Like I commented in several places, including in Jeff's story, this is about right. AMD would have done everything to get a token "win" against the 1080, although in truth we are probably looking at a product that is more or less equivalent at the cost of more power.
However, it would be unfair to judge the FE edition on this sole metric. The 1080, powerful as it is, does NOT have any pro drivers. It is completely destroyed by the Vega FE in anything associated with CAD, 3d modelling and the like. So, for many people the card is an interesting middle ground between strictly validated "pro" cards (that costs several $k) and consumer cards that are not made for the job.
AMD seems to think that game makers might want to switch between making assets (3d modelling, pro drivers) and testing the game (game drivers) will be interested. I guess that is an interesting niche. The price is not at all bad for a pro product and the power consumption is probably not very important in that use case.
The most relevant question for us is what can we deduce for the gaming version that will follow. I would guess slightly higher performance enabled by having to feed less RAM (8GB vs 16GB) with probably similar power consumption. I would estimate $500 for an air cooled version, roughly equivalent to the 1080, and $700 for the water cooled version, scoring above the 1080 but below the 1080Ti.
As I also noted elsewhere, the maturing ROCm platform and the release of the miOpen (equivalent to cuDNN) may make or break the product for several compute applications. The nVidia advantage in compute is probably more about software than hardware. There is a lot of work to do on that front, but things seem to be moving in the right direction.