11-15-2018, 09:30 PM
https://www.extremetech.com/gaming/28065...ses-remedy
Quote:“Test escape” is a term of art that refers to a problem that slipped through quality control testing and wasn’t caught until it wound up in the hands of the end-customer. This certainly does appear to be the case. While I haven’t been running news up the flagpole every single time an RTX 2080 Ti failed, I promised to keep an eye on the situation and have been doing so. Kyle @ HardOCP has also done some of this work — one of the two RTX 2080 Ti GPUs he purchased failed after just two hours last Friday. Other cards have literally burst into flames and there’s been speculation in the GeForce forums that the problem might lie with Micron’s GDDR6. Supposedly cards coming back from Nvidia have Samsung memory instead of Micron.
A memory swap could also simply be evidence that Nvidia’s repair shop / GPU supplier had Samsung memory in-stock instead of Micron. Just because a company is using two different companies for supply doesn’t mean it always has an equal and identical stock of parts from both manufacturers on hand literally every single moment of the day, and we’d need to know more than we do about what issue Nvidia has identified and how it slipped through the cracks to know what happened here. Thus far, that information hasn’t been disclosed.
...
The optics of shipping defective hardware immediately after you jacked up the price on your flagship GPU by $500 are not good, and gamers who promptly sold their old cards upon purchasing new ones may find themselves stuck on Intel integrated graphics while the replacement goes through, but Nvidia has pledged to make the situation right for affected customers. What Nvidia needs to do now, if at all possible, is release information that will help gamers identify if they have a defective GPU before it fails, in order to request replacements as quickly as possible and avoid potential system damage.