Hacker News Reader

Comment by reaperducer | Hacker News Reader

Ray Tracing

I'm not someone into the inner-working of chips much, but is "ray tracing" a new term used for something in microprocessors now? Or is this the same graphic "ray tracing" we were doing back in the 80's on Amigas and Atari STs?

View on HN

We're crossing the threshold this year where real-time ray tracing in hardware isn't just some theoretical concept, it's actually useful and available in affordable consumer hardware (NVIDIA and AMD GPUs, as well as the PS5 and new Xbox all have it).

Yes, the NVIDIA 2xxx RTX series had it two years ago, but this is the year where it's actually viable and not so gimmicky.

It's also shipping in consoles this generation, which is going to drive a lot more games to actually implement it. When it's only being used by 5% of the PC userbase, maybe you don't bother doing that work. If cheaper GPUs can push that up to say 20% of PCs next year, you still might not.

But when every PS5 and XSX has raytracing hardware, suddenly it makes sense. That's going to be helpful for getting it supported in PC titles sooner.

We've been able to do realtime ray tracing in software since forever though and the shadertoys and whatnot have been full of hardware accelerated demos, that's not so interesting.

I remember playing with a number of demos on my intel core 2 duo macbook (not pro) a decade ago.

The new GPUs have hardware specifically designed to do the type of math that raytracing does, such as ray/volume intersections, faster or more efficiently than generic shader hardware. Sufficient quantity (FLOPS) can become a quality of its own.

They weren't doing that in hardware real time accelerated at 4k resolution in AAA games. This year they are. That's a big leap from your core 2 duo demos.

Realtime ray-tracing is a mix of hardware acceleration for ray intersections and de-noising. The de-noising approaches are certainly novel and not really analogous to previous methods, imho, though the accelerated bounding box hierarchies for accelerating intersections certainly appear in previous iterations.

I think this is the key answer, de-noising, using various neural techniques finally allowed acceptable images to be produced from sometimes <1 ray / pixel, where normally images produced at such densities were nearly unusable.

This is still not perfect and many ray tracing techniques rely on accumulation over time which limits certain images from working (I imagine raytracing a small particle cloud, or fast shifting objects to be a worst case scenarios)

The new ray tracing is being done in real time rather than being baked into the scenes.

I did not read it as a list of technologies used in chip fabrication, but as a list of “killer apps” driving the need for more performance.

It's the same graphical ray tracing, although with many modern improvements like path tracing for randomization of the ray bounces to efficiently approximate and converge on the correct lighting, and ML-powered denoising and upscaling algorithms that take low-resolution fast rendering and transform it to higher quality and detail.

The latter. Systems have become fast enough recently to do realtime ray tracing.

its a common enough workload that dedicated hardware for it has emerged, and software written to utilize that hardware is becoming more common place.

Reply on news.ycombinator.com