NVIDIA put the RTX 30-series out on Tuesday, Jensen pulled the new card out of an oven in his kitchen for the camera, and my feeds have been doing nothing but argue about it since. The pricing surprised people in a good way for once. The performance claims surprised people the way performance claims always do, which is to say everyone is waiting for the independent benchmarks before believing the "twice the 2080" line, and they're right to.
I don't game enough to justify any of it, so I watched it from the cheap seats. The part I keep thinking about isn't the framerates. It's that the interesting work in that keynote was the bit nobody's quoting: the memory bandwidth, the new I/O path for streaming assets straight off storage, the unglamorous plumbing. The headline number is the framerate because that's what fits on a slide and sells the card. The thing that actually moves the bottleneck is the boring bit underneath, and it's always the boring bit underneath.
That's the pattern with every launch keynote, and the reason everyone has an opinion before the reviews land. We react to the number we're given because it's the number we're given. The honest response is the dull one: wait two weeks, let someone with a test bench and no stake measure the thing under a real workload, and assume the gap between the slide and your machine is wider than the marketing and narrower than the cynics claim. I'll keep my money in my pocket until then, which is, conveniently, also what I'd have done anyway.