machine to
> pretend it was a 370 class mainframe....in the early 80's
> (I worked for a company that got all the latest IBM wonders...)
It didn't make a PC *pretend* it was an S/370, it used the PC's
motherboard and I/O system as a backplane and accessory panel to an ACTUAL
REAL S/390 processor board.
> PC/370 it was called if I remember correctly....
Correction. P/370.
> made with IBM re-microprogrammed 68K silicon..
> Later more powerful (5-7 mips) versions came out
> for MC ad PCI named PC/390 after the ES series mainframes.
The 68K just ran as a general purpose bus-I/O interface to the PC. The
P/390 uses a proprietary chip that serves the same purpose. The P/370 ran
at 0.8-1.3 (roughly) mainframe MIPS. This has nothing to do with the MIPS
to which most people refer, because the mainframe MIPS is measured using
VLIW instructions. The MCA and PCI P/390 run at 3.5-4 mainframe MIPS.
The brand new P/390E, which incidentally is still being manufactured and
lists for $15,000+ and is the basis for the Multiprise/3000, runs at 7-7.5
mainframe MIPS. Incidentally, the P/390 MCA, P/390 PCI, and P/390E can
all be run alternately (and quicker) in an RS/6000 with that particular
kind of slot.
To rehash, the P/370 is an 8-bit XT-bus card that runs at 0.8 MIPS and has
32 MB main storage with no extended storage (mainframes divide main memory
between main storage and extended storage, kind of like an Amiga, and call
disk either minidisk -> temporary disk, or DASD -> main disk), the P/390
MCA is a 32-bit Microchannel card that runs at 3.5 MIPS with 32 MB main
storage with a daughtercard connection for either a 32 MB or 96 MB
daughtercard, which added that much additional main storage and no
extended storage. The P/390 PCI is a 32-bit busmastering PCI card that
runs at 3.5 MIPS, and has 128 MB main storage, can use PC (or RS/6000)
memory as extended storage. The P/390E operates at 7 MIPS, has either 256
MB or 1 GB of main storage, and can also use PC or RS/6K memory as
extended storage.
> Tthe lowly XT, with it's MFM 10 meg drive would IPL the real
> machine and act as a channel processor, DASD (disk), and 3278 I/O.
> In other words, what Intel chips were meant to do, and no more!
Not exactly. On the P/370, and the P/390 MCA, the DASD would be an
external mainframe DASD box (containing large-platter HDDs on a 3380/3390
interface). The internal HDDs of the PC or RS/6000 would be used for
minidisk (in addition to virtual minidisk on DASD). The P/390 PCI and
P/390E can use PC and RS/6K disk as DASD, and can emulate a 3390 9-track
tape drive using a SCSI DAT.
> That same company also had a beautiful real 68k based
> IBM Instruments Scientifc (unix) computer... too bad IBM opted
> for the Intel platform instead, punishing us forever to
> a world of 64k segments, crappy architecture, small
> medium, large and linear model compilers, Messy-DOS,
> Winblows and all it's breathern... when we might have
> have 68Ks and unix...It would have been a different world.
> Thanks IBM!
No offense, but bullshit. IBM still makes UNIX workstations based on the
POWER architecture even today. Before the POWER machines, it was the
RT/PC based on the ROMP. The IBM IISC was largely an experimental CISC
workstation exercise. Have you ever seen a POWER2SC or POWER3II or even
an old POWER RS/6000 running (PowerPC RS/6Ks don't count, they're
el-cheapo)? They're surprisingly fast. Everything is coprocessed. They
were designed to do CAD, and they excel at it. CATIA was dreamt up and
implemented on RT/PCs and RS/6000s.
> Anyone have either an IBM Instruments Scientific Computer or PC/370
> card ? Both are rare, while PC/390 systems come up on ebay sometimes...
P/370s are FAR from rare. It's just they are largely still being used in
production environments. They are pure mainframe hardware. They never
*ever* crash. If you don't need to replace one, you just don't. I have
four P/390s running myself, along with an S/390 9672 G5.
Peace... Sridhar
> Heinz
P.S. I am an IBM employee, working on low-level mainframe server
technology manufacturing. Here's my sigfile from work.
--
Sridhar Ayengar sayenga_at_us.ibm.com
Tool Control Software Development Sridhar Ayengar/Fishkill/IBM_at_IBMUS
Interconnect Products (845)892-3496
IBM Corp.
East Fishkill Facility, Hopewell Junction, NY
Received on Wed Aug 08 2001 - 00:43:12 BST