PPU and RPU questions

[DarkSideA8]

Just watched Ageia's deformable objects video - which is pretty cool. One of the first criticisms, however, is that the objects have high resolution meshes.

http://www.hardforum.com/showpost.php?p=1030077903&postcount=6 - the criticism being that the "wow" factor is limited to this little room, and that being able to do this throughout an entire game is unlikely due to the calculation problems inherent in having many objects with high resolution meshes.

However, I recently read that raytracing (via an RPU) could take away the need for high polygon objects, and perhaps make the process easier.

Any of you tech-savvy folks have an idea whether this would work?

 

[dotK]

Raytracing is a long ways off from being utilized in games. The processing power simply isn't there.

 

[LuminaryJanitor]

I really can't see how you could substitute ray tracing for geometric complexity. I'm pretty sure models themselves wouldn't come out looking any different than they do under traditional rendering methods.

You might be thinking of normal maps, which as far as I know are built with ray tracing, and got a lot of hype a couple of years ago as a replacement for high-detail geometry. If it's definitely more recent than that, then it might be parallax mapping. No idea how these are generated, but you can fake it using a normal map (as someone has already done for Q4), so ray tracing might be involved.

But whatever it is, I don't think it's ray tracing. And you're definitely not going to be seeing a ray tracing card in the near future - doing this kind of thing in real time at high resolution is the domain of supercomputers.
EDIT: Maybe that's why you don't need high-polygon objects - 'cause you'll be running it at 320x240

 

[DarkSideA8]

This is some of what is being written. Obviously, the RPU is a new idea. I am just wondering if the RPU / raytracing technique - which allows for truly round objects, not ones with a whole bunch of triangles, will work with a PPU and improve overall 3D gaming quality / experience...

See below:

Recently, realtime ray tracing has been developed to the point where it is becoming a possible alternative to the current rasterization approach for interactive 3D graphics. With the availability of a first prototype graphics board purely based on ray tracing, we have all the ingredients for a new generation of 3D graphics technology that could have significant consequences for computer gaming. However, hardly any research has been looking at how games could benefit from ray tracing based graphics. The RTGames working group tries to fill this gap with the development of computer games based on ray tracing. http://graphics.cs.uni-sb.de/RTGames/

"3D-cards are not designed for raytracing. They read polygon meshes and then scanline-render them. Scanline rendering has very little, if anything, to do with raytracing. 3D-cards can't calculate typical features of raytracing as reflections etc. The algorithms used in 3D-cards have nothing to do with raytracing." http://tag.povray.org/povQandT/miscQandT.html

"In a review to be published in the September issue of 3D World magazine, ART VPS's new PURE PCI-X raytracing card has received eight out of ten points. According to Pete Draper, the 3D World reviewer, "The card raytraces quickly and leaves all software renderers in the dust when it comes to speed." The new PURE card contains 16 AR350 raytracing processors, twice the number of previous PURE cards, accelerating dedicated 3D final-frame rendering. Using these ray-tracing processors, the PURE PCI-X produces images that are identical to a high-quality photograph, but with the added value of realistic 3D effects and animation. ART VPS’s ray-tracing technology enables images to be generated up to 35 times faster than with conventional software rendering." http://www.art-render.com/news.2.htm

 

[quadnad]

Something I don't understand from that POVRAY.org site FAQ section is that he mentions that one of the reasons GPUs aren't capable of doing Ray-tracing is because "The algorithms used in 3D-cards have nothing to do with raytracing[...]This means that you can't use a 3D-card to speed up raytracing (even if you wanted to do so). Raytracing makes lots of float number calculations, and this is very FPU-consuming."

Isn't the whole idea behind ATi's Stream Computing and Nvidia's Gelato that since the GPU has such massive floating point computation powers that it's perfect for other uses (e.g. folding)? Perhaps the GPUs STILL don't have enough FP power!

A resource that greatly helped me understand the ins and outs of raytracing: Ray Tracing Wiki

Here's a quote from an interesting page describing the dedicated raytracing hardware a group built:

The ray tracing performance of the FPGA prototype running at 66 MHz is comparable to the OpenRT ray tracing performance of a Pentium 4 clocked at 2.6 GHz, despite the available memory bandwith to our RPU prototype is only about 350 MB/s. These numbers show the efficiency of the design, and one might estimate the performance degrees reachable with todays high end ASIC technology. High end graphics cards from NVIDIA provide 23 times more programmable floating point performance and 100 times more memory bandwidth as our prototype. The prototype can be parallelized to several FPGAs, each holding a copy of the scene. A setup with two FPGAs delivering twice the performance of a single FPGA is running in our lab. Scalability to up to 4 FPGA has been tested.

 

[SuperGee]

I see it's a other way to get 3D done.
The geometry and texturing will be the same. But the rendering method is different.

The pro and con's. Do game have a large need for it compared to the current choice the market is in for many years already.
Well if you games have a lot of mirrors wich aren flat. Or shiny ball's Opaqe or not.
It might be the best we do do it. But misses the Hardware to do it competive. But the current methode will do alright to.

Most things and outdoor scene aren't made of shiny things more matt and ambient light disperence have a large impact.

In those RT PU demo some thing, rarely or ocasional needed in real games, are often showed like a shiny opaque sphere.

Of course there would be a game that could use a sphere. Like a shiny Space ship from the in and outside. But Army games. Don't. Not lot of mirrors in games.

So Ratracing might add something of better correct result in some circumstances.
The market isn't in a hard need.
If the big boy.
The DirectX and OpenGL members see a urgent use need of raytracing it have a chance. If the current way suffice it will not. Except some hardcore pusher operating outside the market.

it could be most game are better of with current state and a few could be better in detail with the RT method. Well they made a choice way back then.

 

[drizzt81]

I really can't see how you could substitute ray tracing for geometric complexity. I'm pretty sure models themselves wouldn't come out looking any different than they do under traditional rendering methods.

http://www.intel.com/technology/itj/2005/volume09issue02/art01_ray_tracing/p06_rt_performance.htm

According to their charts, it is pretty clear that raytracing is less affected by high triangle count that traditional raster methods.

If the viewport size remains fixed, then the ray-tracing performance scales logarithmically with the complexity of the scene. This means that if you compare a HW raster engine and a SW ray-tracing engine using the same input for both engines, although the HW will initially beat the SW, the SW will eventually catch up with the HW. In fact we measured this and found that the intersection point is in the vicinity of the 1M triangle range, i.e., when the scene complexity exceeds 1M triangles, a SW ray-tracing solution will always outperform a HW raster solution.

then again, that is an Intel document, so take it for what it's worth....