Progress in the 3D industry can be overwhelming. Watch a CG classic from the last three decades like, DreamWorks’ Shrek (2001), and notice the difference in animation quality compared to now.
In hindsight, Danny’s musing of us “on the precipice of a modeling revolution” doesn’t seem far off. Along with how 3D animation continues to sweep media, gaming, and architecture, our 2019 coverage of 3D trends hint that we have (maybe unwittingly) been thrust off the edge. The bar rises as industry technology evolves: service quality and expertise rush to keep pace.
For people like us—modest studio heads, 3D freelancers, the good ol’ working class creatives—it boils down to stronger competition, where frequently churning out output, accepting large-scale jobs, and accomplishing tasks faster are tried-and-proven ways to get ahead and provide bread on the table. It’s a race: the name of the game is speed. Although, that may be a problem where rendering is involved.
Rendering is mainly dependent on hardware capability, the root of financial hurt. It can put a real damper on showcasing technical skill and creativity, too. Without the appropriate machines, the photorealism of a sequence may have to be toned down just for it to see the light of day (or more accurately, meet the deadline). Even if “keeping it photo-real” isn’t the goal, the time allotted for a project render often produces the largest bottleneck in most workflows. Rendering is time-consuming and resource-intensive. People want none of either or both.
Regardless, it’s more a matter of circumstance than control. A 3D artist in the architecture niche whose deliverables are interiors may have a more consistent scale and predictable average render time than that of a 3D generalist.
Anyhow, it may be time to ask: Should you offload to third-party render farms? Or should you consider rendering inhouse?
This article will go through the benefits and drawbacks of both options. But before we start, there are other things you can do aside from increasing or improving hardware. For example, there’s a study from Clemson University that compares different render managers and how they affect 3D artist pipelines. Our blog, meanwhile, shares in-depth testing strategies and ways to detect possible problems early on, and trade secrets on optimizing scenes for rendering through geometry and textures in 3ds Max.
These resources are free. Thankfully too, because rendering is hardly mindless work.
There’s a lot to consider when investing in an inhouse (budget, space, electricals, and so on); the best way to go about it is to structure them according to the main categories of software and hardware. Though much of the encouragement to build your own may be met with an equal amount of dread, nothing spells DIY better than building and maintaining your own render farm. Like any independent endeavor, what you get is directly proportional to your investment.
Software sets the floor and ceiling for your budget and determines how the other costs fluctuate. Apart from compatibility, whether the software license is single-use or floating plays a big part in your expenditures and ties directly to how many nodes you would need. The average annual price for Renderman is £680 (840 USD) per license. Creating with Guerrilla Station, meanwhile, is free, but it’s £630 (787 USD) per node in a 10–19 cluster to use Guerrilla Render.
Übel’s review of render engines for VFX, gaming, and architecture and Yamazaki’s updated list of 3D renderers are great places to start surveying for software options. As you go, keep the following points specific for a render farm in mind:
The golden rule is to always consider your deliverables. Render quality is a balance between speed and accuracy; if your line of work (medicine, character design) leans toward stylized outputs over realism, accuracy becomes less of a concern. On the other hand, while Blender’s Cycles and EEVEE are free, it pays to look at the distinctive software features: how do they compare in render times (bias) and render quality (GI)? What, pertaining to your niche, works for you?
Hardware is an upfront cost, but that doesn’t mean it’s one-off. In fact, it’s depreciable: the best stuff becomes second-rate two years in.
There are upsides to that though. For one, you know that one guy who nabs the hottest products fresh out the market?
Another is that customizing nodes from different components improves performance cheaper. As you will need paid software licenses for each node, performance per machine is essential. That would also require maintenance, which may mean additional staffing.
Nodes need to be linked in a shared network to split tasks. An insight on reducing render times is parallelization: according to Amdahl’s Law, speedup diminishes when file volume is spread too thin among too many nodes because of resulting data traffic. In other words, a still renders faster in a network of five nodes instead of 50.
Based on your most common render type, this equation can help estimate your benchmark number of nodes.
Big or small, your render farm will make noise. A lot of noise. Best to keep it secluded. It should have a cooling system, too, else you’ll feel the effects of global warming sooner than 2050.
Your space must also have sufficient ampere capacity for your nodes. Two machines alone could damage your circuit breaker under usual conditions (15–20 amperes per outlet). Unless you’re willing to connect to outlets from all over, it’s best to consult with your electrician.
Though electricity rates and surcharges may vary, electricity overhead can cost more than the sum of all the others if one is not careful. For residents in the UK, you can refer to electrical prices here; for those in the US, click this. Regardless, the formula’s the same wherever you are: Watts × KW/hr.
One way to reduce electricity cost is power-cycling. Although, there’s a reason why big animation studios have 24/7/365 in-house render farms: turning individual nodes on and off may lessen overhead, but significantly reduces your hardware’s lifespan.
There’s also the guilt of CO2 emissions. You can calculate your carbon footprint here.
What matters most when considering the benefits and drawbacks is that an in-house render farm is variable to your situation, and mainly how deep your commitments go. Fortunately having a spare room available, being located near a data center, or owning spare hardware are all convenient, sure, but convenience isn’t an investment. You could also shell out less financially, but it balances out through time and energy spent. Maintenance, staffing, and overall upkeep constitute a different beast altogether.
The best way to frame in-house render farms is that they are a continuous investment. You will break-even eventually, how long until that happens is another matter.
There are tradeoffs to everything. Time, money, quality. For render farms in general, you can only pick two. One way or another, it can’t be inexpensive; it’s unlikely it’ll help save the dolphins either.
It’s the main reason why third-party render farms aren’t free. The changes in the 3D industry also affect how we relate to work. With the focus on content creation, it’s no wonder why bottlenecks happen during rendering instead of anywhere else—and let’s face it, that defaults rendering as a hassle.
In our render farm tutorial, Michał Moś provides an overview of different third-party render farm models. The guide’s also instructional with estimating render farm costs and render times; an invaluable way to assess your current/prospective benchmark and compare with third-party services like ours.
More important is how each model has its pros and cons. But the best render farm service is the most hands-off, and it’s that which makes network rendering one of the more ‘revolutionary’ aspects of the evolving industry.
As renders take place in a virtual cloud, the highest possible costs are no longer an issue: space and electricals. Upkeep is not a problem in flexible pay-per-use models. Additionally, you manage your scene through a web-based dashboard. As soon as the files are uploading, your workstation’s free to work on other projects or cool off.
Related is that the service is automated, which comes equipped with a queue manager that distributes file volume to available nodes. Cloud rental services provide priority systems that allow users to flexibly manage their jobs. Urgent requests could go through the premium service to cut time; low-priority jobs, on the other hand, cost less when time permits. In GarageFarm.NET, the longest turnaround time is 1—3 days, while optimizing scenes and reevaluating production workflows remain as cost-efficient ways to break-even without having to break much in the first place.
As mentioned, each render farm model has its cons. Just the same, network rendering has pitfalls that result in unusable frames:
* For questions specific to our service, we’re flattered. View our FAQ.
These issues rear the ugly side of the pay-per-use model. As the task is automated, the best you can do is monitor your project and cancel the job when renders fail. The absolute letdown is wasting the money or credits spent on node usage for useless frames.
So, why cloud rendering anyway?
The inherent benefits of cloud render farms outmatch the others, sure. Frankly, it’s why there are so many of them. The main feature is that it hands-off, especially compared to in-house, but what makes up most of that depends on the quality service care and tech support of the farm you choose.
Since you’ve read this far, GarageFarm does it by providing users real-time communication lines with a 24/7 support team of sysadmins, TDs, developers, and wranglers. But don’t let their eyebags fool you: they monitor your job along with you, and notify you immediately when issues arise. The best part is that they don’t inform you just so you can cancel anyway and prepare the usage cost. They intervene when necessary to prevent a waste of credits.
In the end, choosing between building in-house and offloading to third-party render farms is yours to make. A common scenario, though, would be to assign small sequences and previews in-house then offload large-scale productions or time-pressured projects. Larger studios often make use of in-house render farms for standard deliverables, but still, outsource heavier scenes and turnarounds to third parties.
What happens after that is up to you. The 3D industry is ever-changing. We change with it whether we do something about it or nothing at all. It’s case-to-case, and you can look forward to circumstances falling into your favor sooner than you would expect.
Depreciable but capable hardware may cheapen. Prices for online render services could drop. New educational sources and hacks spring up, and there’s always honest work for the dedicated creative.
What matters is that great options to improve are hardly in short supply. Enjoy the push and pull. After all, part of being a 3D artist is that you innovate. You never just “make do” with what you have.
You also make with.