3 ms·

It's not the capacitors in 360s that failed... it's something in the GPU, whether it's the interconnect, the die, or bad underfill, people have been trying to

It's not the capacitors in 360s that failed... it's something in the GPU, whether it's the interconnect, the die, or bad underfill, people have been trying to figure it out for years... but the failure analysis tools are expensive, and Microsoft will never come clean on why they failed...

jetbalsa 2y ago
He is making a good bit of sense, if it was the tiny 0402 SMD Caps near the GPU going south that would explain a good bit of what is going on
- NikkiA 2y ago
  Ceramic capacitors rarely die, when people talk about recapping electronics it's the electrolytic capacitors that they're talking about; which dry out.
  - robin_reala 2y ago
    Not dying, but falling out of tolerance specification.
    - eqvinox 2y ago
      That's equally rare for MLCCs, in particular for mass decoupling caps. It's really the Al-electrolytics and Tantalums where this happens.
phendrenad2 2y ago
The prevailing theory back when I was in the scene was that the GPU was warping the PCB itself, causing solder balls to lose connection. Heating in an oven softens the entire board, allowing it to un-torque.
- loa_in_ 2y ago
  My theory was that high operating temperature caused the balling to get partially softened and develop crystalline defects when cooling. I base it on personal experience that they never failed while running, but at least three people I know experienced these failures after powering down PCs for few days. All four GPUs were fixed by a friend of mine by... cooking them in the oven briefly
andreashaerter 2y ago
This is partially mentioned in the linked conversation at https://chaos.social/@gsuberland/113084501943452456 https://chaos.social/@gsuberland/113084501943452456: > @greg this is also why it only ends up working temporarily in some consoles. if other components (particularly electrolytic caps) are showing their age, they start to have higher impedance, which causes more noise on the power rails and more undershoot during load transients, which can cause the CPU/GPU to glitch out. when you do one of the hot air tricks you de-age the MLCCs, which is just about enough to compensate for the other parts aging, but then that early rapid aging stage kicks back in
- rasz 2y ago
  no amount of 'de-aging- mlcc capacitors is going to fix broken solder bonds under GPU silicon

jetbalsa 2y ago

He is making a good bit of sense, if it was the tiny 0402 SMD Caps near the GPU going south that would explain a good bit of what is going on

NikkiA 2y ago
Ceramic capacitors rarely die, when people talk about recapping electronics it's the electrolytic capacitors that they're talking about; which dry out.
- robin_reala 2y ago
  Not dying, but falling out of tolerance specification.
  - eqvinox 2y ago
    That's equally rare for MLCCs, in particular for mass decoupling caps. It's really the Al-electrolytics and Tantalums where this happens.

phendrenad2 2y ago

The prevailing theory back when I was in the scene was that the GPU was warping the PCB itself, causing solder balls to lose connection. Heating in an oven softens the entire board, allowing it to un-torque.

loa_in_ 2y ago
My theory was that high operating temperature caused the balling to get partially softened and develop crystalline defects when cooling. I base it on personal experience that they never failed while running, but at least three people I know experienced these failures after powering down PCs for few days. All four GPUs were fixed by a friend of mine by... cooking them in the oven briefly

andreashaerter 2y ago

This is partially mentioned in the linked conversation at https://chaos.social/@gsuberland/113084501943452456 https://chaos.social/@gsuberland/113084501943452456: > @greg this is also why it only ends up working temporarily in some consoles. if other components (particularly electrolytic caps) are showing their age, they start to have higher impedance, which causes more noise on the power rails and more undershoot during load transients, which can cause the CPU/GPU to glitch out. when you do one of the hot air tricks you de-age the MLCCs, which is just about enough to compensate for the other parts aging, but then that early rapid aging stage kicks back in

rasz 2y ago
no amount of 'de-aging- mlcc capacitors is going to fix broken solder bonds under GPU silicon