BAD RAM, BAD RAM slot, Heat expansion?

[Wild Hare]

Recent upgrade adventure with an 867 MHz DP MDD to a 1.42 GHz DP CPU… provided some “BAD” RAM fun. Booted fine initially with four sticks of 512 MB PC2700 - but as it warmed up after several reboots it would sometime freeze & hang on boot. AHT reported problem with J20 RAM.

Tried 12 different sticks that worked fine in other slots / machines. No go in the J20 slot. DEEP cleaned the slot (and all RAM) and it booted fine twice. Third boot - No go. Had to be the slot and figured that with a little heat expansion as the machine warmed, good contact with the stick in the J20 slot was sketchy. Pulled another 512 MB stick (newer) from another machine and it’s been perfect for the last three days now. Minute contact wear might also have been a factor. (Especially as it warmed up.)

Close-up, contact “wear” comparison between two older sticks and the third (newer) is revealing? RAM does get swapped around quite a bit here and between different machines. *Some “wear” does occur. AND perhaps in the RAM slots too.

 

[ervus]

I would guess the problem is/was a cracked solder joint on that socket and wiggling it enough has restored the connection for now. Also, I've seen RAM and stability issues when overclocking the bus on MDD machines. Sometimes 2x1GB will fix the problem (instead of 4x512MB), but on two machines I had to set the bus back to 133 and adjust the CPU jumpers.

 

[MicroTecture]

I would guess the problem is/was a cracked solder joint on that socket and wiggling it enough has restored the connection for now. Also, I've seen RAM and stability issues when overclocking the bus on MDD machines. Sometimes 2x1GB will fix the problem (instead of 4x512MB), but on two machines I had to set the bus back to 133 and adjust the CPU jumpers.

Most people would say to reset the PRAM/NVRAM/etc but sometimes that might help as well. Of course depending on certain things can results vary as well.

 

[Wild Hare]

Had considered the 2 x 1GB solution but (surprisingly) no 1GB sticks here currently. And I’d think that after having 12 different sticks in that slot, that one of them would have wiggled a cracked solder joint back into working position, at least once. (It took the lucky 13th stick?) Also did an extensive inspection of all solder points for the J20 slot.

And side-by-side thickness comparison / looks like #13 is thicker. (Where is my micrometer?) Thicker one is a TECHWORKS 512 MB stick.

Have not overclocked the CPU (yet). Used a stock 1.42. GHz DP CPU from a FW800 MDD after removing the R676 resistor from 867 Mhz DP mobo. AND yes, necessary to do the PMU reset after the CPU swap.

Have two more of these FW800 / 1.42 CPUs to swap into another 867 MDD and a 1.25 GHz MDD. Won’t need to remove R676 on the latter mobo - as there isn’t one.

Report of another, also attempting a 1.25 Ghz CPU (failed) swap into an 867… but possibly not with the PMU reset. So now, reportedly will try again.

We’ll see how the next two swaps go.