"it's nothing to do with solder. it's capacitors"

In broad strokes, a number of issues caused the Red Ring of Death ("general hardware failure") but the most dominant/infamous one was eventually attributed to a cross-domain electro-mechanical issue causing connections to break in the GPU interposer from thermal cycling in the first generation xbox gpu chips (made at a 90nm process node) which were rushed to market and fixed with second generation chips (made at 65nm process node)[1]. Initial attempts to individually address the solder, heat delta (by reducing the max temperture), and mechanical design of these chips did not fix the issue. Ultimately all three had to be sorted.

The "towel trick" worked by making the resin in the PCB temporarily more pliant - allowing the x-clamp to reseat the gpu and temporarily re-bridging these broken connections on the GPU and to a lesser extant the CPU. Similar idea when swapping out a stronger "x-clamp" - aftermarket or with washers.

Old leadership was given two shots to fix this initially and failed with Zephyr[early 2007] and Falcon[late 2007] generations which were best efforts given the cruel reality that GPU architecture turnaround time of around 2 to 3 years. Ultimately the fix had to come from three sources: ATI (new chips), replacing Peter Moore with Don Mattrick (new leadership and comms), and new hardware revision (Jasper[2008]). Earlier attempts to "replace the lead-free solder with a different formulation" in 2007 were ineffective in addressing the problem.

Fundamentally it was a calculated gamble - Moore got a huge (albeit short-sighted) win for himself and Microsoft doubled their game console market share by pushing their vendors and themselves to compromise their release engineering schedule. It's debatable if they would have captured more share by running a more diligent process - probably no - given the issues were initially latent and didn't manifest for a few years, but certainly they would have lost less market share later on in 2008 and 2009. They couldn't, for instance, have known the PS3 would be so expensive at launch nor that the Wii would be such a sleeper hit in an untapped market segment.

[1] https://www.youtube.com/watch?v=z2d6IMBS8oY&t=1130s

> if you heat the ceramic [capacitor] above its Curie temperature, all of the titanium atoms go back into a regular cubic lattice. if you cool it rapidly (just heat sinking away to the board is fine) it resets the aging process. and you can do this infinity times. suddenly the capacitance is back up to where it was before it even left the factory.

> the Curie temperature is pretty low. like, 120°C ish.

https://chaos.social/@gsuberland/113084484599859593

Worth reading it all!

https://www.youtube.com/watch?v=I0UMG3iVYZI (ripfelix)

It is a long watch, but it goes into a fairly deep dive into the whole affair. The author came to the conclusion that while a small ratio of the problem was the caps or the bga, There was a new process in play where chiplets were bonded to a base(bumps) which was then bonded to the motherboard(bga). And all processing units made with this first generations bonding process had a higher then expected failure rate, That is, the ps3, the xbox360 and nvidea graphics cards of this generation are all failing for the same reason.

Having said that, That post is an amazing deep dive into the chemistry of capacitors.

This also may not apply very much to gaming consoles in particular.

If the wire carrying high current has copper that is too thin gage, it will heat to some extent since its resistance is not quite zero, and at any one current the more resistance then the more heat will need to be dissipated. If there is a point where it is not pure copper from end-to-end but has a solder joint in between, the solder will be the point of highest resistance so it will autoheat to a higher temperature than the copper does.

Plus when a solder joint cools properly it solidifies as a eutectic alloy for the most part rather than a crystalline solid. A "cold" solder joint can have more of a crystalline nature, plus can have lots of metal oxides included if it was allowed to remain molten for too long or at much higher temperature than necessary considering its melting point.

IOW a cold solder joint can be even less of a conductor than normal, so much less it can even sometimes act as a bit of a ~semi~conductor in exactly the spot you would least want that.

For those who are using ceramic coupling caps in their audio circuits, this can help explain some of the differences that can be detected between different types of caps.

Especially in vintage-style vacuum tube assemblies where I would expect sometimes the internal temperature does rise above 120 C, and this would depend on ambient temperature too.

The towel trick warped the board and sometimes it worked and most of the time you wound up with a warped board and a system that continues to RROD.

I fixed mine this way and still have it at my parent’s house, I’ll take a picture when I can, it’s very obvious what happens with the towel trick.

It makes a ton more sense that there is a somewhat of a reset that is randomly happening when people are fixing stuff nearby and thinking their main target was the thing that fixed it.