My barn-find IBM mini

[HoJoPo]

It turns out to be a serial handshaking problem. I dropped the console speed to 9600bps and it works just fine.

In fact, I am typing this on the serial console right now.

View attachment 29429

Either Links does something to the handshaking (how?? why??) or it's just a coincidence that I never ran into a problem before, and my handshaking settings are incorrect anyway.

It's too bad this thing is such a power hog. I sure could get used to using it. I'd probably have to add a boatload of serial ports though. I'll have to see if I can determine how much power it uses; but it must be a lot. Although, it's probably not as much as my A3000 used to draw, and I left that on for years.

Handshaking makes sense, a lot of the older terminals don't like much more than 9600. If you are using XON/XOFF instead of hardware handshaking, it may not be responding fast enough at the higher rate.

You could run NetBSD on a Raspberry Pi with a USB to serial adapter, making a wireless terminal. Much lower power draw, and probably faster... you could probably even power the Pi from the terminal, or, if you were really creative, conceal it inside. It wouldn't be as authentic... though you could occasinally fire up the old iron and SSH to it from the now wireless terminal. :D

 

[NeXT]

Look at the psu rating. It's probably where somewhere between 450 and 500 watts. That is not too bad in comparison to modern gaming machines and consoles.

 

[KC9UDX]

Look at the psu rating. It's probably where somewhere between 450 and 500 watts. That is not too bad in comparison to modern gaming machines and consoles.

It has three identical power supplies!

 

[GiGaBiTe]

The thing that makes those old big irons power hogs are the SCSI drives. The CPUs themselves don't use that much power, I used to have an NEC big iron with dual Pentium Pro 200s and without the SCSI drives, it'd run just fine on a 350W PSU.

The trouble you had getting newer distros running on that machine may be due to bad memory or a bad memory controller. I've had plenty of old machines that I tried installing Linux/BSD on which would constantly fail, only to figure out that there was bad memory in the machine or in rare cases, a bad memory controller. Windows on the other hand would run absolutely fine on those machines and some of them had up to a couple of megabytes of faulty memory.

 

[ClassicHasClass]

It has three identical power supplies!

I love IBM redundancy.

Anyway, break out the Kill-A-Watt, man. You could probably reduce the draw somewhat by going down to two redundant supplies, or maybe just one; each PSU draws some additional nominal juice as well as spreading the load out. When I put the second redundant supply in my POWER6, it added about 40W to its total draw (but it's a server, so I'll pay that penalty gladly).

 

[KC9UDX]

They say 7A @ 125V, which, there's no way they continuously draw that much. I have the server, LCD monitor, and serial terminal on a dedicated 20A breaker, which doesn't get hot. I could not power this setup on a 15 A breaker with room lighting, whatever that actually means. I'm guessing these are somewhere in the 400W range.

I don't have a Kill-A-Watt, but I can measure current; not easily, and likely won't bother.

One power supply is failing, I think. There are two LEDs on the back of each, and on that one, when the server is powered up, one goes out. And, I get a P/S warning light on the front; and an unknown warning light on every drive bay for the 2nd SCSI bus.

I did a bunch of memory testing today and once two bytes showed bad, but I haven't seen that happen more than once. If I hadn't just shut it off, I'd be posting this there; I was using it all day.

My terminal occasionally slips up at 19200bps, but not often. If I get ambitious, I'll investigate why the handshaking is failing.

I opened it up and got the 3.5" floppy working. I found that the heat sinks released themselves from the two processors, and were lying on the floor. The processors weren't hot; there's a ton of airflow in there. So, I installed CPU#3, with one of the heatsinks, and said to heck with the other two.

I hate to say it, but I'm really starting to like the thing, despite that it's a run-of-the-mill x86.

And, I'm getting to the point where I appreciate NetBSD more than I ever appreciated Solaris, which says a lot. I kinda wish I'd kept an Ultrasparc and run NetBSD on it. But on the other hand, I don't have room to have kept one.

 

[ClassicHasClass]

One power supply is failing, I think. There are two LEDs on the back of each, and on that one, when the server is powered up, one goes out.

That may just mean the system only supports 1+1 PSUs, and the third extraneous power supply is recognized and then shut down as unneeded. Maybe try swapping it and seeing if it acts differently?

I get a P/S warning light on the front;

Every IBM box I've ever had flashes a "power problem" light on just about every home circuit. My POWER6 did it from the very first day I bought it, 6 years ago.

and an unknown warning light on every drive bay for the 2nd SCSI bus.

Termination issue, maybe?

Love how the system runs so well even without the heatsinks ...

 

[KC9UDX]

That may just mean the system only supports 1+1 PSUs, and the third extraneous power supply is recognized and then shut down as unneeded. Maybe try swapping it and seeing if it acts differently?

I tried powering them up individually and got the same results. So unless the power supply in the middle is somehow configured to be the extra-redundant supply, I think it's just a failing supply. I could do some more playing round with that though. Well, I could until last night, come to think of it. I can't freely get at the back of the machine anymore.

Every IBM box I've ever had flashes a "power problem" light on just about every home circuit. My POWER6 did it from the very first day I bought it, 6 years ago.

I wonder what it doesn't like? On this one, the "power problem" light is always on or off coincident with the opposite state of the second LED on the suspect power supply.

Termination issue, maybe?

Possibly. I unfortunately haven't noticed if this indication indeed coincides with the power problem light. In either case, sometimes they are on, sometimes they are not. But when they are on, they are all on.

Love how the system runs so well even without the heatsinks ...

It kind of bothers me to run it like that. But, since it was apparently running like that for a long time anyway, and, doesn't seem to get hot, I'll leave it. In the worst case scenario, I end up with two dead Pentium Pros. I could make a heatsink, I suppose. But, there is a ridiculous amount of air circulation inside that case.