RR collects a lot of statistic information.
Please see the description of the python module libpyRR2_datafiles.pyd
At first, you need to be able to render an image that represents the average render time of one of your final frames.
If you have such a scene file, Submit the job it to one machine on the farm (Preferable a Daywalker).
You might want to increase the frames step that for example every 20th frame is rendered only. As you only need to get an average of a few frames.
Wait until the job finished and check the average frame time information in the job table in rrControl.
If you want to calculate without RR, then you have to get a stopwatch.
Start Maya from scratch that it does not have anything pre-loaded. Wait until Maya has loaded all plugins.
Open the output console and the script console to have some additional information about the current status of Maya.
Start the stopwatch and load your scene file. Stop after the scene is loaded and everything is displayed. The scene load time is the first time you need to take into account.
Start the stopwatch, render a frame, stop it. You can collect multiple frames as well for a better result.
Add the frame time with 1/10 of your scene load time. Now you have the average frame time on a farm.
In this example I use 25min/frame.
As almost all calculations are in hours, we have to convert it.
25min/ frame / 60 min = 0.417 h/frame
Now you have to know the speed of your test machine in a value that can be compared with other clients.
In the industry the speed of a client is measured in Ghz.
You multiple the number of real, physical cores by the ammount of Ghz they have.
If hyperthreading doubles the amount of virtual cores, then do not multiple it by 2, multiply it by 1.3. (+30% is the increased speed with Hyperthreading that we and other companies have tested)
RR tests the speed of all clients at installation.
This value is more reliable as an older generation of CPUs is slower than a newer generation of CPUs, even if they have the same Ghz.
The value is displayed in RR as PS (ProcessorSpeed).
You can either use Ghz or PS, but I recommend using PS.
In rrControl, PS and Ghz are listed in the client table.
(If it is not visible, then enable the column via the menu Layout)
In this example, the machine has as PS of 200 and a total Ghz of 30.
Of course you need to total number of frames that are rendered on the farm.
In this example 1000 frames.
In a perfect world you would setup and shade your scene to render it only once.
But in the real world, you see animation errors in final rendering or some shot does not fit with the next one and has to be re-rendered.
Therefore you have to multiply the number of frames by some value from your experience.
As far as I remember the vfx feature Gravity had a rate of 1.4, which is reeaaly low, it is usually higher for other productions (1.7).
If the shading has some new look, the re-render rate will be higher.
In this example I use a rate of 2. Which means that I render 2000 frames on the farm.
At first, we have to know the summed amount of render power.
With RR, you can just select all online clients in rrControl (or just a subset of special clients) and you get the summed amount of PS at the bottom of the table.
In this example we have 5000 PS for the whole farm.
As there are always two projects, we assume we only get 1/2 of the farm for our project.
5000 PS / 2 = 2500 PS.
Now we can calculate everything:
Total PS*h of the project:
2000 frame x 0.417 h/frame x 200 PS = 166800 PS*h
Total time on the farm:
66680 PS*h / 2500 PS = 66,72h
66,72h /24h = 2,78 days
In most cases companies use a price per Ghz*h. (see help section rrConfig, cost settings )
If you have the costs per hour of your machine, you can calculate the total costs.
In this example we use a price of 0.1 €/Ghz*h
Total Ghz*h of project:
2000 frame x 0.417 h/frame x 30Ghz = 25020 Ghz*h
25020 Ghz*h x 0.1 €/Ghz*h = 2502 €.