We render 10-20% of all our work on cloud render farms like FoxRender to keep costs down, whilst maintaining our competitive edge when deadlines get tight.
Rendering shouldn’t feel like spending.
But when entirely reliant on cloud-based rendering, that’s often exactly what it feels like.
In my experience running a CGI Creative Studio, when artists are given the option between rendering in the cloud (faster, but costs credits) or locally (slower, but “free”), they’ll almost always choose the local option. Even if they’ve been told the project budget already accounts for cloud usage, and even if we tell them to use the cloud — If they can wait, they’ll rather render locally to avoid “spending.”
And honestly? As a studio owner, I appreciate saving money! haha
But there are times when time really is money — and waiting just isn’t an option. Like when our local render farm is at full capacity. That’s when cloud rendering becomes the hopefully obvious solution.
So which is better? Cloud rendering or building and running a local render farm? Like most things, the answer isn’t all or nothing. It’s balance.
That’s why I’ve built our studio on a hybrid solution where we render most of our work on our own local farm — and lean on cloud rendering services like Fox Renderfarm (more on this choice later) when we need to scale.
It’s not much, but the bottom three machines house 10x RTX3080Tis, totalling 10 GPU render nodes, and beefy CPU nodes. This little farm really helps to churn out our renders when the rest of the artists’ machines are in use.
Why Local Render Farms Still Matter
Cloud rendering has come a long way. It’s fast and –relatively— frictionless.
Back in the day, sending 3D scenes to a cloud render farm meant spending time to conform your entire scene to that farm. Tedious tasks like baking caches, fixing file paths, and praying you wont have errors or missing plugins etc. These days, for us at least, it’s drag and drop — Fox Renderfarm, for example, mirrors your entire folder structure and just gets on with it.
So if it’s so easy now… is there still a point to building and maintaining a local farm?
Short answer: Yes.
Long answer: Unless your whole studio is fully cloud-native (AWS, etc.), there’s more friction than you think.
Our studio is NOT on the cloud. Real people on real physical machines on a 10Gbe network.
The ‘mental’ hurdle, the silent bottleneck.
Just the thought of sending something to an external farm — even if it’s only a few extra clicks — still presents a “mental” hurdle. For quick tests, lighting passes, and R&D lookdev, this slight hesitation or extra steps quickly compound into a silent bottleneck, or a exponential blow to our annual expenses.
While local render farms are expensive to build and require maintenance and troubleshooting, they afford the studio speed, availability, and the mental freedom to experiment without “cost”; In my view, one of the most important parts of creative work is ease of exploration. – You seldom find the best version of something on the first try.
Keeping beautiful ideas alive.
If each route to explore is met with compounding hurdles, it’s not long before the artist is compelled to prematurely end the creative process and settle for whatever is left on the table. When every render becomes a cost decision every single time, exploration gets throttled, ideas don’t fully form. And that’s a problem.
Having a local farm means the team can test, break things, run ideas, make mistakes and iterate without hesitation. And that is really important because I believe ‘hesitation’ is the regretful place where beautiful ideas quietly fade and die.
So… When Do We Use the Cloud?
The trick is finding the right balance.
From my experience building and maintaining our entire workstation/render infrastructure, we need our local render capabilities to at least cover about 80-90% of our render load. That’s our local baseline.
For the remaining 10-20%—typically only during crunch-time—we scale up on the cloud.
This gives us the best of both worlds:
-
- Artists don’t feel pressured to “spend” when iterating work.
- We can instantly expand when we’re slammed.
Final Thought: Build for Freedom, Not Just Power
I used to fantasize about having a fully local render farm that could handle everything.
But running a farm that size just to handle the extra 10-20% occasional crunch-time, doesn’t warrant its 300-400% added cost and complications.
Cloud rendering gives us elasticity–It scales when we need it, and saves us from the heavy upfront cost.
But relying entirely on cloud rendering? That limits us too.
In 2021 we bought 10 RTX3080Tis as they provided a great cost-performance for the work we did. Giving us 10 render nodes with only 3 machines and 3 redshift licenses.
Over the years, I’ve learned to build our render capabilities for freedom — not just power.
This freedom has given our artists more room to grow, our pipeline more flexibility, and our projects more headroom to explore.
With that said, our 30-series cards are nearing their end of effectiveness for current CGI demands. And we’ll be looking to make an upgrade to the new cards some time in 2026/7. 🙂
If you’re looking to build your own local render farm, you might want to consider reading about my experience building ours. – Read Here
