The carbon footprint of a VFX production is largely a result of high energy consumption, necessitated by powerful computer systems running complex rendering software, and the management and storage of vast amounts of data. Clients continue to request high-fidelity, photo-realistic imagery and massively complex scenes on shorter deadlines. Furthermore, VFX often requires holding complete copies of original media as well as plates and composited images.

The sheer volume of data required to produce and deliver high quality VFX can dwarf that of the original media created by the shoot. To meet these needs, energy requirements to render and store images multiply, as does carbon emission.

To address this environmental impact, it is crucial for VFX professionals to adopt sustainable practices.

By leveraging effective strategies for data and resource management, the industry can make meaningful strides towards sustainability.

• Optimize resource allocation based on task requirements using virtualization technologies and predictive models.

• Strike a balance between over-provisioning and under-provisioning to avoid wasted energy and project delays.

• Use machine learning models, profiling tools, resource allocation software, benchmarking, and feedback loops to accurately predict resource requirements.

To optimize the allocation of computational resources, it is essential to employ strategies that avoid both over-provisioning and under-provisioning.

Virtualization dynamic provisioning can help properly manage resource allocation. Virtualization allows multiple virtual machines to operate on a single physical host machine.

By precisely assigning the CPU, memory, and storage resources needed for each virtual machine, hardware resources can be maximized, minimizing waste.

• Transition to energy-efficient CPUs and GPUs.

• Consider server virtualization for more efficient utilization of physical servers.

• Upgrade storage systems to solid-state drives (SSDs) where possible.

• Choose power supplies with higher efficiency ratings.

• Invest in energy-efficient cooling solutions.

• Carry out regular hardware maintenance and upgrades.

Utilize energy-efficient hardware whenever possible. As technology advances, GPUs, CPUs, and storage devices become more efficient.

Five-year-old equipment will be far more expensive in terms of power consumption and carbon emission.

Transitioning to newer, more energy-efficient equipment can significantly reduce overall energy consumption.

Keep in mind that while these components may be more expensive upfront, they can lead to considerable savings in the long run through reduced energy consumption and lower cooling costs.

• Use efficient codecs like H.265 or EXR (with DWAA compression) to reduce file sizes and save energy.

• Encourage a culture in the VFX industry of artists working with low resolution proxy files (maintaining the same aspect ratio) or compressed files, until the versions are finalized for delivery.

• Then render the Final delivery version as high quality, uncompressed files. (i.e. work in DWA-A compressed half-res EXR files, and deliver Uncompressed full-res EXR files) This will reduce the render energy consumption during the project, and create a smaller server footprint for final project archiving.

• Implement a system to optimize render farm usage and minimize energy waste.

• Utilize lower-resolution models during previsualization and editing stages to save computational power without sacrificing quality.

Higher resolution requires more computational power and energy for rendering, longer data transfer times, increased storage space and energy consumption, higher energy consumption on workstations, and potential delays in review and retake processes.

As the resolution and fidelity of VFX projects increase, so do the computational requirements. However, it is important to question whether the increase in quality justifies the rise in computational power and energy consumption.

It is crucial to ensure that the chosen fidelity is truly necessary for the product.

• Implement a tiered storage strategy based on data access frequency.

• Use cloud storage solutions where appropriate.

However be aware:

With cloud based storage and cloud based render farms, you should encourage the use of and/or installation of cloud based storage & render farms only by companies which use servers that shut down or go into idle mode when not in use.

Products such as "Amazon CloudWatch" and Microsoft "Azure Virtual Machine" will go into idle mode when no activity is detected.

• Regularly clean up and archive old data.

• Implement Data Deduplication Software.

• Use Storage Virtualization.

Effective storage management is a cornerstone of sustainability for sustainable VFX workflows.

A strategic, tiered approach to storage can lead to substantial energy and cost savings.

Storing frequently accessed data on high-performance drives, and moving less frequently accessed data to slower, energy-efficient storage options like LTO, or to the cloud when suitable can reduce storage-relate emissions.

• If possible, implement contractual clauses that minimize data retention.

• Regularly audit data storage.

• Establish clear policies for managing client materials.

• Develop data archiving and deletion policies.

• Ensure secure data disposal.

 VFX production generates a significant volume of data, much of which is archived to cater to potential future modifications, present past work to prospective clients, and repurpose completed elements in new projects.

This practice has traditionally been seen as preserving the relationship with clients. However, the comprehensive storage and archiving of every asset and iteration can lead to substantial financial and energy costs, necessitating a reconsideration of these storage practices.

VFX houses may benefit from implementing data retention policies that encourage a more selective storage approach.

Clauses can be added to contracts specifying that VFX houses will not store or maintain materials that they don’t explicitly own. They would retain only those materials required to recreate the final deliverables, reducing storage needs and ensuring data security.

For models and assets developed for clients, these should only be retained if the client gives explicit permission or directives to do so and expire or return them after two years. This not only reduces the amount of storage needed but also respects the intellectual property rights of clients.

• Implement automated quality checks.

• Use procedural generation for creating visual elements.

• Automated Task Scheduling.

Leverage automation to optimize workflows and reduce the amount of rework and wasted computation.

This can range from automated systems for managing resource allocation, to machine learning algorithms for tasks like upscaling footage or denoising renders.

Furthermore, some tasks can be scheduled to run during off-peak hours when energy costs are lower or when renewable energy production is at its highest.

Resources for designing energy efficient systems:

Energy efficiency white paper

Energy efficiency tool

Virtualization technologies and predictive models:

VMWare

Aria Operations

Machine learning models, profiling tools, resource allocation software, benchmarking, and feedback loops:

Tensor Flow

PYTorch

Prometheus

Grafana

Energy-efficient CPUs and GPUs:

AMD

NVIDIA

Server virtualization:

Hyper V

KVM

Virtual Box

Energy-efficient cooling solutions:

Legrand

Mitsubishi

Efficient codecs:

FFmpeg

Data Deduplication Software:

Opendedup

Diskover

Storage Virtualization:

Freenaz