Rated: 95% (4 votes)
Loading ...The death of the graphics processor?
garethogden
There’s been a lot of talk on the web recently about Intel’s efforts in the field of real-time ray-tracing (RTRT) and Custom PC is no exception, as you can see here. The debate has centred on whether ray-tracing is set to replace traditional raster-based graphics in games in the not too distant future, and this is certainly an exciting prospect given the potential benefits in terms of visual quality that ray-tracing allows. The question is, is it likely?
There’s certainly good reason to think that Intel is gunning for control of ATI’s and Nvidia’s traditional stomping ground when you consider how much research effort the company is investing into highly-parallel software applications, such as game graphics. And with multi-core processors becoming the norm and Intel’s ‘Larrabee’ product – which, according to Intel CEO Paul Otellini, will be ‘a highly parallel, many-core product comprised of an array of Intel architecture cores’ – creeping ever closer, Intel is certainly making all the right noises about becoming a major player in the high-end graphics business.
Intel’s interest in the area is understandable when you consider the rise of companies such as Nvidia, which owes much of its success to its graphics processors, plus the fact that the popularity of computer games shows no signs of abating.
Also, after Intel’s principal rival AMD (not IBM, as Otellini joked recently) bought ATI (a company also best known for its GPUs), a move by Intel to make traditional graphics processors redundant would certainly make AMD’s purchase look like a rather poor investment. Plus, it’s very much in Intel’s interest to make CPUs, rather than GPUs, the primary processor for running games anyway, since Intel is much better at making CPUs than it is at making GPUs, as its woeful ‘Extreme’ graphics processors illustrate.
However, despite Intel’s aggressive stance on ray-tracing, not everyone thinks it will replace GPUs. Ars Technica has put forward a balanced argument to say that ray-tracing alone isn’t the answer and therefore won’t replace traditional GPU-based rendering any time soon. It argues that a combination of ray-tracing and traditional rendering is probably more likely, with ray-tracing used to calculate so called ‘secondary’ or ‘indirect rays’, which are the results of, for example, reflections and refractions, while a GPU handles the rest. According to Ars Technica, trying to employ ray tracing for both types of ray (also known as global illumination) would lead to a massive step back in game performance. However, a hybrid design using a GPU for primary rays and ray-tracing for secondary rays would still lead to more realistic effects than can be simulated using shaders alone, while also allowing for a fluid frame rate.
Ultimately, we won’t know how this will pan out until Larrabee becomes a reality and more is known of Intel’s plans for real-time ray-tracing. Will ray-tracing replace traditional rendering completely, thereby consigning Nvidia and ATI (and possibly AMD) to the scrapheap, or will the industry adopt a hybrid design that takes advantage of both multi/many-core CPUs and graphics processors? Time will tell, but if ray-tracing does take off then it certainly looks like a win-win situation for Intel either way.
While the idea of ray tracing is quite exciting it is so computationally intensive that it is never going to catch up with conventional methods. Its destined to be the technology that never was.
If it is going to take 2 to 5 years for for the hardware to be able to ray trace at current game standards then it will take a similar time frame for that high end hardware to filter down to the mainstream. Only at that point would it become commercially viable for developers to make a ray traced game.
The idea of hybrid ray traced and standard graphics games is also probably a non starter commercially. Game developers are not going to go to the additional expense of programing for an add on ray trace feature for the niche extreme system owner gamers who lets face it would buy the latest games with or without a ray traced add on.
So to my mind you are looking at 6 to 8 years before ray traced games to match current standards could on both technological and commercial considerations hit the shelves.
But of course during that period conventional games, graphics cards and displays will have progressed 100x fold. We will all be playing games at ultra extreme full HD resolutions on 3D holographic displays.
So I probably would not be going into the loft to get my 24 inch wxga 2D TFT monitor to play a ray traced game. That’s even if I could buy a suitable connector on Ebay to connect my DVI monitor to the uubbber display port on my 80 core larabee extreme supercomputer.