Everything seems to be 8-bits off from where it should be - k becomes c means 107 becomes 99, so check / touch up the solder on all the address pins before attempting to reflow the gpu. Ugh, that thing's all BGA?
The GPU predates the bumpgate / early lead-free solder flip-chip issues so reflow doesn't carry much risk - just use flux and ideally pre-heat the back side of the card before heating the card. Maybe dehydrate it before soldering too.

Could you run VMTCE to determine which memory address / series of addresses is failing? Since it's an Nvidia card that should work without issue.

Thermalwrong wrote on 2023-12-17, 02:06:

Everything seems to be 8-bits off from where it should be - k becomes c means 107 becomes 99, so check / touch up the solder on all the address pins before attempting to reflow the gpu. Ugh, that thing's all BGA?
The GPU predates the bumpgate / early lead-free solder flip-chip issues so reflow doesn't carry much risk - just use flux and ideally pre-heat the back side of the card before heating the card. Maybe dehydrate it before soldering too.

Could you run VMTCE to determine which memory address / series of addresses is failing? Since it's an Nvidia card that should work without issue.

I've discovered a way to get it working again - if I wiggle the card back and forth while its still connected i can get it into a position where it works beautifully but it definitely takes some time. I wedged an eraser between the card and the hdd cage (its a shuttle xpc so the distance between the two is a few cm) to keep it pushed. Moving the pc from my desk and back to storage sometimes stuffs it up again however. The agp slot on the board and contacts on the card are squeaky clean as i only restored the whole system around october ish. I dont know how to solder so thats out of the question

I have yet to run vmtce - I cant even get the system to boot when the card is playing up so would I be able to run it when its working fine and still see whats wrong?

If it goes away when you move the card around it's a cracked solder joint. This is a common problem from this timeframe. I have one of these cards doing this as well. It's a frequent point of failure for cards with early BGA DDR2 memory that was hot, they didn't cool it properly, and the solder couldn't take the thermal stresses.

swaaye wrote on 2023-12-21, 18:17:

If it goes away when you move the card around it's a cracked solder joint. This is a common problem from this timeframe. I have one of these cards doing this as well. It's a frequent point of failure for cards with early BGA DDR2 memory that was hot, they didn't cool it properly, and the solder couldn't take the thermal stresses.

Would putting flux around the suspect memory chip/s and heating it up fix it? If I can even identify which one (or all of them) is stuffed. I've got these little heatsinks on them now but they're pretty pathetic

https://www.aliexpress.com/item/4000303524775 … .1b1a1802aWdoTv