Those pics look like crap. Why would anyone want to use this tech if it takes way more power and looks worse?

So, my back of the enveolope math suggests we would need about 40 cores (assuming Core2 style cores) in order to play a ray-traced game at 640x480 (presumably a ray-traced game at 640x480 would be much better than a conventional game at 1600x1200, but perhaps others will disagree).

Assuming clock speed and IPC improvements, maybe it could be done with 32 cores. So, if we assume a doubling of the core count every two years, we'll have 8 cores in 2008, 16 cores in 2010, and 32 cores in 2012. Add in another year or two in order for that to become mainstream, and we're looking at 2014 before we get raytraced games at 640x480. Unless, of course, someone comes out with specialized hardware that can do better. Still -- this is a ways off...

I remember wolfenstien engines - those were "ray traced" on a 2 d graph, the distance of the nearest wall trace inverted to give the y value to draw on screen - fnun times. and one dimention of trace.

Current graphics is pretty but an approximation. It's why rendering programs only use the gpu for effects previews.

a ray traced render accelerator would be interesting. The hit on a ray traces with image complexity isnt bad, but the hit from increased res is serious.

we're not ready for this tech until a hardware specific solution is optimized to accellerate this.

Intel is clearly hot for raytracing. Its the flavor of the month candidate right now for "next big thing". Which doesn't necessarily mean much.

At any rate, Takahashi talking to Intel/AMD/NV about raytracing: http://blogs.mercurynews.com/aei/2006/08/the_coming_comb.html

Ummmm, what's the benefit of ray-tracing? Seems like an awful lot of processing power to do ray-tracing.

"Real Time Ray Tracing" = PR Plug for cpus (usually dragged out of the closet and dusted off by Intel every time it releases a newer and faster cpu.)

Sure, I could chain together several computers to do "real-time ray tracing," but that would be about as useful to me as deciding to start wearing 20 pairs of socks every day so that I could don a pair of OJ Simpson's Bruno Mali 12 EEE's...;)

What's amusing to me is that every time this old turd is trotted out and polished, the so-called "technical pundits" among us never fail to act like it's something they've never heard before in their entire lives...;)

So we can process sound on todays video cards and even Fold but it can't help at Ray Tracing?

RPU: "A working prototype of this hardware architecture has been developed based on FPGA technology. 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"

http://graphics.cs.uni-sb.de/SaarCOR/
http://graphics.cs.uni-sb.de/~woop/rpu/rpu.html

i'd like to take a moment to pimp a (mostly) ray-tracing oriented forum i frequent: http://ompf.org/forum/

Awesome picture at the end of the author.

if ray tracing is so great why do all the screenshots from ray traced q3/q4 look like the game is running 16bit with low graphics settings.......... :S

at least the shadows are clean, well maybe too clean

Hmph. Rather than have accurate water effects or shadows right now, I'd rather spend my processing power with more complex "things" on-screen or more things in general.

The thing that I immediately thought of when reading this, and which at least one other poster has thought about, is that we could be almost here anyways with stream processing. 96 shader units gives you a heckuva lot of processing power right now.

I still remember doing raytracing calcs on my amiga for hours and days just for a single image. Would be funny if it made a comeback. Would on the otehr hand also mean the dead of GPUs...

I remember hearing about this project or a similar one a while back. The same basic idea anyway, adding ray tracing to existing game engines and rendering it with multiple cpu cores.

How much?

That's my concern... would it be cost-effective to go this route? I can see lord of the rings pulling it off when the return on investment is so high... but what about a game?

Anyone with some experience in this area wanna provide some insight? :P

Glad to see "touted" is still as popular as ever,Im sure VRock will smile when he sees it.

Well, it's no surprise Intel likes the idea. They've been trying to grab back video processing from the GPU since the beginning. In Intel's perfect world the only real valuable components in the system are the CPU and chipset, both made by Intel. Any dollars spent on a GPU is dollars they didn't capture, and the fact that most games are GPU-bound, not CPU-bound, means they get fewer upgrade dollars too.

But we're a long way from real-time raytracing being usable in games. That performance graph at the end is pretty telling: a single core does 4fps on a 256x256 framebuffer, a quadcore 16fps. For a minimally playable game, you'd want 1024x768 and 30fps, which requires 96x the throughput of a single core (in some combination, like 48 cores running at twice the clockspeed); at 1600x1200 and 60fps you'd need roughly 480x the cores/clocks. (The nice thing about raytracing is that it is linearly parallizable; the bad thing about raytracing is the load rises with every pixel processed). CPUs have been advancing at a remarkable rate, but that's quite a ways away still.

However, a CPU optimized for raytracing wouldn't necesarily have many general-purpose cores, or even specialized cores with local memory like Cell. You'd be looking at something more like a set of stream processors. And who is already on their way there? If Intel doesn't get out in front of it they may find nVidia (and possibly AMD/ATI) GPUs are once again stealing value from the CPU. Which might suggest one of the reasons they are hiring graphics expertise.