Skip to main content

What Does It Mean?

As gamers around the globe try to get the best quality visuals and highest performance out of Doom III, NVIDIA marks the launch of a new (and potentially regular) discussion of technical wizardry by explaining Floating Point Precision.

Dark blue icons of video game controllers on a light blue background
Image credit: Eurogamer

We don't know about you, but every so often when we're reading about some wonderful new trickery in a videogame - be it graphics, audio, networking, you name it - we're forced to fire up another web browser and head off hunting for a simpler explanation. Buzzwords, much as we love them, can be a big problem for us all, and that's one of the reasons we've decided to kick off a new, semi-regular area of the site dedicated to working out what the hell some of these things mean. Hence the name. In these days of normal mapping, specularity and a whole host of other exciting things that only a scant few can fully comprehend, What Does It Mean will attempt track down the sort of people who came up with the whole idea in the first place, and then try and extract some sort of reasonable explanation from them.

Without further ado then, here's NVIDIA's explanation of a topic close to the hearts of Doom III fans at the moment: floating point precision. Tell us, NVIDIA, what does it mean?

"Floating point precision refers to a computer operation that uses a 'floating' decimal point to attribute values to a given function. In relation to the way games are displayed on your PC, it refers to image quality and way a graphics card renders an image on your monitor.

"Previously, when graphics cards were less advanced, they processed information in terms of 'integers' or 'whole numbers'. The disadvantage to working with whole numbers is that it limits the level of detail you can achieve in a given scene in your game. When you are able to render a scene in a game using more refined values, through decimal places, it enables the designer to attribute far more subtle variations in terms of colour and design.

"To illustrate the point we'll use a very simplified example; imagine three bars of colour - black, grey and white - lined up next to each other. This represents three values of colour using an integer value scale - 1, 2 and 3. As you can imagine, the stark contrast between the three colours would not be useful for recreating photo-realistic imagery. Looking around you, you can see that the real world is created through gradients of light in the way shadows are formed and light falls on the surfaces of objects in your environment.

"Introducing floating point math to the technology means that, using the same example above, you can now introduce variations black, grey and white using values with decimal places. So, between black (1) and grey (2), you can now introduce a combination of the two colours, black/grey (1.5). This subtle variation already smoothes the transition between black and grey. Taking this idea further you can introduce further refinements using smaller and more numerous values until you achieve a seamless transition of colour from black to white and every subtle variation in between.

"In modern gaming terms, this distinction is very important. NVIDIA's latest generation of graphics chips support 32-bit floating point precision to produce very detailed, cinematic quality graphics. The movie and TV industry already adopt 32-bit precision for the kinds of visuals that you'll see in the latest Hollywood blockbusters. This is called the IEEE (Institute of Electrical and Electronics Engineers) standard and is considered the benchmark for graphics quality.

"However, it isn't always necessary to render a scene to such a high standard in the context of a computer game so the default 'fallback' for rendering more simple images is 16bit floating point precision. Using less precision reduces the workload of the Graphics Processing Unit (GPU) on the graphics card and so improves other factors such as the frame rate. In other words, 32-bit and 16-bit can be used together to give users the best image quality and the fastest performance.

"Just as the very latest movies require the very latest technology to realise the best special effects, so do games designers. Most next generation games, such as Doom III, have been designed to take advantage of 32-bit floating-point precision capability of NVIDIA's latest range of graphics cards, the GeForce 6800 series, unlike some of our competitors who can only support fixed 24-bit floating-point partial precision.

"NVIDIA is proud to be leading the technological race in terms of computing visual solutions and giving the consumer the choice to play games as they were intended to be seen by the developer."

Besides the technical tomfoolery, opinions and views expressed in What Does It Mean are those of their respective authors and do not necessarily coincide with those of Eurogamer or its staff. If you'd like to suggest a topic for What Does It Mean, email contact@eurogamer.net and we'll do our best to find out for you.

Read this next