Bumpgate: Difference between revisions

Vindicator4021 (talk | contribs)
m Sony's Response - The "Yellow Light of Death": Added a reference for the soldering iron "poke test".
Vindicator4021 (talk | contribs)
m Made it clearer why readers need to know about the fault/defect's cause. Also added more citations where needed.
Line 1: Line 1:
'''Bumpgate''', also known as '''Nvidiagate''', was a scandal where [[Nvidia]] and ATI Graphics Processing Units (GPUs) were prone to high failure rates due to a design flaw that led to cracked solder bumps under the die. Despite the "Nvidiagate" name, this defect not only affected many Nvidia GPUs made from approximately 2006 to 2010, but it also affected ATI GPUs from 2006 to 2008. Among retro console enthusiasts, the defect is best known to have likely been the culprit behind the high failure rate of Nvidia GPUs in [[Sony]]'s early PlayStation 3 models and ATI GPUs in [[Microsoft]]'s early Xbox 360 models.
'''Bumpgate''', also known as '''Nvidiagate''', was a scandal where [[Nvidia]] and ATI Graphics Processing Units (GPUs) were prone to high failure rates due to a design flaw that led to cracked solder bumps under the die.<ref name=":0" /> Despite the "Nvidiagate" name, this defect not only affected many Nvidia GPUs made from approximately 2006 to 2010, but it also affected ATI GPUs from 2006 to 2008. Among retro console enthusiasts, the defect is best known to have likely been the culprit behind the high failure rate of Nvidia GPUs in [[Sony]]'s early PlayStation 3 models and ATI GPUs in [[Microsoft]]'s early Xbox 360 models.


==Background==
==Background==


===Graphics Processing Unit (GPU) Faults===
===Graphics Processing Unit (GPU) Faults===
From approximately 2005-2010, GPU manufacturers Nvidia and ATI developed some GPUs that had a serious design flaw. They had engineered GPUs that electrically connected the silicon chip (die) to the substrate (the "green square part") using high-lead solder bumps. High-lead solder bumps were chosen in order to fit the power delivery specifications that these GPUs needed<ref>{{Cite web |last=Williams |first=Rob |date=29 Sep 2008 |title=NVIDIA at a Disadvantage Due to their Choice of Solder? |url=https://techgage.com/news/nvidia_at_a_disadvantage_due_to_their_choice_of_solder/ |url-status=live |access-date=1 Jun 2025 |website=Techgage}}</ref>. To give these solder bumps more strength (especially when operating at high temperatures), it's standard to use an epoxy with silica filler known as underfill. Underfill needs to fit certain specifications, depending on how hot the processor it's used on is expected to get. If it's too hard, the underfill will crack the die. If it's too soft, it'll crack the bumps. It also needs to still fit within the right specifications at both high and low temperatures.
From approximately 2005-2010, GPU manufacturers Nvidia and ATI developed some GPUs that had a serious design flaw. This flaw led to failures in many of their GPUs during that time period, and Nvidia even saw a lawsuit from it. In order to understand what truly happened during this controversy, though, it's important to understand what exactly led to the faults.


In 2006, IBM and Amkor published a study that explained that use of a low Tg underfill was not acceptable with high-lead solder bumps, and high Tg would be necessary to avoid defects. Therefore, because Nvidia and ATI chose to use high-lead solder bumps, they needed a high Tg underfill. However, this study was not out at the time that GPUs from 2005 were made, and the companies ended up using low Tg underfill in these processors.  
These manufacturers had engineered GPUs that electrically connected the silicon chip (die) to the substrate (the "green square part") using high-lead solder bumps. High-lead solder bumps were chosen in order to fit the power delivery specifications that these GPUs needed.<ref>{{Cite web |last=Williams |first=Rob |date=29 Sep 2008 |title=NVIDIA at a Disadvantage Due to their Choice of Solder? |url=https://techgage.com/news/nvidia_at_a_disadvantage_due_to_their_choice_of_solder/ |url-status=live |access-date=1 Jun 2025 |website=Techgage}}</ref> To give these solder bumps more strength (especially when operating at high temperatures), it's standard to use an epoxy with silica filler known as underfill. Underfill needs to fit certain specifications, depending on how hot the processor it's used on is expected to get. If it's too hard, the underfill will crack the die. If it's too soft, it'll crack the bumps. It also needs to still fit within the right specifications at both high and low temperatures.


This low Tg underfill would become too soft at high, but normal operating temperatures for these GPUs. When the processor went through normal thermal changes, the solder bumps would soften under heat and harden as they cooled.<ref>{{Cite web |last=Demerjian |first=Charlie |date=1 Sep 2008 |title=Why Nvidia's chips are defective |url=http://www.theinquirer.net/inquirer/news/1004378/why-nvidia-chips-defective |url-status=dead |archive-url=https://web.archive.org/web/20090520152257/http://www.theinquirer.net/inquirer/news/1004378/why-nvidia-chips-defective |archive-date=20 May 2009 |access-date=1 Jun 2025 |website=The Inquirer}}</ref> This would happen over and over again until they cracked under the thermal stress. When enough solder bumps cracked, it would cause a failure in the unit, hence the term "Bumpgate".
In 2006, IBM and Amkor published a study that explained that use of a low Tg underfill was not acceptable with high-lead solder bumps, and high Tg would be necessary to avoid defects.<ref>{{Cite journal |last=Ray |first=S. |last2=Kiyono |first2=S. |last3=Waite |first3=K. |last4=Nicholls |first4=L. |date=2006 |title=Qualification of low-K 90nm Technology Die with Pb-free Bumps on
a Build-up Laminate Package (PBGA) with Pb-free Assembly Processes |journal=56th Electronic Components and Technology Conference |pages=139-144 |via=IEEE}}</ref> Therefore, because Nvidia and ATI chose to use high-lead solder bumps, they needed a high Tg underfill. However, this study was not out at the time that GPUs from 2005 were made, and the companies ended up using low Tg underfill in these processors.
 
This low Tg underfill would become too soft at high, but normal operating temperatures for these GPUs. When the processor went through normal thermal changes, the solder bumps would soften under heat and harden as they cooled.<ref name=":0">{{Cite web |last=Demerjian |first=Charlie |date=1 Sep 2008 |title=Why Nvidia's chips are defective |url=http://www.theinquirer.net/inquirer/news/1004378/why-nvidia-chips-defective |url-status=dead |archive-url=https://web.archive.org/web/20090520152257/http://www.theinquirer.net/inquirer/news/1004378/why-nvidia-chips-defective |archive-date=20 May 2009 |access-date=1 Jun 2025 |website=The Inquirer}}</ref> This would happen over and over again until they cracked under the thermal stress. When enough solder bumps cracked, it would cause a failure in the unit, hence the term "Bumpgate".


==Companies involved and responses==
==Companies involved and responses==