gorobei
Diamond Member
- Jan 7, 2007
- 4,020
- 1,519
- 136
the primary uses appear to be the water/wave behavior (not look) , the depth of focus blur of background, and some ambient occlusion from what is commented on in the video.
from a technical standpoint this isn't groundbreaking stuff.
1) the water doesn't appear to be full fluid particle simulation, just surface modulation. you can achieve wave effects without full dynamics with just multi-layered displacement maps (either bitmap or procedural).
2) the bokeh blurring is interesting, but could just as easily be done/faked on cpu or with generic programmed shader effect.
3) the ambient occlusion doesn't look very impressive. the only noticeable effect was a nondescript shaped soft shadow below the character running on the beach. very low resolution and could just as easily have been done with an oldschool shadowmap. Also ambient occlusion is part of dx11 now so there is no benefit to doing it with cuda unless you are going to do a full radiosity montecarlo sample(which this doesnt appear to be). SSAO is the super cheap realtime version of ambient occlusion. the look is similar but nowhere as accurate/impressive as the stuff that comes from mentalray or renderman. Doing it in realtime is nice to claim in the features checklist, but in order for AO to look good you need some hand tweaking in the compositing stage (which i'm not seeing here).
in film and commercial vfx, we usually have a choice: do an effect with full math/simulation/control or do the effect the cheap/easy/simple way. Most of the time you end up doing the full-out/expensive way because you're sitting around waiting for final plates or timing so the producers have you do the hard version since they are already paying for it timewise. 90% of the time you could do it the cheap way and no one in the audience would notice.
Nvidia is funding the game developer to do the slightly more math/sim version, but it could be done the cheap way with no real loss.
This is fine since dx10 doesn't have access to some of the dx11 stuff and most likely have plenty of horsepower to spare for older/simpler/console type games. It doesnt mean you really want to take this approach when your are struggling to make min framerates on some of the newer dx11 engines. It is working harder instead of smarter, just for the sake of claiming you did it the "mathematically accurate" way.
The cryengine 3 footage of the radiosity demo was way more impressive and i would imagine hardware independent.
from a technical standpoint this isn't groundbreaking stuff.
1) the water doesn't appear to be full fluid particle simulation, just surface modulation. you can achieve wave effects without full dynamics with just multi-layered displacement maps (either bitmap or procedural).
2) the bokeh blurring is interesting, but could just as easily be done/faked on cpu or with generic programmed shader effect.
3) the ambient occlusion doesn't look very impressive. the only noticeable effect was a nondescript shaped soft shadow below the character running on the beach. very low resolution and could just as easily have been done with an oldschool shadowmap. Also ambient occlusion is part of dx11 now so there is no benefit to doing it with cuda unless you are going to do a full radiosity montecarlo sample(which this doesnt appear to be). SSAO is the super cheap realtime version of ambient occlusion. the look is similar but nowhere as accurate/impressive as the stuff that comes from mentalray or renderman. Doing it in realtime is nice to claim in the features checklist, but in order for AO to look good you need some hand tweaking in the compositing stage (which i'm not seeing here).
in film and commercial vfx, we usually have a choice: do an effect with full math/simulation/control or do the effect the cheap/easy/simple way. Most of the time you end up doing the full-out/expensive way because you're sitting around waiting for final plates or timing so the producers have you do the hard version since they are already paying for it timewise. 90% of the time you could do it the cheap way and no one in the audience would notice.
Nvidia is funding the game developer to do the slightly more math/sim version, but it could be done the cheap way with no real loss.
This is fine since dx10 doesn't have access to some of the dx11 stuff and most likely have plenty of horsepower to spare for older/simpler/console type games. It doesnt mean you really want to take this approach when your are struggling to make min framerates on some of the newer dx11 engines. It is working harder instead of smarter, just for the sake of claiming you did it the "mathematically accurate" way.
The cryengine 3 footage of the radiosity demo was way more impressive and i would imagine hardware independent.
Last edited:
Facebook
Twitter