Did you all see that the 1 Quadro RTX 8000 was just as fast as 4 V100's in doing the Star Wars Demo?
 
So no idea what Stanford could do with that kind of power and AI and Tensors core but it could be awesome.

Quadro did not seem well received - not really supported - when I was looking into them about 2 yrs ago, before buying my Titan X(p)
 
The drivers that are specific to the Quadro can make them perform very well - at a significant price

its only 4x as fast when under ray tracing (RT) loads. under non RT loads is where i really want to see how well it can perform, because that is where F@H will run. i strongly doubt F@H will ever update to where it could utilize the RT specific hardware, though i may be wrong.

We need more specifics from nvidia about it. The ray tracing uses less precision as it's an AI style algorithm. The split between integer and floating point units looks interesting though. They could have increased the size of the floating point units to support the extra 8bit fp units Deep learning is all hot and bothered about.

would be nice
 
Sure hope there are about 10K units world wide on Launch day & not just a couple hundred

The Ti series is almost never the first announced. Probably be an xx70 and xx80 like Pascal's launch. The 1080Ti was months later.

 margin improvement through recycling

Dang is it interesting. The split on int and FP looks awesome on paper. 50% more perf across board supposedly. Wonder if folding can take advantage if it. The rest of it Tensor and RT units might prove useful in long term but I don't see any short term benefit as the coding is probably gonna be long in coming.

I'm willing to find out if they are good at Folding

Microsoft® DirectX® 12 API, Vulkan API, OpenGL 4/5  -  This from EVGA's website.  I haven't seen anything about OpenCL support, it only talks about OpenGL.  Since I don't anything about this side of things will it still work with FAH with it needing OpenCL?  I was just curious, my pocket book won't let get any of these new cards this year.  

If they were going to use the Tensor cores, then the Volta that is already out would have a HUGE increase in score.  As it is, the performance increase isn't that huge, and the Tensor cores on the V100 cards are not being used.  Stanford would also need to redesign their cores to take advantage of the new hardware.  Currently everything is designed to run on Kepler, that is how far behind they are.  Not Maxwell, but Kepler, currently 6 year old architecture.  Stanford only has 1 programmer as they are either too cheap, or too underfunded to pay anyone else to update their code.  This was pushed hard when crypto was booming because of the sheer amount of farms using Pascal cards and they were told to update their ancient cores, and Stanford basically said "no", but not as politely.
 
Stanford doesn't care about this new hardware, because the majority of users don't have it.  That's their attitude, and it won't change.

Cool GTX
2080 Ti seems to be the new Titan Xp replacement 

Thats what I've seen from a few reviewers say.  Which makes since to pull the Titan name from the Gaming cards.  

it will not do that with RTX on, that i can guarantee.

Real Time Raytracing in the home has been a fascination for me ever since I saw this demo that was created on a Pentium 4 back in 2000
 
https://www.youtube.com/watch?v=CFRZfCgIs2E
 
That Nvidia managed to recreate cinematic levels of RTR (Yes, I know the framerates are limited at THIS time, give it a gen or two to catch up!) in a price range that will allow this tech to be put in the hands of the consumer? THAT'S the big leap in the technology for me.  Imagine the creativity that is going to come pouring out now with this kind of visual output available to the average end user.  Machinima will never be the same, for sure, and you better believe that Hollywood studios will be leveraging the technology as well!
 
Now, imagine what COULD happen if a clever coder were able to take the AI decision making paths that RTX relies on and apply them to folding? THAT could be a game changer right there!
 
d

10 gigaray (10million?) is a far cry from traditional ray tracing. The 10Giga rays might even only refer to effective equivalence to traditional ray tracing for all we currently know. They are reducing the number of rays by an order of magnitude by reverse bounding the visible rays to the viewpoint.