Cool new rendering technology... maybe?

[Soul]

Now, i may just be even more confused then before, but if you aren't rendering things you can't see what would hapend if a creature was going to attack you from behind?

As i see it, these points that don't render when not on screen is all good when it comes to static unchanging things, but things that move would still need to be processed right?

It might not need to be rendered, but all the "search points" of the unit and then the enviorment the unit is moving on would have to be processed i think.

Processed, yes, rendered, no way. I.e. you do not care what state your model is, and you do not need to animate it, as long as player does not see it, all you care about is "physics" and "sound", NOT under any circumstances "vision"

That is all good.

But if the models aren't animated, would that mean that everytime you spin around, the models will be rendered and then be dropped down in the T-Pose before the animations start? Ò_ó

That or every unit of the same kind would make the same first animation when first being seen/rendered?
That would be kinda... boring.

Why in Q3 if you stare at wall your FPS can just as high as triple the average?

True.

But i think its fair to assume that the fps would not just be trippled but dozenquaddrupled when staring into a wall, if you only use your RAM and not your grafic card when/if Q3 would be remade in point cloud right?

This is not me saying your fps would be duzenquaddrupled when using point cloud and staring into a wall, this is me saying that you will have regual fps when looking at a wall, but have "regual fps"/"dozenquaddrupled" when not looking at a wall.

[==Troy==]

Depends on your architecture and the way you design your game to be. Physics can always just specify "frame YYY of model XXX of animation ZZZ", still you do not have to render the model or the polygons if you are not looking at them.

[Soul]

Depends on your architecture and the way you design your game to be. Physics can always just specify "frame YYY of model XXX of animation ZZZ", still you do not have to render the model or the polygons if you are not looking at them.

Again true.

If having a massive world prerendered then that would reduce your fps alot too, you would of have to balance it out.

Alot of games solve this by making regions inside the game, where you preload/prerender the next segment of the game.

All these have their pros & cons ofc, and i can see that point cloud can have som very intresting applications,
but i think its quite stupid to make grafic cards "obsolete" better to walk a middle way, so you don't need to spend vast amounts of money on either grafic cards or RAM-sticks, but a moderate amout on both.

[==Troy==]

Depends on your architecture and the way you design your game to be. Physics can always just specify "frame YYY of model XXX of animation ZZZ", still you do not have to render the model or the polygons if you are not looking at them.

Again true.

If having a massive world prerendered then that would reduce your fps alot too, you would of have to balance it out.

Alot of games solve this by making regions inside the game, where you preload/prerender the next segment of the game.

All these have their pros & cons ofc, and i can see that point cloud can have som very intresting applications,
but i think its quite stupid to make grafic cards "obsolete" better to walk a middle way, so you don't need to spend vast amounts of money on either grafic cards or RAM-sticks, but a moderate amout on both.

GFX card is nothing more but a very specialised piece of hardware to allow high memory bandwidth and parallel computing of the scene. It does not matter WHAT fills up the port on your LCD screen, as long as it can do that fast and snappy.


If the OP's search algo cannot be paralleled well, but is still faster than conventional poly graphics, then be it, the gfx cards will just become optimised for the search algo (and even if it cant be paralleled, there is still a LOT of things you can optimise for hardware-wise)



Edit : most games have the world pre-rendered (Q3 is no exception).

[Soul]

Depends on your architecture and the way you design your game to be. Physics can always just specify "frame YYY of model XXX of animation ZZZ", still you do not have to render the model or the polygons if you are not looking at them.

Again true.

If having a massive world prerendered then that would reduce your fps alot too, you would of have to balance it out.

Alot of games solve this by making regions inside the game, where you preload/prerender the next segment of the game.

All these have their pros & cons ofc, and i can see that point cloud can have som very intresting applications,
but i think its quite stupid to make grafic cards "obsolete" better to walk a middle way, so you don't need to spend vast amounts of money on either grafic cards or RAM-sticks, but a moderate amout on both.

GFX card is nothing more but a very specialised piece of hardware to allow high memory bandwidth and parallel computing of the scene. It does not matter WHAT fills up the port on your LCD screen, as long as it can do that fast and snappy.


If the OP's search algo cannot be paralleled well, but is still faster than conventional poly graphics, then be it, the gfx cards will just become optimised for the search algo (and even if it cant be paralleled, there is still a LOT of things you can optimise for hardware-wise)

Fair enough.

Edit : most games have the world pre-rendered (Q3 is no exception).

Yeah, but i think its all the fancy effects and not the rendered world that reduces your fps.

If using point cloud data system then you would chose to not fully utilize a very efficient "tool".
If you give your RAM all the grafic cards work and it's own aswell, and just have your grafic card stitting there as a spectator.
Then I'm quite sure that that your fps would give a new meaning to groud zero.


One quick question, all the ingame effects, such as explosions and lightning projectiles, do they utilize your grafic card or your RAM the most?

[Hobo Joe]

If having a massive world prerendered then that would reduce your fps alot too, you would of have to balance it out.

I think you don't have a very good understanding of the way graphics technologies work. For example, in a map on a standard engine, the mapper builds the world and places lights and textures. Then, once that's done, they compile, or 'bake' the map, which applies the lighting information to the textures on the walls. Now instead of being a fullbright texture, it has shadows, but they have already been rendered so it applies no extra load onto the GPU, because it simply sees it as a texture.

Alot of games solve this by making regions inside the game, where you preload/prerender the next segment of the game.

All these have their pros & cons ofc, and i can see that point cloud can have som very intresting applications,
but i think its quite stupid to make grafic cards "obsolete" better to walk a middle way, so you don't need to spend vast amounts of money on either grafic cards or RAM-sticks, but a moderate amout on both.

Yeah, but i think its all the fancy effects and not the rendered world that reduces your fps.

That just depends on what it is. Although this is being moved away from in more recent years, in the past the majority of 'effects' (explosions, muzzle flashes, bullet tracers etc.) are just 2D sprites, which apply a very small load to the GPU, the heaviest loads on your GPU are poly counts, texture sizes/counts, and lighting/shadows.

One quick question, all the ingame effects, such as explosions and lightning projectiles, do they utilize your grafic card or your RAM the most?

Both. You can't just designate a visual effect or model or texture or anything and make it use exclusively one hardware component, it ALL uses your HDD, it ALL uses ram, it ALL uses the CPU, and it ALL uses the GPU. It's not just 'more' dependent on one piece of hardware or another, it requires all of them to function.

[SwiftSpear]

Now, i may just be even more confused then before, but if you aren't rendering things you can't see what would hapend if a creature was going to attack you from behind?

As i see it, these points that don't render when not on screen is all good when it comes to static unchanging things, but things that move would still need to be processed right?

It might not need to be rendered, but all the "search points" of the unit and then the enviorment the unit is moving on would have to be processed i think.

Processed, yes, rendered, no way. I.e. you do not care what state your model is, and you do not need to animate it, as long as player does not see it, all you care about is "physics" and "sound", NOT under any circumstances "vision"

That is all good.

But if the models aren't animated, would that mean that everytime you spin around, the models will be rendered and then be dropped down in the T-Pose before the animations start? Ò_ó

no, animation is frame based, what frame needs to be loaded when the model is loaded is trivial to identify. What frame of what animation the model is currently executing must be tracked, but the entire animation has no good reason to render, at least from the standpoint of saving processing time.

That or every unit of the same kind would make the same first animation when first being seen/rendered?
That would be kinda... boring.

once again storing "model:blah; animation: dancing; frame: 79" and incrementing frame count or swapping the current animation state when the simulation behavior changes is both trivial, and not the same as "rendering an animation". What bones are where and what polygons are in what position is not calculated until the model is forced to render. Only the animation state information is known before that point.

Why in Q3 if you stare at wall your FPS can just as high as triple the average?

True.

But i think its fair to assume that the fps would not just be trippled but dozenquaddrupled when staring into a wall, if you only use your RAM and not your grafic card when/if Q3 would be remade in point cloud right?

This is not me saying your fps would be duzenquaddrupled when using point cloud and staring into a wall, this is me saying that you will have regual fps when looking at a wall, but have "regual fps"/"dozenquaddrupled" when not looking at a wall.

That's the interesting thing about the claim. They are saying you can have alot more data then a computer professional would say is possible rendering using a trick of dropping information that doesn't need to actually be shown on the screen. It sounds plausible, but sort of in the same way the 3 houses 3 utilities problem sounds plausible to solve.

[PicassoCT]

our hoax has unlimited flaws no exposed. Also they have a bad taste for colours and there logo is very colourfull of itself, like bruce. If the tech realy exists, mjam.. spring goes voxel

[TheRegisteredOne]

One thing I am curious of is how they are going to be calculating things like illumination, reflections etc, will normals because calculated using adjacent points? Also, the appearance of the points inside the visible square also depends on other points outside of it. This is will there are shadow problems on the edges of the images I think. How are they going to know which points will influence the final image through refraction, reflection, shadows etc. The only way I can think of is some sort of ray tracing, which is terribly expensive. Also, note that in the demo, as they zoom in close, there is obvious degradation of quality in the form of some sort of pixelation due to the sparseness of the points.

[irfanmusadiq]

Great piece, keep up the terrific work.

[user was banned for being imaginary]

[PicassoCT]

you have to give it to those bots, they give you some sort of longterm memory.. unlimited tech. unlimited silence. What ever happend to the ..

yep, checked it out..

still so full of themselves as before. Why dont they see, that when you got stuff to see, you dont need a chatterbox. All the nvidia, intel grafixdemos have one thing in comon- leaving silence for the gasps.