Remote Node try or evaluation

Hello,

I am in the process of evaluating a new compositing solutions

I currently own AfterEffect but I am not happy with it, i consider it is weak for compositing EXR from V-Ray or other 3D render, After effect is also weak in network rendering. Since version CC2014, after effect is no-longer good in using workstation with a lot of CPU core.

I have recently tried cryptomatte that is a must have in 3D compositing and a .Fuse version is available.

I have Fusion 8 free, and start to become more and more productive with it.
I have also tried NUKE, but I found Nuke quite expensive and is the last choice for me.

My big issue is: FUSION is not as fast as I need, but I have quite big render farm for 3D.
Is there a way to try or evaluate remote node before purchasing fusion. I need to evaluate the performance boost if i use 10 computers. I already do that with V-Ray Distributed rendering and it is great for 3D.

Best regards

I don't know if you can evaluate before purchase, but the render nodes help not just the final render. They work in the interactive mode too.

You can setup Fusion to use your render farm to render and cache frames around your current timeline position while you are working in the comp.

When you move forward and backward, the frames are ready without any rendering.

You can setup Fusion to use your render farm to render and cache frames around your current timeline position while you are working in the comp.

Are you sure about that? I thought clustering was removed from the current version.

Well, I do not use it myself, but accordingly to the Fusion 8 manual it is present:

Clustering
Clustering is a feature that allows the memory cache for a tool to be flled by remote workstations
during an interactive session. When the Cluster mode is enabled on a tool, Fusion will attempt to
maintain a cache for the tool in memory, using Remote Slaves on the network to assist with flling
the cache.

Well, I do not use it myself, but accordingly to the Fusion 8 manual it is present:

Clustering
Clustering is a feature that allows the memory cache for a tool to be flled by remote workstations
during an interactive session. When the Cluster mode is enabled on a tool, Fusion will attempt to
maintain a cache for the tool in memory, using Remote Slaves on the network to assist with flling
the cache.

That mode doesn't exist anymore. But it's in the docs, so it's probably valid to ask support when they expect it to be back.

Thanks all for the info.
Yes I want remote node for boosting render and interactive work.

That mode doesn't exist anymore. But it's in the docs, so it's probably valid to ask support when they expect it to be back.

@Chad Capeland This is why, I want to try before purchase Maybe somebody from BlackMagic can inform if this feature is definitively gone and if not, when it is planned to comeback and answer the try/evalutation question?

This is one feature I really wish worked in fusion. In theory it would be a significant performance boost. In practice it doesn't work.

This is one feature I really wish worked in fusion. In theory it would be a significant performance boost. In practice it doesn't work.

It used to work great. I bet they would change some aspects of the design if it were to return, but it was really neat.

HOWEVER... Remember that for many comps, the bottleneck is the I/O, not the processing. Computers have been getting faster at a rate greater than the commonly deployed network and storage infrastructure.

I find I still get faster performance when I add savers to the flow and use the network to render a preview, and use the Loader_from_saver script and work down stream with a rendered proxy than I have moving footage, renders, etc to a SSD RAID. Now there are some things that get crazy fast when moving to the SSD RAID, but overall I've consistently found using network rendering faster then using fast I/O. But again, it all depends what it is.

This is why, I want to try before purchase

Buy it, try it out, if you don't like it, point out how the documentation is false and demand a refund.

I dunno if this tactic works if I've suggested it on the company's official forum, though...

No no refund for me

One more time thanks to everybody, no blackmagic employe in this forum ?

Based on my experience with other software, regarding I/O performance on network there is some good strategy to get around:

1) copy source media on all render node.
On windows network that can be done using a batch with the command "robocopy" and the parameters (/mon:1) that monitor source folder and reexecute copyjob after each modification.
I you have many PC you can cascade copy, using one batch on each PC:
round1 = PC1->PC2
round2 = PC1->PC3 + PC2->PC4
round3 = PC1->PC5 PC2->PC6 + PC3->PC7 + PC4->PC8
Network Switch can easily support concurrent connection from PC1->PC5, PC2->PC6, PC3->PC7...

2) many mother board has two or more gigabyte connectors, with some network skills and a good managed network switch you can TEAM the connection and obtain 2Gb/s network, we have teamed 4 connection on two file server and we are near 4Gb/s max transfer.

I have applied this strategies with success for 3dsmax VRAY, I just finished a projet with 500Gb of texture and external assets, I dont want these assets to be copied nor read thru the network each time I use the distributed rendering of VRAY and the network rendering of 3dsmax.

The only question left open is how FUSION work with remote node, can we use local path / can we use UNC path, is the performance really improving, is there some optimisations... in short real performances/features versus marketed performance/features.

I also understand that this kind of use, is not the main focus of the majority of FUSION Studio customers.

But the price gap between FUSION Studio and NUKE + RENDER LICENSE is huge and can be invested in other software/plugin or hardware. At this price FUSION STUDIO can be the "killer compositing app" especially with the cryptomatte support.

The only question left open is how FUSION work with remote node, can we use local path / can we use UNC path, is the performance really improving, is there some optimisations... in short real performances/features versus marketed performance/features..

pathmaps-example.png (60.61 KiB) Viewed 1984 times

You should start exploring the Pathmap feature in the Fusion preferences which can be defined globally for all of Fusion, or defined per comp file.

A pathmap would allow you to use a relative filepath like "Pierre:/Project7/Media/image.0000.exr" in a loader or saver node. Then you can adjust what you want the first "Pierre:/" part to map to per render node/workstation so you can have different folder structures on Windows/Mac/Linux and they can all access the same media assets.

Using the pathmap approach would allow you to easily switch the media assets in your composites from being accessed off a main fileserver, over to a copy that you might have cached locally on each of the render nodes.

Also, Fusion CAN (it's an option) cache files locally to disk as they are requested. This helps in some corner cases where you might read a file off the network once, but use it multiple times in a comp that is memory limited or use it in multiple comps. So you path the files to the server, but Fusion will check to see if you have a local copy before reading off the server. Very useful to assets that get reused.