Red light showing from processor/memory board

[FilthyMuppetInnuendo]

Ram controllers are in the CPUs, you see Ram error blinking code.

If I take off the side panel when trying to boot it, will any of the on-board error lights be lit if there’s actually a RAM problem? Or a CPU problem?

Did you take good photos of the sockets and the CPUs, if not do some macro photos of the sockets and photos where we can see the CPUs, both sides.

That would… require me to tear it all down again. Took me seven months to feel comfortable doing it the first time. I swear, they looked totally fine as I put them into place. No bent pins, no dust on the bottom of the chips, etc. After 16 years, the thermal paste on the CPUs was even still soft, for heaven’s sake (it was baked onto the northbridge, though), and I was surprised by that.

EDIT: I do just happen to have two images. One of my CPUs remained stuck to the heatsink when I took it off, so I have a shot of both the bottom of the chip and the socket. The other one stayed in place, but I did have to pick it up and set it aside later (had to take the daughterboard out of the tray and tilt it on its side to disassemble and reassemble the northbridge heatsink, and since they’re unlidded CPUs it would have just fallen out). But again, they were still pristine when I put them both back later, and I noticed zero pins out of place and no dust between them in the sockets.

EDIT 2: I reset the RTC (button while plugged in, rather than taking the battery out) because it hadn’t been plugged in since November. I booted it again with the side panel off. None of the diagnostic/error lights on the daughterboard are on, but the sleep LED still blinks roughly twice a second. Can’t reset the SMC because that requires a boot chime first.

 

[FilthyMuppetInnuendo]

I reseated my RAM a third time and thought, “She’s already dead. I may as well try.” I tightened both CPU heat sinks down even further (I knew one was a little loose for my liking) and… she booted. I guess it was just saving itself from self-destruction from “not registering” the presence of the heat sinks.

Installing a temperature monitor tells me the northbridge is sitting at 75°C idle, which I’m reading is a correct/normal temperature for a Mac Pro without an additional fan added. My CPU A diode is 51°C, but my CPU B diode is 38°C. Does that mean I should tighten A’s heat sink just a tiny bit more? Or was that just the system running a few cores more on the other one for some reason? Thanks everyone who, over the last few months, gave me the courage to finally try this out.

So for anyone doing this sort of thing: Don’t be afraid to tighten your heat sinks down, even on unlidded chips. Fingertip tight, but get them tight. I want to do a tiny stress test (gaming in windows) to check northbridge and CPU temps next…