In house render farm Vs. Online farm

[Devin Johnston]

For the last 17 years I've been fortunate to have a 100 node render farm at my disposal but the office is transitioning to laptops and now we are forced to consider other options. The issue is whether it's more economical to have a small number of machines (25) that will be used as dedicated nodes or offload all of our rendering needs to an outside farm. In either case we are looking at significant costs but my own personal feeling is that having the in house machines will be less expensive in the long run. It would be great to hear from those that have dealt with this and hear your thoughts on either option.

[Marco Manunta]

We don't have many many nodes in our office but enough to understand that local render farms are perfect for still images, while online render farms fit perfectly for high end animations.

 

That said, we are aware that new technologies for animations (like unreal engine for instance, a software that we are using more and more) are manageable with local render farms, so if we take account of future improuvements (especially for exterior animations) i guess we will be good if we will enhance just our local nodes.

 

For the moment the quality that you can get with 3DS + Vray is still the best possible imho, so, for at least the next 2 years i guess we will stick locally for still images and online for huge animations.

[Justin Hunt]

using an online farm only starts to make sense if you can offload the cost to the project budgets, but quickly falls over if doing many test renders are part of your workflow, especially for design development projects. Also you have to factor in the upload and download time/cost as well.

[Chris Harris]

using an online farm only starts to make sense if you can offload the cost to the project budgets, but quickly falls over if doing many test renders are part of your workflow, especially for design development projects. Also you have to factor in the upload and download time/cost as well.

 

100% agree with Justin. Animations get even more costly with the online farms/uploading/downloading. If you can find a client to pay for it as you go, great...and good luck getting that to happen.

[Scott Schroeder]

Where I work, we are all on laptops now as well. However, we kept 22 of our old user machines for our render farm. We primarily do stills and SD/DD level design by rendering work, so the online rendering just wasn't going to work with the data transfer needs. Even something set up with AWS and Deadline was not anywhere close to being as fast and iterative as our in-house farm.

 

Our corporate IT was looking at cutting down our render farm as they are concerned with fees associated with it such as windows licenses, etc. Yet, we were able to prove that will an in-house render farm may have more steep upfront costs, it pays for itself within a year of use and then those fees are not coming back until you upgrade in 5-6 years if you purchased quality equipment.

 

When we do our 1-2 animations per year, we tend to do them with an online render farm. We roll the cost into the animation fee but we are very clear in our contract with the client that after the animation is sent off, any changes will require additional fee as we'll need to resend it all back to the online service. What tends to sell the extra fee for online rendering to the client is that if we do it in house, it will had 1-2 weeks to the production time. If we do it online, we can get it in their hands faster and they usually will always go for the faster timeline.

 

While I'm a fan of the mobility provided by the laptop, the workflow with it while sitting at your desk isn't there with what a desktop can provide.

[Francisco Penaloza]

We had a similar issue, but it was triggered for a new security system that IT needed it to install for a few 'special projects'

We used to have access to the floor computers about 40 or so, that we thought they had enough RAM for our project.

Fast forward on time now we have only 2 dual 28 cores dedicated render nodes. These are very robust machines actually and they can process in the same time frame that our older 40 machines or faster. Less hassle also updating software and debugging issues.

If we need more power than that then we use rebusfarm or similar.

 

Now if we require to use an external render farm, then the project will be able to pay for that 'extra' cost. Other than that our render nodes also have a large chassis so we placed a couple of 1080 ti to use when Lumion renders are good enough.

 

For us I think is almost a happy medium, I would like to have other two of those render nodes to reduce the amount of outsourcing usage, but honestly lately I am doing more Lumion animations than VRay.

 

On the flip side, those render nodes also help us to bake our Unreal projects and it is interesting to see since everything seems to be moving into real-time production software, having a large number of render nodes in house doesn't seem to be necessary anymore, at least for us.

[Scott Dombrowski]

We've had just two dedicated nodes in-house for years. They help crank out still renders either via DR or Backburner. We only do a couple of animations per year, and our workflow is switching over to realtime for that. The render nodes help on those by handling the light bakes. I used Rebus a few times in the past and the process seemed pretty finicky. I had to bake all the Civil View traffic because their plugin checker didn't recognize Civil View, despite it being built-in to Max. Getting the scene ready for Rebus took more time than anticipated, and downloading frames afterwards took a lot longer than anticipated. Overall, it was still a lot faster than rendering 6000 frames in-house. Our clients are perfectly happy with game engine quality animations and we like not having to set aside as much time for output as we would with traditional rendering. Plus, we get a VR capable scene out of it as a bonus.

 

tl;dr: Small in-house farm for us.

[Devin Johnston]

Thanks everyone for these great comments!

 

I ended up testing 3 different farms and considering that we do 10-15 animations a year it makes a lot more since to go with the in-house option. Some of that workload can be done with Lumion or Unreal but we also do a lot of composite animation using drone footage. As far as I know you can't import tracking data into Unreal yet even using Datasmith so we have to stick with doing it the old fashioned way. Outsourcing sounds great because you don't have to deal with upgrading and licenses but it's easily twice as expensive as buying and maintaining the machines yourself.

[Chris Harris]

Out of curiosity Devin, what kind of drone and camera on-board with the drone are you using?

[Scott Schroeder]

Thanks everyone for these great comments!

As far as I know you can't import tracking data into Unreal yet even using Datasmith so we have to stick with doing it the old fashioned way.

 

Pretty soon you'll be able to import an animated camera into Unreal via Datasmith. I imagine the workflow would be AE for tracking -> Max -> Datasmith -> Unreal.

 

It may already be available in 4.22. I have not tested it at all yet as I'm just discovering this myself.

https://forums.unrealengine.com/unreal-engine/unreal-studio/1563173-datasmith-camera-animations

[Devin Johnston]

I use either the DJI Phantom 4 Pro or the Inspire, they are both terrifically stable platforms and the footage you can capture is amazing.

[Devin Johnston]

Pretty soon you'll be able to import an animated camera into Unreal via Datasmith. I imagine the workflow would be AE for tracking -> Max -> Datasmith -> Unreal.

 

It may already be available in 4.22. I have not tested it at all yet as I'm just discovering this myself.

https://forums.unrealengine.com/unreal-engine/unreal-studio/1563173-datasmith-camera-animations

 

I should have been more clear, you can import animated cameras into Unreal from Max but that doesn't mean it will work with motion tracked footage. The last time I tried this I was able to get the camera in but I couldn't get the footage and camera to match up.

[Corey Beaulieu]

I realize that this post took a turn towards drones, but in the interest of Farm Local v. Farm Outsource I have to say this:

 

I am lucky enough to work for a company that has over 100 nodes overnight. We have 80-ish during business hours too. It's as dreamy as you would imagine. We are a fully rendering studio though and we complete over 50 animations a year as well as many, many different VRs and Stills.

 

What I have to offer to the conversation is that Rome wasn't built in a day. We started years before me and small and have added over time. We upgrade workstations (which are not as buff as they could be to save on cost) every few years and add to our farm every few years. Local doesn't need to be so crazy when you can check in the farm. When very large projects come in we buy more machines, maybe 5 or 10. When very, very large projects come in, ones that would take down any farm, we budget new machines in the contract to better the schedule. This is rare, but fits for some clients when you say you can have it on 2-3 months v. 6-8 over their render requirement (4k films are a beast!).

 

I know I am a lucky artist and I know that my situation is rare, but building a farm a little bit here or there is a great way to go. Get good stuff at first and upgrade memory over time. CPU's can last a while if you start high.

 

The one cost that you MUST have if you are building a farm though, is an IT Manager. Ours is invaluable on many levels, but when it comes to our large farm, I could not imagine having to wrangle that as well.

 

You cannot forget the cost of managing, Cooling, Housing, and Electrifying a farm when looking at the cost of it. I would imagine that I could make a few more dollars annually were it not for this, but I don't know that I would make that trade having it.