It turns out that although these crazy server motherboards have all these slightly different PSU connectors on them, when in reality, they’re just an ATX mothermoard. Admittedly not everything worked, but I had enough working to know that it really was the old PSU that was the problem rather than the new motherboard.

To make this work, I had to understand what the various other connectors on the board are actually for.

The server motherboard I have has the following PSU connections:

• 24 pin ATX power connector
• 4 pin 12V PWR connector ‘to provide adequate power to the system’
• 8 pin CPU PWR connector

A normal desktop PSU has the following connectors:

• 20 pin ATX power connector
• 4 pin 12V PWR connector

The 20 pin ATX connector pinouts and 20 of the 24 pins of the ATX socket on the motherboard pinouts match up, which is what I was expecting. Rather helpfully, the 20 pin conector even fits perfectly into the  24 pin ATX socket on the motherboard if you line up the correct parts. Obviously four of the pins in the socket have nothing going into them, but it seems to be that they’re just used for additional power feeds. I hoped they weren’t required to at least turn the motherboard on for my testing.

The 4 pin 12V power obviously fits into the 4 pin 12V power socket on the motherboard.

So at this stage, I have nothing connected to the 8pin CPU PWR socket. I hoped that this would still let me turn the system on, but unfortunately it doesn’t. According to the server motherboard manual, the CPU PWR connector is just 4 Ground connections and 4 12V connections. I tested the various CPU PWR socket pins with a multimeter and it seems that two of the 12V pins connect together on the board and the other two connect together on the board. But not all four 12V pins are connected. This suggests that 4pins were for CPU PWR to one  CPU and 4 pins were CPU PWR to the other CPU.

Now again, working on assumption, if you need 12V CPU PWR what on earth are the additional 12V PWR lines/socket for? I’m sure when this socket was introduced a number of years ago onto desktop motherboards it was to provide additional power to the ‘power hungry’ new CPUs that had been released at the time, was it not?

Seeing as the 12V PWR and the CPU PWR are all 12V/GND combos, I unplugged the 12V PWR and tested whether it conneced up to any of the 8 pin CPU PWR connector. Needless to say, it plugged straight into the first four pins (presumably for CPU0). I checked the motherboard manual against the colours of the wires going into each pin and the yellow wires were connecting to the 12V pins and the black wires to the GND pins. So it all made sense in my head.

So at this stage, I have a motherboard, 1 CPU installed into CPU0 socket, and two sticks of RAM (the setup requires memory to be installed in pairs). I have the 20 pin ATX power connector plugged into the 24 pin ATX socket. And the 4 pin 12V PWR connector plugged into the first 4 pins of the 8 pin CPU PWR socket.

I bridged the pins that a power on switch should connect to with a small screw driver…. And the system powered up!

So testing my server motherboard with a desktop PSU was actually pretty simple once I thought about it.

One thing to note… As we only hooked up to four of the pins in the CPU PWR socket. Installing the second CPU into the CPU1 socket. The system powers on but never POSTs. This kinda makes sense as the second CPU doesn’t have its CPU PWR connections.

But yeah… I’ve now invested in a new PSU for the server… I paid additional money for a dual redundant PSU caddy, one redundant PSU (yeah I know… only one, just in case the testing i’d done above was telling me lies!). I now have a working new server.

There’s things online that suggested he could literally wire one end of one of the LEDs to a switch and the other side of the switch to the middle pin of the variable resistor that creates the variable trigger in the controller. Presumably the crystal on the board is supposed to be hooked up to drive the LED in the controller. But it didn’t seem to work. I didn’t have an oscilliscope handy so I couldn’t see what the signals were doing, so I wasn’t sure if it should work or not anyway. So it was back to the drawing board…

Luckily on his search around the Internet, he’d come across this blog

The principle is simple enough, take a 555 timer, stick a few components on it to set the interval of the timer and hook the timer output up to the controller.

A little measuring around things suggested to me that the variable trigger is simply just a variable resistor. The resistance between ither CW or CCW and the common pin were 2K at one extent and 3K at the other extent. I guess that’s a nice simple way to make a variable trigger work. Knowing the resistances and such, I then measured the voltages between CW and GND, CCW and GND, common and GND. One result was 0V, one was about 1.6V and commond was about 1.3V. That kinda makes sense, as the voltage of the common pin moved between 0V and 1.6V presumably the system would think it was either fully on or fully off. Nice and simple.

Now the Xbox360 controllers are USB, so i’d got 5V and GND to power the 555 chip. I knocked the circuit up quickly, using a 100K variable resistor in place of the 22K and hooked the chip up to the USB 5V and GND pins to power it. Again i’d not got an oscilliscope, but using a voltmeter I could see the output of the 555 was oscillating to some extent.

I was mildly concerned about hooking up the 5V output of the 555 directly to the common point of the triggers variable resistor as the controller was only providing 1.6V maximum as far as I could tell. But seeing as the 3.3V line that my brother had hooked up to it based on the ‘one wire and a switch’ method hadn’t broken anything, we gave it a try anyway.

A few minutes later (and after tweaking the variable resistor a bit) my brother was happy that he would empty the entire clip of what appeared to be a single shot hand gun in Call of Duty in around 1 second, rather than the 3 or so seconds it would take to mash the trigger frantically otherwise.

I recently bought a new server motherboard, processors, RAM… planning to swap out the parts in an old server. Up until a week or so before the bits arrived the old system was working fine.

Queue swapping out the motherboard etc. with the new ones once they arrive. Hrm, no power. Fiddle around a little and eventually realise the PSU appears to not be working. I fear the worst and wonder if it’s something that i’ve done, but I can’t see any issues with how the new board mounts in the case so i’m certain that there’s no shorting or anything crazy like that.

So it looks to me like the PSU in this old system has just died between now and when it was working perfectly a week ago. New PSU… £145 + VAT. Unless I want the redundant PSU option, which is £160 per PSU, £50 for the PSU ‘connector’ and £20 for a caddy for it to all connect together.

So now, I have no way to work out whether the motherboard actually works, without shelling out at least £145 for a new PSU, which (you never know) could be the actual problem in the first place and have for whatever reason broken my original PSU.

I hate computers sometimes.