Who knows, maybe you’ll still be able to run it just to see....Id love to see how few frames my card can pump out trying to calculate ray traces.
This is my hope, it will also be very interesting to see how Vega fairs versus the 1xxx series too. AMD pretty much need to put out an RT driver for it as they appear to have nothing else to offer for quite come time, so even if it's slow, it will show they can do it too. If it turns out to be faster than 1xxx series, that will also be good for them , even if it just means you can get some better shadows in SOTTR (no pun intended!).
If it perfroms OK against the "RTX" cards with RT on it will of course be even more interesting.
As had been mentioned elsewhere, Vega has a lot of horsepower but it is not well utilised a lot of the time. If RT can be done on the e.g. 20 - 25% that goes unused, it may work.
Additionally, I wonder if this "kind of async" the NV presentation showed (the new pipeline) where integer ops are done in parallel means that DXR etc. are implemented where these ops are fully parallel, if that's the case (and the apparent new push for 2 cards hints at it), maybe if you have a 2nd card (e.g. a 2nd Vega) if the s/w can then utilise it better by e.g. putting the needed RT/integer etc. calcs onto the 2nd card - or optimally spreading the various calculations over the total cores to get the right balance for the best performance.
It would seem that's the way things are going.
If this does turn out to be the case, for those wanting more performance, what would a TR with 4 x V56/64s perform like and how much would it cost if 2080ti is 2x the cost of a V64? Obviously either option is expensive and the power bill/cooling could be a factor but less likely if you can afford that kind of system.
Wouldn't it be funny if cards can be given a calculation to do with very little data xfer and the results also don't need much and you could run your own super computer with 10+ cards on oe of those mining boards with 1x slots - or some board made for it with a load of 4x. (This last bit is a bit of fun BTW!).
/ramble