An apology for what? A "benchmark" that has to be run improperly to produce anything different for the 970 than it does the 980? Or even Kepler and Fermi cards? Run Nai's "benchmark" on any Fermi or Kepler card (as --Tj-- did) and you'll see the exact same results. That is, of course, unless you run it improperly. The fact that it produces the same results on every card, when run properly, proves it's inaccurate for the purpose it was written for. NVidia has not always admitted to mistakes. Even after being found guilty in a class-action suit concerning the "excessive failure" issues for the G84m and G86m, NVidia still denies any wrong doing and made every attempt to avoid compensating the affected customers, against a court order.