There are few creations in modern science fiction that can make your stomach drop much like the sandworms of Arrakis. Stretching for several hundred metres across the desert landscape, these massive creatures emerge from beneath the dunes, displacing vast quantities of sand as they move.

This apex predator of the Dune universe (known by the indigenous Fremen as Shai-Hulud), has established itself as a hallmark of this series of films by Denis Villeneuve, playing a key role in the visual identity of the planet. But the challenge of convincing the viewer was far greater than just drawing and animating a digital animal.

em360tech image

The visual effects teams also had to accurately map the physics of how the surrounding environment modelled the behaviour of the environment when this huge body moved through it. It had to fall on the ground around the giant, the dunes had to crumble and fall down, and the dust clouds must interact persuadingly with the surrounding landscape.

It needed complex physics-based simulations, professional visual effects software, and a lot of computational power.

The preparations for the sandworms are an interesting example of how modern film production depends on the use of high-performance computing (HPC). HPC is also used for scientific simulations, engineering studies, and increasingly demanding business applications.

Why Did Creating Dune’s Sandworms Require So Much Computing Power?

An improbably large digital creature is only part of the magic of the effects. The environment must react naturally to its movement. For the sandworms, this meant representing how vast quantities of sand moved in their wake as they dragged themselves along or emerged from the dunes.

DNEG was the lead visual effects partner on both Dune films, having also contributed to acclaimed productions such as Blade Runner 2049, Interstellar, Inception, Oppenheimer and Tenet. For Dune, its teams' creature effects, special effects and sand simulations helped bring the vast, unforgiving world of Arrakis to life.

One of the most difficult tasks was capturing the movement of sand on such an enormous scale. Sand is composed of granules that move, collide, collect and settle. When a sandworm passes beneath a dune, the surrounding environment is affected by these interactions. The task for the filmmakers was to produce these effects convincingly, without making them appear too mechanical or detached from their physical surroundings.

The original Dune involved sand simulations at an extremely fine visual resolution. As DNEG reported, this placed enormous demands on its computing and storage resources, generating petabytes of simulation data. Keep in mind that 1 petabyte is equal to a staggering 1 quadrillion bytes. The studio also studied how the sand thrown up around a colossal creature could resemble the displacement of water when a whale moves through the sea.

In Dune: Part Two, visual effects were further developed and sand simulations and worm trails extended over hundreds of metres of desert. The resulting effects demanded much computational resources to calculate the movement of particles through the environment.

The distinction between traditional animation and computational simulation is thus highlighted: rather than drawing by hand every movement, one uses mathematical models that calculate how the materials and objects will behave under certain conditions.

But as the complexity and scale of the simulated universe increases, the calculation difficulties increase proportionately.


Infographic showing the technology behind Dune's world

From Individual Particles to More Complex Simulations

To understand why these effects require so much computing power, it helps to look at the simulation of physical phenomena. Physical simulations use mathematical models to approximate how materials and objects behave.

A simulation might calculate an object's position, speed, acceleration and the effects of collisions with surrounding surfaces. These principles can also be applied to more complex phenomena, including flowing water, smoke, explosions and moving sand.

One commonly used technique is particle simulation. Particles are computational elements that represent individual components within a system. Each particle has properties such as position, velocity and direction, alongside rules governing its interaction with surrounding particles.

These simulations can also be combined to recreate more complex physical phenomena. For instance, in a sand simulation, individual particles can be used to show how material moves across a dune, collapses around an object or becomes dispersed in the air.

The more particles there are, the more calculations are necessary. When these particles begin interacting with one another, the computational complexity increases further.

A sequence involving vast quantities of moving sand is likely to exceed the capabilities of a conventional workstation. This is where high-performance computing becomes important.

What Is High-Performance Computing?

High-performance computing is the use of powerful computing systems to process complex calculations and large datasets more rapidly than conventional computers could individually.

Rather than relying on a single computer, HPC systems can combine multiple servers, specialised processors, storage systems and high-speed networks to handle demanding workloads.

It is useful for calculations that are too extensive, time-consuming or complex for conventional computing systems to manage efficiently.

A typical HPC environment relies on three interconnected components:

  • Compute: The processors and servers responsible for performing calculations. These can include CPUs, GPUs and other specialised accelerators.
  • Storage: The systems responsible for holding the large datasets required by simulations and other computational workloads.
  • Networking: The connections that allow computing systems to exchange information and access shared data efficiently.

All three components are important. A system may have powerful processors, but if it cannot access the data required for its calculations efficiently, its performance will suffer. Similarly, large amounts of memory are of little use if processing resources are inefficient.

The ultimate objective is to create a system environment in which data transfer, storage and processing work together. This allows studios such as DNEG to execute simulations, render digital images and continually refine their work.

Why Are GPUs So Important to Visual Effects?

One of the most important technological elements behind modern visual effects is the Graphics Processing Unit (GPU). Beyond gaming and graphics programmes, GPUs are also used for a wide range of computational tasks.

The Central Processing Unit (CPU) is a general-purpose processor designed to perform a broad range of instructions and coordinate the computer's operations.

A GPU, on the other hand, is designed to perform many calculations simultaneously. This makes it particularly useful for workloads involving large numbers of similar calculations that can be processed in parallel.

Consider a simulation involving millions of particles. Each particle may require a series of calculations to determine its position, velocity and influence on its surroundings. A GPU can handle many of these calculations in parallel, reducing the time required to run a simulation.

However, this does not mean that GPUs completely replace CPUs. Some calculations involve sequential operations or tasks that depend on one another, which are often better handled by the CPU.

Modern computing systems typically use both types of processor, allowing them to handle different workloads more efficiently.

During the first Dune production, DNEG used Dell Precision workstations and Dell PowerEdge servers equipped with NVIDIA RTX and server GPUs to support its visual effects work.

This combination of hardware and software allowed artists to experiment with demanding simulations and refine their results.

The same basic technology is used across many other industries and organisations, including those that build and operate artificial intelligence (AI) systems and run scientific simulations.

Render Farms and Distributed Computing

Running a complex simulation is only one part of producing a finished visual effect. Once the digital environment and its elements have been prepared, they must be converted into the images that audiences see. This is known as rendering.

Rendering is the process of computing how the interplay of objects, materials, light, shadow, and camera will produce an image. To produce a sandworm effect, for example, the process may involve combining the digital creature with simulated sand, dust, ambient lighting and photographs taken on location.

One single image may require considerable processing, particularly if it involves complex visual effects. To manage these workloads, visual effects studios often use render farms.

A render farm is a group of computers, linked together, that share their resources in order to process the renderings. Instead of confining the rendering to a single machine, the studio can distribute the work between several of them.

For example, a 60-second film with twenty-four frames per second contains 1,440 frames. If each frame takes several minutes to be drawn, the processing of the entire sequence with a single machine would be far too time-consuming.

The render farm allows several frames to be rendered at the same time, thus reducing the time required. A 'distributed' work is an application of a distributed system, whereby a certain number of workstations are connected in a network, and a task is distributed over a number of them.

In this case the rendering of individual frames and the simulation of individual elements are not at all identical. A simple rendering is an independent operation, which can be easily divided up. A complex simulation can have several of its elements whose calculations depend on previous calculations, or on the interaction between several of these elements, and thus requires more complicated communication between the nodes.

Yet the processes both show why the resulting digital image depends on much more than powerful computers.

It requires an environment which coordinates the computation, the memory and the communication facilities of many systems.

The Enterprise Connection: The Same Infrastructure Powers AI and Scientific Research

The computational challenges of Dune may appear specialised to the entertainment industry, but the underlying technology supports a much wider range of applications.

Modern companies are increasingly using complex calculations to develop products, analyse information and operate sophisticated systems. High-performance computing provides the infrastructure needed to support many of these tasks.

AI and machine learning

AI is one of the most prominent consumers of computing resources. Training an AI model involves processing large amounts of data and repeatedly performing mathematical calculations to adjust the model's parameters.

GPUs are particularly useful because they can carry out many of these calculations simultaneously.

The relationship between visual effects and AI training is not that the two processes are exactly alike. They are not. However, both can make use of specialised hardware, parallel processing and the high-speed movement of large amounts of data.

As AI applications become more complex, enterprises will need to ensure their existing IT environments can cope with the additional processing demands.

Engineering and product development

Are you enjoying the content so far?

To analyse how a product or system will function before manufacture or use, engineering companies and manufacturers rely on mathematical simulations.

For example, an automobile manufacturer may simulate how a car responds under various collision conditions. An aeronautical engineer may calculate airflow over an aeroplane wing, while an energy company can study how its machinery responds to changes in the environment.

These simulations make it possible to model complex physical interactions and perform extensive calculations.

HPC allows companies to run more sophisticated models, compare different scenarios and potentially reduce the time and cost of physical testing.

Weather prediction and climate modelling

Another important application of high-performance computing is forecasting the weather (or weathercasting). Meteorological machines divide the atmosphere into calculating planes, apply mathematical equations to them, and foretell the future state of the environment.

A number of climatic factors, the temperature, atmospheric pressure, wind speed and humidity, have to be taken into account, the models have to take account of all this. The more complicated the forecasting system, the more computations are needed.

It’s also more complex in that it deals with the long-term trends of the environment and the interaction between the various systems.

These same general capabilities, which can be employed to re-create fictitious surroundings, are used to perform scientific research and support real-world decisions.

Medical and science research

HPC is also essential for genomics and molecular modelling, and it can support the optimisation of drug design. Computer models help chemists study molecular interactions, analyse large biological datasets and identify potential drug candidates.

These activities can involve extensive calculations that are difficult or impossible to perform using conventional computing resources. The technical capabilities that make modern visual effects possible therefore have applications far beyond producing increasingly realistic films.

Why Computing Infrastructure Is Becoming a Strategic Consideration

As AI, simulation and other data-driven applications become more common, organisations need to rethink their approach to computing.

Enterprise platforms may not have been designed to support workloads that require extensive parallel processing, specialised hardware acceleration and rapid data movement.

This creates several infrastructure considerations:

  • Processing power: Organisations need sufficient CPU and GPU resources to meet the demands of increasingly complex workloads. This could mean investing in specialised hardware or accessing additional computing power through cloud services.
  • Storage requirements: Simulations and AI applications can produce huge datasets. Businesses need storage systems that can store, access and transfer information without creating bottlenecks.
  • Network performance: Distributed computing environments require reliable, high-speed connections between processors, storage systems and other infrastructure components.
  • Scalability: Computing needs may vary depending on the workload. Organisations need to consider how they can scale their resources without maintaining expensive infrastructure that remains underused.
  • Energy and cooling: High-performance computing systems can consume significant amounts of electricity and produce substantial heat. Data centre operators must consider the additional power and cooling required to support increasingly intensive workloads.

The cloud offers one way of accessing HPC resources without having to establish an entirely dedicated environment. However, its suitability depends on factors such as workload frequency, complexity, data transfer costs and available resources.

For some organisations, on-premises systems may be more appropriate; for others, a hybrid approach makes sense.

The critical consideration is whether the computing infrastructure has the capacity to support the applications that depend on it.

Beyond Faster Processors: Building Infrastructure for Increasingly Complex Workloads

The Dune sandworms show that creating realistic virtual experiences requires more than powerful graphics cards. It requires physical models, specialist software, parallel computation, large amounts of memory and networks capable of supporting complex computing workloads.

The same principles are increasingly relevant to enterprise technology. As companies expand their use of AI, scientific simulation and computational modelling, their infrastructure requirements will become more demanding.

The challenge is not simply acquiring faster machines. It is building systems capable of storing, transferring and processing information efficiently while supporting increasingly complex applications.

DNEG's task was to simulate vast quantities of moving sand and give gigantic creatures the illusion of existing in a natural environment. The same high-performance computing capabilities can help organisations advance engineering, healthcare, climate research and AI.

The sandworms of Arrakis may be fictitious, but the scale of their computational needs reveals something very real: the future of enterprise computing increasingly depends on the infrastructure needed to support its most demanding workloads.

Following the latest tech developments through channels like EM360Tech’s Emerging Technologies feed can help your organisation stay ahead of the game and confidently meet future demands.