The card is fast. My ballpark estimate is 40% faster than a GeForce RTX 2080 Ti in situations where you're limited by the GPU.
The die is huge, at 628 mm^2. That's not really a shock, but it is a reason why the card is expensive and always will be.
To get that jump in performance, there's also a considerable jump in power consumption. Using 30% more power than an RTX 2080 Ti in order to deliver 40% more performance is only a modest increase in energy efficiency. That's still enough to make it the most efficient GPU ever, but not by that large of a margin, and you'd have hoped that Nvidia would have gotten more efficiency out of a die shrink than that. Then again, AMD didn't get an enormous jump in efficiency out of the die shrink to 7 nm, either, so die shrinks probably don't do as much for you as they used to.
What the card is not, however, is available. Micron says that GDDR6X is still only sampling, not in mass production:https://www.micron.com/products/ultra-bandwidth-solutions/gddr6x/part-catalog
That certainly lets you make review samples, and lets you have a handful of cards show up at retail. But don't expect to see the RTX 3080 widely available at retail at the MSRP of $700 this year.
The RTX 3070 is far more likely to have widespread availability soon, however, as it relies on the now mature GDDR6 memory, rather than the upcoming GDDR6X. From some back of the envelope arithmetic, it is probably a little slower than the RTX 2080 Ti. However, at $500, it's going to be massively cheaper than the latter card. That's definitely something good that we got out of the die shrink.
There are two other cards that loom large in this analysis, of course. One is Nvidia's own GeForce RTX 3090. That may or may not be a fully functional GA102 die, while the RTX 3080 uses a considerably cut down die. From the paper specs, the RTX 3090 could be about 20% faster than an RTX 3080. It will also cost more than twice as much.
And there is also Navi 2X, which AMD has said will be in the Radeon RX 6000 series. Hopefully that heralds a return to sane naming after the RX Vega 64, the Radeon VII, and the RX 5000 series being the successor to the RX 500 series. AMD has promised that Navi 2X will offer a 50% energy efficiency improvement over Navi. I'm guessing that it doubles the performance of a Radeon RX 5700 XT, which would put it as slightly slower than an RTX 3080, while using significantly less power. Of course, if that's where the card should land, then AMD might decide to just clock it higher and make it match an RTX 3080 in both metrics.