Server boxes

[AcadiusLost]

If the power is coming on, then it's shutting itself back down immediately, I'd suspect CPU overheating issues. AMD processors of that vintage had thermal cutoff failsafes IIRC, which are one of the few things that could command the computer to power off before an operating system or anything else loads.

I would recommend confirming that the fan on the heatsink is spinning when power is applied; if that's not spinning, you'll need a new heatsink (maybe 10 GBP). If it is spinning, it may be that the thermal paste which conducts the heat away from the processor die and into the heatsink has degraded, keeping the heatsink from doing it's job. The fix for this is removing the heatsink, cleaning off the old paste (usually need to use rubbing alcohol or something similar), getting a fresh tube of thermal paste (also very inexpensive), and applying a fresh (thin and even) layer. Reinstall heatsink, and try again.

There should be films/walkthroughs on how to do this on youtube and techie sites, if you haven't done it before yourself.

If, however, the motherboard lights come on, but there are no beeps and nothing else happens (lights stay lit, etc), you may have a problem with power supply, motherboard, or some other systemic problem.

Just to confirm; is the new video card one that requires an additional power line in from your power supply (similar to the ones going into the hard drives and optical drives, most all PCI-E cards require this)? Forgetting to link in that extra power can also cause some strangeness, as the video card generally initializes before the CPU; without the extra power coming in it might stall things out partway.

[fluffmonster]

vid card should get enough juice through the pci-e connection to run in 2d mode. not plugging in the 12v connector shouldn't stop it from booting up the OS.

[Cassiel]

Sorry - working in A&E at the moment (ER, if you prefer), and my shifts are if not actually insane then at least quite seriously unhinged. I will try to get this up and running soonest.

If the power is coming on, then it's shutting itself back down immediately, I'd suspect CPU overheating issues. AMD processors of that vintage had thermal cutoff failsafes IIRC, which are one of the few things that could command the computer to power off before an operating system or anything else loads.

Ok - that is what's happening, so this may well be the issue...

I would recommend confirming that the fan on the heatsink is spinning when power is applied; if that's not spinning, you'll need a new heatsink (maybe 10 GBP). If it is spinning, it may be that the thermal paste which conducts the heat away from the processor die and into the heatsink has degraded, keeping the heatsink from doing it's job. The fix for this is removing the heatsink, cleaning off the old paste (usually need to use rubbing alcohol or something similar), getting a fresh tube of thermal paste (also very inexpensive), and applying a fresh (thin and even) layer. Reinstall heatsink, and try again.

Will confirm this (though not tonight as it's closing on midnight) - reapplying thermal paste is no problem - I built this box myself originally, so can do.

The video card has the necessary adapter plugged in.

Thanks again to all for advice. Will update asap, hopefully this weekend.

[Cassiel]

So: the box has ubuntu installed, but is still intermittently refusing to turn on.

*Thermal paste: not this - I reapplied it
*CPU fan is spinning
*As indeed is the GPU fan, case fans, and drive fan
*CPU temp running around 40 degrees
*Power is definitely getting to the MB (green power light comes on)

Crashes take two forms:
(1) Refusal to turn on. Power switch does nothing at all (green light on MB remains lit)
(2) Crashes. Suddenly power cuts during operation. Frequently followed by (1)

Only change I can think of is that I plugged back in one of the chassis fans into a MB chassis fan power slot - but I'm not convinced this has been consistently the case with the crashes.

I may try swapping out the 2GB of memory that's in there for the old 512MB pair and see if that prevents the crashes (the RAM has a lifelong guarantee, so in some ways I hope it is that).

Will try getting the box on the network and letting zic / some of you guys run diagnostics on it. Any thoughts gratefully appreciated.

[gribo]

Did you try another power supply?

[zicada]

yeah sounds like a faulty PSU

[Cassiel]

Based on the spec of the box, would an old 200W PSU do the job, at least for testing purposes? My folks have an old desktop box I could dig one out of. If I can identify that that's definitely the problem, then happy to pick up a new one.

[gribo]

An old 200W might not have enough on the 12V rail.

[Cassiel]

An old 200W might not have enough on the 12V rail.

Would it at least get the box to a certain point, i.e. to a boot screen? I only want to use the old PSU to confirm that's actually the problem before laying out for a new one...

[fluffmonster]

Does the 200W PSU have a separate 4-pin plug for the cpu power port? I believe s939 requires that. Without it, there will be problems whether the gear is good or not. Old 20-pin PSUs don't usually have that separate cpu power plug.

[Cassiel]

Hokay - I'm going to have to give up on this. The old PSU did nothing apart from create an unpleasant burning smell which I hope was just the PSU. If anyone in or around London wants to have a look at it, or if anyone thinks they can get it working, happy to try to organise that.

However, happily, I now have an 9-day old daughter who's taking up most of my time, so tinkering with difficult hardware's dropped right down the list!

[hollyfant]

Congratulations!

[Wynna]

Oh my dear heavens! Congratulations, Cassiel!

[darrenhfx]

Congrats

[Holy_Rage]

Way to go, sir. Congrats!