Revolution or Hype? Seeking the Limits of Large Models in Hardware Design