Here's another perspective on things:
http://news.cnet.com/8301-1001_3-10119277-92.html
Basically what the article says is that the problem back in june was the combination of:
*High lead bumps + weak underfill = 'Material set' which was buggy*
Then nvidia switched to
*High lead bumps + robust underfill = 'Material set' which ought to be fine*
According to nvidia intel shipped hundreds of millions of chipsets in this combination.
So now the question is, was the exchange of just the underfill enough to fix the problem of cracking/dying chips?
Points which speak against it:
1. Now all new nvidia chips don't use high lead bumps anymore (eutectic solder + robust underfill)
2. A chipset (which has been sold by intel) has different thermal properties than a graphic chip. A GPU like a CPU is exposed to much higher temperature and stress levels
3. What does a company do if it has a big stock of already produced (but buggy) chips? It can't change the chip itself but it can change the underfill (which doesn't belong to the chip) - which of course improves the situation but definatly doesn't solve the problem in the long term
Just my two cents,
ToM
Message was edited by: tom_1st