Correct composition? and performance issues

#1

I have a basic question about making a composition.

Usually I only use a tree footages, a beauty pass, a idmatte pass and occlusion pass, and I do compositing by linear mode, adding color correct operator and others operators, with masking. Following this schema:

Well, I don’t know if this method is correct thinking in a performance composition, or is better some this:

Or, what method is correct for a simply but big composition?

I work with 8K resolution still frames, get footages from a Network Attached Storage, and I have serious problems for interactive rendering proccess, if not in proxy mode, include using ROI.

I try to caching footages and some operators but the performance is still slow.

System is a workstation 8-core Intel Xeon, with 12GB RAM and Quadro FX3700 Graphic card.

#2

Your computer specs seem to be enough for this workload, except for I/O speeds.

I’m not sure what performance you are getting out of your network, or what hard drive speeds you have available, but to work with an 8k frame interactively you need a very fast access speeds, hopefully something around 500mbps.

I would also stay away from working off of a network if you are looking for interactivity, unless you have some massive bandwidth.

#3

Well, my network speed is 1Gbps, but its rate transfer is not great of 500/600Mbps, because not by the NAS and switch, else because gigabit ethernet of Fusion workstation I think is a little bottleneck.

Local hard drives are SATA, and its average read-write benchmarks are about 60-40MB/s respectively.

Usually the footages that I use in my compositions are .tiff without compression and 16bit colordepth. This entail about 200MB by file. Note that only one file (beauty pass) is 16bit. occlusion and idmatte pass are 8bit and .tif with LZW compression, and entail about 20MB and 2MB respectively.

Roughly, the base workload are about 300MB only for footages.

Another thing, is that I put a CineonLog2 to convert beauty pass footage at starting point in composition, change linear to log, and also have control for colordepth; change a 16bit float or 32bit float. I know this increase the workload, but in some situations is absolut necessary working in float point.

However, I’m testing somethings this days.

-I also compressing a beauty pass with LZW and reduced to 8bit in the tif file, entailing about a 50MB/file.

-I turn to 8bit the cineonlog2 operator for interactive rendering. For production render, turn to 16bit again

-I put a set domain at start of each footage, when I need view all composition in full resolution (proxy off), but only for view a specified region.

Basically with this three points, at moment I have been able to enhanced my workflow with bigs footages.

But I have some questions about Fusion manage the loaders.

Would be the same load a indentical bitmap but with or without compression? I have this question because some applications, when load a bitmap from disk, if file is a jpg of 5MB or a tif of 80MB, in memory is always 80MB.
However I see how Fusion read the loader more one time, when compute the compositiion.
Is not loaded in memory for the first time is readed?

#4

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.