on the topic of HDR and motion blur

[Cons Himself]

i have a question for the more technical members of the board.

luddites stear clear!

the way i understand hdr is that it allows for realistic motion blur in real time rendered scenes.

following from this assumption, it is a fact that TV and movies run at 24FPS, but motion blur enables them to appear perfectly smoothly.

if motion blur can be enabled in real time rendered engines, we can have silky smooth gameplay at 24FPS.

does my logic follow and if not why not?

cheers

 

[Cypher19]

the way i understand hdr is that it allows for realistic motion blur in real time rendered scenes.

It's not a requisite, but HDR really helps (as usual )

following from this assumption, it is a fact that TV and movies run at 24FPS, but motion blur enables them to appear perfectly smoothly.
if motion blur can be enabled in real time rendered engines, we can have silky smooth gameplay at 24FPS.
does my logic follow and if not why not?

Yup, but you need a LOT of motion blur sampling for it to look silky smooth and really good at 30. Over 16 samples starts to look really nice at 30fps in RTHDRIBL, but having that many becomes REALLY costly when you need to start adding in things like shadows. The RTHDRIBL demo is an insanely simple demo in regards to the rendering required for each sample. Add in stuff like skinned animation, and all of a sudden you need to recalculate each vertex on each sample (imo though, extra vertex calcs needed for binormal/normal/tangent for bump mapping would be unnecessary since using only one for the entire frame would be enough). Put in shadows (either shadow maps or volumes) and you need to rebuild/re-render THOSE every sample too which is insanely costly. And that's before having to account for stuff like the extra fill rate.

In the end, I would prefer running a game at 500fps than running it at 30fps with 16x sampling.

 

[Cypher19]

Btw, is it me or does at least one or two of these HDR-related topics come up every time Kawase's demo is posted?

 

[Cons Himself]

im not a tech man im a money man.

so from what youre saying it is way more demanding on the gpu since youre not just rendering one solid object, but an entire scene, replete with shadows, animations etc which make it prohibitively expensive to use this method of rendering.

is there a better way to model motion blur? do you see motion blur as the long term future of computer graphics? would you invest your own money in a motion blur for CGI startup?

 

[Cypher19]

Imo, I don't think we'll be seeing a lot of realistic motion blur in computer graphics for awhile (I'd say 5 years is the minimum in the best case situation with ATi/NV blowing Moore's law out of the water). For the most part, that's something reserved for non-realtime solutions.

I would have to sit down and think for awhile to think of a better solution (something I have very little interest in due to the relative unpracticality for real-time graphics), but I have no doubt that far, far better solutions exist, be they discovered or undiscovered.

 

[Soldier X]

GPUs have to render the scene. Cameras can capture the scene.

 

[subtlesnake]

There's something similar that games use called triple buffering. Perhaps a techie will explain it.