3 ms·

It's the same issue that was behind NVDAs "soldergate" fuck-up that ended up permanently souring the relationship between them and Apple. The core is EU's regu

It's the same issue that was behind NVDAs "soldergate" fuck-up that ended up permanently souring the relationship between them and Apple. The core is EU's regulation on lead free solder, which led to a number of people finding out that thermal cycling on the solder led to thermal stresses. Workarounds were identified and any solder formulations since then don't suffer from that issue, so the fix is a complete re-balling of affected chips... a work not for those faint of heart.

mulmen 2y ago
Sounds like a R&D problem. Why the dig at the EU?
- mschuster91 2y ago
  I'm European, I actually support RoHS - it was just the original cause because everyone up to it getting in force was accustomed to classic, decades-proven leaded solder.
  - hot_gril 2y ago
    Did they need to have a longer transition period? Looks like it went into enforcement only 2 or 3 years after it was approved.
    - throwaway48476 2y ago
      No, the issue didn't affect the IBM CPU in the 360 or Intel products because they did the R&D work. TSMC and their packaging partners were just lazy.
    - nolok 2y ago
      It was not so much a "not enough time to transition" and more like "there is no consequences yet so why bother OH DAMN WE NEED TO MAKE IT TODAY IT GOES INTO FORCE". Many had no issues, but a few companies didn't bother to do their homework, problem would have been the same if the period was twice as long.
- gjsman-1000 2y ago
  Why not blame the EU? It is just a well known fact that non-leaded soldier has inferior properties to leaded soldier, which require careful engineering to work around, and still remain somewhat unresolved. At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused.
  - tecleandor 2y ago
    You should remember the bioaccumulative properties of lead. Lead: The most extensively spread toxic environmental contaminant (2024): https://www.sciencedirect.com/science/article/abs/pii/B9780323904001000069 https://www.sciencedirect.com/science/article/abs/pii/B97803... The Urban Lead (Pb) Burden in Humans, Animals and the Natural Environment (2022): https://pmc.ncbi.nlm.nih.gov/articles/PMC8812512/ https://pmc.ncbi.nlm.nih.gov/articles/PMC8812512/
  - mulmen 2y ago
    > At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused. “May” is doing a lot of work there. Can you substantiate the claim that the risk of lead is lower than the switching cost?
  - mschuster91 2y ago
    > At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused. The problem is where the e-waste ends up - some ditch or desert in Africa. From there it ends up leeching in the environment due to corrosion or, worse, as widespread aerosols when the people there burn the waste to get to the copper.
somat 2y ago
Complicating the issue is that this was also an early generation of chiplet so there are two levels of bga. motherboard to processing unit and processing unit to chip_actual. the latter commonly are referred to as "bumps" to distinguish from "bga" which attaches the chip_structure to the mother board. A lot of the problem was in the bumps for this chiplet like sub assembly. and while reballing bga is a tricky but well understood process. my understanding is that reballing bumps is nearly impossible.
- throwaway48476 2y ago
  It's called FCBGA, flip chip ball grid array. The only other option for packaging is wire bonding but that doesn't scale well with pin count. Technically you could recall and repackage the dies but you'd need millions of dollars in equipment.

mulmen 2y ago

Sounds like a R&D problem. Why the dig at the EU?

mschuster91 2y ago
I'm European, I actually support RoHS - it was just the original cause because everyone up to it getting in force was accustomed to classic, decades-proven leaded solder.
- hot_gril 2y ago
  Did they need to have a longer transition period? Looks like it went into enforcement only 2 or 3 years after it was approved.
  - throwaway48476 2y ago
    No, the issue didn't affect the IBM CPU in the 360 or Intel products because they did the R&D work. TSMC and their packaging partners were just lazy.
  - nolok 2y ago
    It was not so much a "not enough time to transition" and more like "there is no consequences yet so why bother OH DAMN WE NEED TO MAKE IT TODAY IT GOES INTO FORCE". Many had no issues, but a few companies didn't bother to do their homework, problem would have been the same if the period was twice as long.
gjsman-1000 2y ago
Why not blame the EU? It is just a well known fact that non-leaded soldier has inferior properties to leaded soldier, which require careful engineering to work around, and still remain somewhat unresolved. At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused.
- tecleandor 2y ago
  You should remember the bioaccumulative properties of lead. Lead: The most extensively spread toxic environmental contaminant (2024): https://www.sciencedirect.com/science/article/abs/pii/B9780323904001000069 https://www.sciencedirect.com/science/article/abs/pii/B97803... The Urban Lead (Pb) Burden in Humans, Animals and the Natural Environment (2022): https://pmc.ncbi.nlm.nih.gov/articles/PMC8812512/ https://pmc.ncbi.nlm.nih.gov/articles/PMC8812512/
- mulmen 2y ago
  > At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused. “May” is doing a lot of work there. Can you substantiate the claim that the risk of lead is lower than the switching cost?
- mschuster91 2y ago
  > At this point, the directive may have caused more e-waste and environmental damage from part failures than the damage the original leaded soldier would have caused. The problem is where the e-waste ends up - some ditch or desert in Africa. From there it ends up leeching in the environment due to corrosion or, worse, as widespread aerosols when the people there burn the waste to get to the copper.

somat 2y ago

Complicating the issue is that this was also an early generation of chiplet so there are two levels of bga. motherboard to processing unit and processing unit to chip_actual. the latter commonly are referred to as "bumps" to distinguish from "bga" which attaches the chip_structure to the mother board. A lot of the problem was in the bumps for this chiplet like sub assembly. and while reballing bga is a tricky but well understood process. my understanding is that reballing bumps is nearly impossible.

throwaway48476 2y ago
It's called FCBGA, flip chip ball grid array. The only other option for packaging is wire bonding but that doesn't scale well with pin count. Technically you could recall and repackage the dies but you'd need millions of dollars in equipment.