What's the advantage of this vs, say a 1080 ti, since stock is available for most vendors?
From our testing the main difference would be wattage usage per algorithm and the hashrate output. The p102 rates listed above were given at a wattage of 250 watts. The 1080TI on Ethash will do 32Mh/s at ~285 watts and Equihash at about 710 Sol/s at 285 watts while the P102 will do 47Mh/s (we've gotten up to 52) at 250w and Equihash at 660 with 250w. These GTX 1080TI of course could be scaled higher with overclocking, but the rates could peak at around 325-350 watts. That difference in wattage can add up over time, and especially if you're a larger operation, that efficiency in savings can help a ton on electrical cost and reaching card ROI as well. We are also still doing tests on some other algorithms, but there may still be some different ones that are similar to Ethash that this card will do very well with. We are also still testing overclocking capabilities on these cards as well to see what we can get them up to and the wattage that comes with that.
Basically what groovy said

Thats not good hashrate per wattage. My 1080ti Blower cards with ethlargement do 50mh/s at 190 watts, along with 725 sol/s at 190 watts. The p102 is not worth it unless you can get them cheaper than a 1080ti.
1080ti is a consumer grade card with 3 year warranty, and using ETHLargement will give you equivalent hashrates as a p102.
P102 is not a consumer grade card and only has a 3 month warranty and might have less than a 5% increase past a 1080ti using ethlargement at a slightly higher TDP along with very little driver support (I believe there is only 1 driver available and doesnt support cuda 9+), and overclocking support, and support in general.
Overall, dont waste your time with a p102. 1080ti's are under $800 shipped now.