Computer-generated graphics have been steadily improving in quality over the past few decades. We have advanced from crudely rendered 8-bit graphics to photorealistic images due to improvements in computational hardware. One of the critical elements of photorealism is the behaviour of light.
In the real world, light behaves like a ray and reflects (bounces) and refracts (bends) while interacting with various surfaces based on physical properties like colour, emissivity, refractive index, etc. To create an accurate rendition of a scene on a computer, we would have to program all these properties into virtual objects, render numerous light rays from all light sources, and trace the paths of these rays as they interact with the virtual objects. This process is known as ray tracing.
Ray tracing using computer hardware has been attempted as far back as 1982 by the LINKS-1 computer graphics system in Osaka, Japan (http://museum.ipsj.or.jp/en/computer/other/0013.html). However, due to the heavy computational requirements, it was reserved for use in pre-rendered scenes. In 1995, the animated movie “Toy Story” was released by Pixar studios, and it was rendered on 117 computers. It took between 45 minutes to 30 hours to render a single frame. (https://www.insider.com/pixars-animation-evolved-toy-story-2019-6). To perceive smooth motion, we must see at least 24 image frames in a single second, but computers obviously couldn’t render so many frames in real-time with ray tracing enabled. Hence, most computer graphics generated in real-time on consumer hardware was rendered using a method called “rasterization”, which approximated how light would be rendered and could result in glaring visual flaws, which reduced immersion.
In short, the main factors which have prevented ray tracing from becoming mainstream are as follows:
- Cost: The amount and type of hardware to render images with ray tracing would be prohibitively expensive, costing tens of thousands of dollars.
- Speed: Even after spending a lot of money, the speed of rendering would be too slow for real-time applications, taking several minutes to even hours to render a single frame.
- Electricity consumption and heat production: A large amount of hardware would consume a lot of electricity and thus produce a lot of heat, making it impractical for average consumers.
This changed in 2018 when NVIDIA launched the RTX 2000 series of graphics processing units, with specialized hardware to drastically speed up the mathematical operations required for ray tracing calculations. It was possible to get a computer costing less than 1000$ and consuming only 300-400 watts of electricity capable of rendering ray-traced images in real-time at a reasonable speed. Since then, graphics processing units have been steadily improving in performance following Moores’s law, and now it’s possible to render games with real-time ray tracing enabled at high frame rates. The most recent RTX 4090 graphics card achieved 60 Frames Per Second or more at an image resolution of 3840 by 2160 pixels with ray tracing enabled on numerous games from AAA studios (https://www.eurogamer.net/digitalfoundry-2022-nvidia-geforce-rtx-4090-review-extreme-performance?page=5), once thought impossible. The number of games supporting ray tracing has rapidly increased, reaching 141 as of October 15, 2022 (https://www.pcgamingwiki.com/wiki/List_of_games_that_support_ray_tracing).
Given the visual benefits of ray tracing, increased support from games, and the primary constraints of Cost, speed, and power consumption are mitigated to a large extent, the answer to whether ray tracing is worth it for consumers in 2022 is a resounding YES!!! The scope for real-time ray tracing could also extend to the metaverse and other digital content consumed by the general public as prices for ray tracing hardware reduce.
