• Please review our updated Terms and Rules here

At Last, The Six Processor IBM PC Server 720... Wait, the WHAT?

ButINeededThatName

Experienced Member
Joined
Nov 29, 2018
Messages
176
Location
Fort Wayne, IN
So this past Sunday I made an almost nine-hour round trip to pick up a lot of MicroChannel gear for the bargain of $300. Honestly at that price I really couldn't turn it down. So, one round-trip to the Cincinnati area later and I had come home with the following;

A very sad looking and dirty Model 60 (though I've definitely cleaned worse)

IMG_20200825_091546-min.jpg
IMG_20200825_091533-min.jpg

A box of SCSI cables and a box of power cables as well as filler trays and drive rails for the Model 95

IMG_20200825_091157-min.jpg

A pair of complete Hot-Plug racks for my Model 95A (FINALLY). I can't tell you how long I've been looking for these

IMG_20200825_091106-min.jpg

A seemingly NOS 768 Watt Astec PSU, as well as a box of SCSI trays and three SCSI backplanes for the PC Server 500 / 720

A... hold on a minute... A PC Server with six processors? What?

IMG_20200824_154444-min.jpg
IMG_20200825_092711-min.jpg

Okay, joking aside I actually had no idea the PC Server 720 was a multi-processor system. From what little information I had about it, I assumed it was effectively a PC Server 500 that just had PCI slots. Turns out I was completely wrong and this in fact is a system capable of taking up to six processors (either 100 / 133 or 166Mhz Pentiums) and 512Mb of ECC memory, or up to five processors and 1Gb of ECC memory. It also has eight combination MicroChannel / PCI slots and much like it's brother, the Server 500, it can take up to eighteen hard-disks. Mine also as a CD-ROM drive, DDS drive, a DDS tape and some floppies that undoubtedly have system software and a backup on them and even the key for the front panel.

I have yet to power it on and test it (want to clean it and check for bad capacitors first), but as of yet I know very little about it. Aside from the page covering parts of it over on Ardent Tool, I've found very little information on specific stuff, like how it utilizes more than one processor. I doubt it functions like SMP as we know it now.

Anyways, I'll be sure to update this post once I clean it, check it and attempt to power it on. So far though, this is looking like it's possibly my greatest vintage computing find!

Oh, and here's a photo of the diagram on the side panel for those who are interested;

IMG_20200825_092800-min.jpg
 
Definitely a very interesting machine.

I wonder how all of the CPUs are arranged on the bus, if they're all on a shared FSB, or have dedicated buses to a central arbiter of some sort.

The ALR 6x6 had six Pentium Pro 200 CPUs on a shared front side bus, which caused horrible bottlenecks. The server was great at number crunching, but not moving large quantities of data around.
 
So that's a Unix machine or Will NT 4 run on 6 x P1's?

I kind of wanted the ALR (or later Gateway) 6 PPro server but shipping was a deal killer back then (and I had no real use for one).
 
Definitely a very interesting machine.

I wonder how all of the CPUs are arranged on the bus, if they're all on a shared FSB, or have dedicated buses to a central arbiter of some sort.

The ALR 6x6 had six Pentium Pro 200 CPUs on a shared front side bus, which caused horrible bottlenecks. The server was great at number crunching, but not moving large quantities of data around.

I'm not entirely sure how they're arranged, but apparently this system supports mix and matching different speed processor cards.

I've always been interested in the ALR systems. They seemed pretty unique.


So that's a Unix machine or Will NT 4 run on 6 x P1's?

I kind of wanted the ALR (or later Gateway) 6 PPro server but shipping was a deal killer back then (and I had no real use for one).

Not that I'm aware of. Seems like your choices for operating systems were OS/2 2.11+ or NT 3+. Apparently it shipped with DOS though.
 
Alright, so after a bit of troubleshooting I can safely say the Server 720 clears POST! I had some issues with the PCI bridge card causing various errors at different stages of POST, though cleaning its slots and jumping the "Program Enable" jumper seems to have cleared it up (I have no idea what that jumper did by the way, though its default state is supposedly not jumped). I am also happy to say that all six processors work and pass verification! Now I can move onto the "fun part" of configuring SCSI devices.

Currently I'm trying to configure the system with two "Cheetah" (SCSI-2 F/W Streaming RAID Adapter/A) SCSI / RAID adapters in the following configuration:

- Fixed Disks 1 thru 6, Bank C on Adapter One, Channel One, ID's 0-5, in a RAID 0 Array
- CD-ROM Drive on Adapter One, Channel One, ID 6
- DDS Tape Drive on Adapter One, Channel One, ID 7

- Fixed Disks 7 thru 12, Bank D on Adapter Two, Channel One, ID's 0-5
- Fixed Disks 13 thru 18, Bank E on Adapter Two, Channel Two, ID's 8-D
- Fixed Disks 7 thru 18 in a RAID 5 Array

So far though I haven't made much progress on this front as adapter one complains that drives 1,2,3,4,7 and 7 (not sure why there are two drive sevens) are unresponsive and the system hangs shortly after. As of yet I'm not sure what the issue is though I'm sure I'll figure it out in time. I'm also currently unsure of how I'm going to load the system partition onto the machine, as I believe you can't IML off of a RAID (I know the 95A with a PasssPlay can't).
 
Last edited:
Alright, so after a bit of troubleshooting I can safely say the Server 720 clears POST! I had some issues with the PCI bridge card causing various errors at different stages of POST, though cleaning its slots and jumping the "Program Enable" jumper seems to have cleared it up (I have no idea what that jumper did by the way, though its default state is supposedly not jumped). I am also happy to say that all six processors work and pass verification! Now I can move onto the "fun part" of configuring SCSI devices.

Currently I'm trying to configure the system with two "Cheetah" (SCSI-2 F/W Streaming RAID Adapter/A) SCSI / RAID adapters in the following configuration:

- Fixed Disks 1 thru 6, Bank C on Adapter One, Channel One, ID's 0-5, in a RAID 0 Array
- CD-ROM Drive on Adapter One, Channel One, ID 6
- DDS Tape Drive on Adapter One, Channel One, ID 7

- Fixed Disks 7 thru 12, Bank D on Adapter Two, Channel One, ID's 0-5
- Fixed Disks 13 thru 18, Bank E on Adapter Two, Channel Two, ID's 8-D
- Fixed Disks 7 thru 18 in a RAID 5 Array

So far though I haven't made much progress on this front as adapter one complains that drives 1,2,3,4,7 and 7 (not sure why there are two drive sevens) are unresponsive and the system hangs shortly after. As of yet I'm not sure what the issue is though I'm sure I'll figure it out in time. I'm also currently unsure of how I'm going to load the system partition onto the machine, as I believe you can't IML off of a RAID (I know the 95A with a PasssPlay can't).

I believe these systems don't need an IML provided they have a complex with an enhanced bios. My PC Server 500 does not need an IML partition, it loads just fine from the Cheetah (or from other SCSI drives). Unresponsive drives is a common fault with the Cheetah. M Mine was forever dropping drives out of the RAID 5 arrays. I scrapped the RAID 5 and popped in a couple of 72Mb U320 drives (you need the 68-pin version) and mirrored them. Its fine with those. Often the problem is the drive status cable which has a small (10pin?) IDC on one end and a funny connector on the other. Somewhere there is an IBM document on recovering.

More info here.

http://www.tavi.co.uk/ps2pages/ohland/raid_cheetah.html
 
oh that's a thing of beauty.

gotta love that foam from the Model 60/80. I had the same issue and even hours of work, and it's STILL sticky on the inside.
 
I've found very little information on specific stuff, like how it utilizes more than one processor. I doubt it functions like SMP as we know it now.

The PC Server 720 used an architecture IBM licensed from a third party called the "Corollary C-bus II"; several other machines were based on the same design, particularly some Olivetti and ALR products. I remember there being a discussion about this on... some forum, I don't remember if it was this one, and the long and short of it *I believe* is it was a proprietary hack to Intel's MultiProcessor Extension to support more CPUs at the cost of making it a little less symmetric. (IE, some limitations on secondary CPUs handling interrupts or something, documentation for it is extremely scarce.) Apparently at least SCO Unix and early versions of Windows NT could run on it in some capacity, but you're probably SOL with anything much newer.
 
Back
Top