CDW, AMD, NVIDIA & Dell Help M&E Studios Meet Their Rigorous Computing Demands

With the growing complexity of media workloads, studios often hit roadblocks in fulfilling their compute and artist workstation resource requirements. Here are five of their biggest challenges:

Significant financial pressure placed on the M&E industry in the last few years has resulted in budget cuts.  As a result, studios were forced to reduce in size and were challenged by a lack of investment to refresh compute and artist workstation infrastructure. Many of the studios scaled back their offices and infrastructure alongside a reduction in head count. 

As the industry turns around, studios that have been awarded projects need to ramp up with artists, secure artist workstations and scale up their render capacity. When it comes to artist workstations, the cost, physical footprint and complexity to deploy and support a fleet of towers isn’t practical.

Unfortunately, the large demand for CPU and GPU for AI workloads has made it cost prohibitive to rent from the hyperscalers since these resources are being rented at a premium price outside of M&E for machine learning and other AI workloads. As a result, the hyperscalers are either too expensive or there is limited availability. 

Despite the advancement in GPU technologies, D’Amato noted, “The majority of media rendering is still CPU-based today for final frames, particularly for live-action cinema.”

While GPUs are used for some content creation and animation, typically, they haven't replaced CPUs for critical, photorealistic final renders. “Legacy render software and the specific computational approaches for photorealistic lighting and shading in final renders seem to be better suited to CPU workflows,” Kostiuk detailed.

This explains why demand for high-density CPU clusters is still increasing as studios explore ways to render their media workloads most efficiently.

In parallel, demand for new GPU for artist workstations and generative AI workloads has increased. Virtual desktop infrastructure (VDI) meets the requirement to offer flexible options for hiring remote artists and delivers the benefits to productivity, density of deployment along with the cost savings in reduced power and office footprint. The high core count CPUs and large memory GPUs that make up the VDI server can also be used after hours to add capacity to the render farm and for generative AI workloads.

The upfront capital expenditure to purchase new render server and artist workstations may be difficult for the studios.  After some challenging years, studio balance sheets and credit have been impacted at the same time as interest rates have risen, resulting in higher borrowing costs. 

Studios require a partner, like CDW StudioCloud, that can offer greater flexibility when it comes to cost-effective financing, preferably a monthly payment plan that can align with the project-based nature of the M&E industry and doesn’t tie up the studios’ access to credit.

While core count is important, D'Amato emphasized that “Main memory or RAM is really the constraint that most of our clients are seeing in their render jobs.”

Studios are often constrained by the memory demands of their large-scale rendering tasks, leading to render bottlenecks if main memory or GPU memory is insufficient. When it comes to artist workstations, the GPU memory required for each artist is also increasing. 

Constraints around memory create longer render times, particularly for high-consumption frames. This can lead to production slowdowns and less artist iteration, impacting creativity.

The increased demand for main memory and GPU memory for real-time image generation and other AI workloads is increasing the price to purchase or rent from hyperscalers.  There is also a scarcity of supply, making it difficult to secure increased memory capacity, driving up the price or eliminating the availability and/or viability of spot rentals.  

Increasingly studios need a partner, like CDW StudioCloud, that can leverage vendor partnerships and buying power to deliver preferred and volume pricing for main memory and GPU memory.

High-performance computing requires significant power and that power generates a lot of heat, resulting in massive and expensive cooling within the data centre facility. The latest generation of CPUs and GPUs are each 500W or more, placing significant strain on power and cooling.

However, retrofitting existing data centres with liquid cooling is expensive and time-consuming. As Kostiuk noted, “To deploy next-generation CPUs and GPUs at full scale, the industry would have to cross the chasm from where we are at currently, going from 35 kilowatt racks in a complete air-cooled system to 100+ kilowatt hybrid liquid cooling infrastructure.”

Moreover, the cost and lead time required for building new data centre halls with hybrid liquid cooling can prove to be challenging, making it difficult to scale rapidly and cost-effectively.

Studios need a partner, like CDW StudioCloud, that has negotiated long-term contracts for space and power to provide the most cost-effective hosting.

Studios need high-performance CPU and GPU servers for sustained usage during the production process. This footprint of CPU and GPU servers required throughout the entire production process is referred to as the studio’s baseline or sustained requirement.

This baseline requires either large, upfront capital investments or premium priced long-term rental rates charged by the hyperscalers. This upfront payment or premium rental puts financial pressure on the studios that often leads to deploying lower performance solutions or going for longer periods between technology refreshes. 

This can lead to reduced capacity for the studio to generate images or increased overtime strain to complete projects. 

For periods of peak demand, short-term burst render and workstation rentals and their associated premium rental rates from the hyperscalers place further financial strain on the studios. These premium priced short-term rental charges balloon further in size and scope when cloud storage, data movement and premium bandwidth charges are layered into the total hyperscaler cloud rental cost. 

To add further insult to injury, even though the bandwidth is charged at a premium price, this bandwidth could also represent severe latency challenges for the studios as they become bottlenecked by network latency to the cloud. That is because hyperscalers usually offer lower bandwidth direct connects peaking at only 10Gb/s. 

There is also no flexibility to deploy studio owned technology in the adjoining racks within the cloud, so workflow architectures become complex and expensive.

Studios need a partner, like CDW StudioCloud, that offers flexible rent-to-own technology and rental rates that match and align to a competitively priced three-year or five-year total cost of ownership. They also need a partner that can host the studio’s owned assets next to the rental fleet for efficiency and 100Gb/s bandwidth that has a simple and secure air gapped architecture.

Studios understand the critical nature of security threats and recognize the very real fact that a security breach or ransomware infiltration has become a 100 percent inevitability. 

Despite recognizing the threat is real, significant financial pressure placed on the M&E industry in the last few years has resulted in budget cuts. 

As a result, studios reduced their size and may have had to prioritize spending on other critical needs, sometimes at the expense of security investments. Many of the studios scaled back their offices and infrastructure alongside a reduction in head count. In some cases, this has led to embracing hybrid cloud-based architectures and artists working remotely, which has introduced new security threat vectors.

As studios are awarded projects, they need to ramp up with artists and provide assurance to their clients they have effectively secured their hybrid cloud infrastructure and remote artist workflows. In parallel, as new and more sophisticated AI enhanced security threats emerge, there is a requirement to refresh, expand and harden their security posture against new security breaches and next generation ransomware attacks.

Studios need a partner, like CDW StudioCloud, that has their own robust security and risk advisory service. They need a partner that embraces MPA best practices and is a Gold Shield Trusted Partner Network member that operates their service in Trusted Partner Network data centres.

CDW StudioCloud offers a comprehensive solution designed to tackle render and VDI server challenges faced by studios and meet performance requirements. Here are some of the ways we can help M&E studios tackle these challenges.

CDW StudioCloud maximizes performance for studio workloads with efficient infrastructure design. It leverages AMD EPYC CPUs that can fit more cores in the same rack space without increasing power or energy consumption.

This architecture yields higher CPU core capacity for rendering jobs. Using this design, the StudioCloud solution can now offer 192 cores of compute per CPU with up to 3 TB of RAM.  

“Over time, as AMD rolled out its new processors, we started to consolidate our offering. Using AMD processors, we can get more performance into a single rack,” said Kostiuk.  Each generation of AMD EPYC CPU increases in core count but also increases in per core performance with improved memory bandwidth.

This way, in the new render deployments, CDW StudioCloud can not only deliver core density but also 20 percent faster performance that is paired with up to 8 GB of RAM per core. This allows for more efficient core utilization, especially for RAM-intensive jobs, and significantly more concurrent frames per rack, even for monster frames.

For virtual desktop infrastructure (VDI) use cases, the emphasis is on higher CPU frequency over maximum core count. The AMD EPYC 9575F, with its higher frequency, is considered an optimal SKU for VDI, as most content creation applications benefit more from higher frequency than extensive hyperthreading. The StudioCloud VDI servers deliver the perfect balance of higher frequency AMD CPU while still delivering the core density to support even more artists. The CPUs can therefore be leveraged extensively for the render farm after hours too.

NVIDIA GPUs are integral to CDW StudioCloud capabilities, especially for VDI and emerging AI/ML workflows. 

CDW recently introduced the NVIDIA RTX PRO 6000 Blackwell Server Edition as a key upgrade for M&E clients’ artist workstations. This high-performance data centre GPU features 96 GB of GPU memory and a memory bandwidth of 1.6 TB per second using DDR7. When paired in a Dell R7725 server, this doubles the GPU memory to 192 GB per server.

NVIDIA RTX PRO 6000 Blackwell Server Edition was specifically designed for the deployment of powerful VDI environments.

This holistic approach, integrating NVIDIA GPUs with optimised configurations with AMD CPUs tailor made for high core count and high frequency on Dell servers, allows for remarkable density. StudioCloud deployments can accommodate up to 16 artists per server and as many as 224 artists in a single rack.

Such studio workflows, leveraging VDI, prove to be more efficient than traditional tower-based setups while offering scalability in server-based deployments.

A standout feature of the NVIDIA Blackwell Server Edition is its Universal MIG technology. This allows the GPU to be segregated into hard, dedicated portions for each user.

This eliminates the noisy neighbour problem, where one artist's heavy workload could impact others sharing the same GPU. This ensures guaranteed quality of service (QoS), vital for critical tasks like client reviews or complex simulations generated by effects artists.

“If they put two artists on a single MIG instance, they're still going to affect each other potentially. But the frequency of that happening or the number of people who would be affected by that happening would be greatly reduced,” D’Amato explained.

CDW StudioCloud seeks to provide an air-cooled solution as an interim step, maintaining a 35 kilowatt headroom per rack. This strategic choice allows for immediate technology rollout within existing data centre infrastructure, avoiding the high cost and lead times associated with hybrid liquid cooling upgrades at the data centre.

Kostiuk reiterated, “We're able to deliver an upgraded air-cooled solution and better still we are able to roll out this technology immediately within our existing data centre partner infrastructure.”

While new, high-performance CPUs necessitate a shift to one-unit (1U) servers with two processors from previous two-unit (2U) servers with four blades, CDW is still able to deliver more performance in a single rack. This re-imagined approach, focusing on a fully air-cooled solution, ensures density while managing heat.

This immediate offering smooths out the transition for our clients to adopt hybrid liquid cooled solutions when the infrastructure has scaled up to support those future deployments.

CDW StudioCloud leverages the best technology to deliver the perfect intersection of greatest density and most power-efficient solution at the lowest price. Power efficiency is especially important as saving power also saves money. 

CDW standardizes on Dell servers for StudioCloud, owing to the strength of our partnership. CDW leverages our Dell Titanium Black Cloud Service Provider Partnership and CDW’s volume buying power with Dell. That partnership and commitment to success is extensive.

Keeping power and cooling costs down while maximizing server rack density is the name of the game. The Dell PowerEdge R7725 server, for instance, offers significant CPU consolidation by replacing multiple legacy servers, which lowers power usage and boosts performance. The StudioCloud Operations team and our clients also appreciate the server management capabilities offered by Dell servers. Its orchestration-friendly management interface also simplifies server scaling, monitoring and automation tasks.

Combined with its robust security features and cyber resilient design, the Dell R7725 enables data centres to operate with lower costs while maximizing core density, running AMD-based processors packaged with NVIDIA GPU with greater agility and efficiency.

The new StudioCloud NVIDIA GPU servers are capable of handling a variety of production workloads, including VDI for dense deployment of VFX artist workstations, CPU and GPU rendering but also generative AI and machine learning.

Studios can position their investment in artist workstations while expanding their render footprint and deploy infrastructure for generative AI workflows.

Furthermore, when not being used for production, studios benefit from leveraging all the performance of the investment in NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs to generate millions of AI tokens at a fraction of the cost charged by traditional subscription based AI services like ChatGPT.

CDW StudioCloud ensures its services are fully compliant with the Motion Picture Association's Trusted Partner Network (TPN) security protocol. In fact, CDW StudioCloud is recognized as a TPN Gold Shield partner. This eliminates a significant hurdle for studios, which face rigorous and frequent audits themselves.

To achieve this high level of security, CDW StudioCloud maintains completely air-gapped environments between clients. This involves a dedicated core switch for every client in each data centre, hardwired directly to their studio, with no virtualised networking or software-defined networking between client environments.

“The fully air gapped architecture of StudioCloud and our TPN Gold Shield status means we're the most straight forward part of any studio's audit. We know our clients must be MPA compliant, so, this is what we do to make sure StudioCloud is easy for their assessment,” D’Amato said.

The team behind CDW StudioCloud includes long-time M&E veterans each with over 20 years of technology as well as M&E vertical expertise. This means clients benefit from real-life, practical M&E experience, which is fully dedicated to resolving the unique challenges of clients in the industry.

With a large footprint of servers, CDW StudioCloud is able to offer simple OPEX or CAPEX cost structures for clients, including rent-to-own, improving price-to-performance ratio as well as total cost of ownership (TCO).

In comparison to hyperscalers, where compute instances are designed to be cost-effective for on-demand needs, StudioCloud is optimised for sustained periods of time for final renders and offering extra capacity during a studio’s peak use. This aligns perfectly with studios' needs and offers better economics.

Clients also benefit from dedicated 100 gigabits per second (Gb/s) connectivity to their compute resources. Critically, CDW supports clients bringing their own storage, allowing direct, fast connections to both render and storage without egress charges.

CDW StudioCloud offers a variety of ownership options in terms of server purchase, rent-to-own or pure rental. The rent-to-own model, particularly popular for larger investments like studio storage, offers monthly payments over time, allowing clients to eventually own the hardware.

This flexibility allows studios to manage capital expenditure (CAPEX) versus operational expenditure (OPEX) and benefits from CDW's lower cost of capital, translating to further savings for clients. “Clients love the flexibility of working with us and not having to build out and support their infrastructure nor tie up their credit,” Kostiuk said.

The upfront capital expenditure to purchase new compute, workstations, data storage and security infrastructure may be difficult for the studios. After some challenging years, studio balance sheets and credit have been impacted at the same time as interest rates have risen, resulting in higher borrowing costs.  

CDW StudioCloud provides the flexibility to make monthly payments on one-year to five-year term rent-to-own plans for new security infrastructure that provide very low financing without tying up a studio's access to credit.

The bar for high-quality visual effects and innovative storytelling shifts every year. As your studio plans to work on the most challenging projects, CDW StudioCloud can help meet your computing needs in a unified offering that’s high-performing and predictable.

CDW StudioCloud equips you with a dense, secure and flexible rendering platform as well as the best of breed technology for high performance artist workstations. StudioCloud delivers high-performance compute that accelerates production pipelines, optimizes costs and scales securely, making it a formidable solution for heavy generative image creation and rendering demands.