[bnoji]

I acquired a base model Octane (CMNB015ANF175) this weekend along with a spare PSU. The previous owner said it was in use, froze while running and wouldn't boot again after that. He bought a new PSU and it still didn't work.

I plugged it in and both the chassis fan and PSU fan spin up for a second until the relay clicks and it shuts off and sits idle with a slight buzz. It sounds (and smells) like a short somewhere. I pulled the power supply out but I can't tell if that smell was from the chassis or the PSU. I opened the PSU and there's nothing obviously smoked and both fuses were fine.

The other PSU doesn't do anything so I'm assuming that's the bad one.

Are there any points on the boards where shorts are "common"? I'd hate to buy a third power supply only to find out it's a short on one of the boards instead.

[weblacky]

You're walking into an area that hasn't really been well documented yet. Octanes have only started to do this in the past 10 years or so. They're now old enough that rebuilding is going to start becoming a thing. But no one has yet taken it on as there are many more expensive SGI systems ahead of it that people will actually pay for rebuilding.

Octane parts are still relatively cheap, and so unfortunately the repair versus replaced with another used part is a real battle on the older systems.

There is no common mainboard or XBOW failures that are known as a quick fix on Octane. We have seen a few octane's that had a failure inside and not the power supply. 98% of the failures we hear about are normally the power supply. They normally result in what we call the click of death. Where you actually plug everything in and you just hear a very faint click and nothing else comes to life. This is normally an overcurrent protection in the power supply triggering because the power supply itself has a problem.

In your case if you have the fan starts on both the chassis and the power supply then the problem is likely not the power supply. On systems like that we've seen scorch marks on the computer side of the PSU power connector when it gets really bad.

The octane is actually a very complex machine (that runs very hot) with no monitoring, except for a primitive thermal system for the graphics that changes fan speed to maximum. Later systems have a complex environmental monitoring system to try to compensate for these types of symptoms. Unfortunately octane just doesn't give you any feedback.

You're going to have to try the modular approach. As you can understand there's basically four standard modules in octane. The power supply, the chassis with the midplane inside (known as the XBOW), the system mainboard module and the peripheral/graphics module. Anything that's modular slides out with a tooless configuration.

What you're experiencing has to be in either the motherboard, the midplane, or the peripheral/graphics module.

From what I know is the octane itself has absolutely no cut offs that are resettable, outside of the polyfuses that SGI uses but those have a tendency to just explode. It is a software controlled PSU start however and a member here has decoded the circuit for that. But obviously if you can start your octane then that circuit is not the problem.

Since your fans are starting I'm going to make the assumption that it's probably not your 12V line. This isn't a big leap because SGI motherboards and most SGI graphics cards don't really use a 12V DC line they tend to use enormous 5V and 3.3 V lines.

It sounds to me like some capacitor or similar component has gone short on one of your modules.

There are no pin outs, there are no diagrams, there are no boardviews. The best way you're going to accomplish this is remove power from the system, remove the power supply from the system and then remove the system board and finally remove the graphics/peripheral module.

Use a multimeter and actually use diode mode and ohms mode to check all the coils on the mainboard (assuming you find any) looking for power conversion (VRM) rails. And you'll want to randomly probe around the VCC pin of any chips you can identify. Just find some random small easy to access semiconductors and look up their data sheets for the VCC pin and you'll either be on 3.3v or 5 V rails and you can use a multimeter with diode mode to check those to see which of the two you have a short in.

I'm going to make just a wild guess that the short is probably behind a transistor that's turning power on during boot up on your 3.3 V line. At this stage I would do your mainboard first. If you can't find a short on your mainboard, then it might be worth plugging the mainboard back in but leaving the entire peripheral/graphics module out and try to power on the system and see if it just complains but it doesn't actually shut down due to the overcurrent protection.

This would then tell you if it's in your graphics/peripheral module or if it's in your midplane.

The basic problem is that octane mainboards can be had for under $300 if you look online. Sometimes people ask more but if you keep an eye out you'll find one for under $300. Graphics modules have a higher failure rate, so I'd probably lean more toward that but either way your basic chassis and you're very basic XBOW (old version) can be had for cheap as well. And because they could be had for so cheap most people are not going to pay someone to repair the system under most circumstances these days. As octane prices increase and parts become more scarce this situation will change.

That's the primary reason you don't see people working on them is most people are not going to accept a $500 bill to fix their octane. You'd rather spend $300 replacing the part and moving along with your life when possible.

I understand if the troubleshooting becomes a bit much, however that's the area we live in right now as the systems become older there's an increase in failures. Proactive maintenance will eventually become a thing for those left that haven't started their systems in decades and wish to spend the money to recap their systems and power supplies. For right now there's enough used octane components on the open market that no one is going to potentially spend that, they'll just replace the part with another used part.

So again, I would suggest that you use a multimeter on diode mode and check for shorts randomly around easily accessible parts of your mainboard. Should you find any go ahead and post pictures of where you found those points and we may be able to give you some further guidance. And the most part you're looking to identify which of the main modules has the short in it. My guess is going to be it's probably in your graphics/peripheral area only because that's the lowest hanging fruit.

Also the face of the compression XIO connectors on the modules are not to be touched with either tools or your bare finger. In reality they were supposed to have special caps on them prevent people from touching them. Those caps are thrown away once the boards are installed so just a word of the wise, do not be tempted to touch them with your finger at all. You can use a light, clean unused, makeup brush to perchance dust off various things but don't touch them and don't try to put a multimeter probe on any of the pins you see on the compression connectors face. You can always use the solder joints on the PCB board side. Just don't ever touch the interface (faces) sides.

We've heard of all the modules having failures before. So just because I say it's most likely the peripheral Bay doesn't mean that's what it is in your case. But it's easier to start at the mainboard, then potentially move to the peripherals.

Just so you're aware there's an additional connector outside of the power supply standard connector that I believe does something like 24V. There's power connectors next to the compression connectors on a lot of the boards. Those power connectors would also be a great place to probe with your multimeter to find if one of your modules has a short on one of those rails.

Spend a good hour looking around, and let us know what it is you find. If you can at least find a smoking gun on one of the modules then you know where the problem is localized.

[stormy]

This thread might be helpful for you: https://forums.irixnet.org/thread-4349.html