Jump to content

Render Machines

Recommended Posts

Perhaps I would suggest to read the whole forum first (at least the first page), there are few choices that you can gravitate towards based on your priorities. (maintenance, performance/value,etc.. )

All possible configurations on latest architecture (from 4 cores Haswell 1155, 6 to 12 cores Haswell-E 2011-3, i7 to 2p Xeon systems) were outlined and discussed to death. It's good to go through that and than ask perhaps more specific questions.

Link to comment
Share on other sites

Hi there,

We have been getting a huge increase of clients recently and we need to improve the speed of our renders from 9-24+ hours per render to ideally 1-2 hours per render.

We have a rough budget of $15-20k.


What machines would you recommend to get?






What machines do you have now that render?

You can calculate a rough aggregate of cores x GHz for all machines contributing to a 9-24Hr render.




Eg. 2x Quad Core i7 @ 3GHz (2 machines x 4 cores e.a. x 3GHz = 24GHz) + 3x 2P Hex Xeon @2.6GHz (3 machines x 12 cores e.a. x 2.6GHz = 93.6GHz) = 117.6GHz* Total.

So, 117Ghz gives me 9Hr per job, I need 2Hr** per job or 9/2 = 4.5x faster -> I would need an aggregate of ~525GHz to achieve that, or the equivalent of ~ 33x 4790 i7 machines (4x4GHz ea).



Computers get "better" every generation = there is roughly a 7-10% performance increase @ same clocks, so if the machines in the 1st example yielding 9Hr/job are 1st gen i7 (5yo or so) I would guestimate that we are 3 gens ahead, i.e. roughly 20-30% faster clock for clock.



So we could actually get 2Hr with 33 - 25% = 25x 4790K machines, or 525GHz - 25% = ~ 400GHz aggregate.



There are more than one ways to get to this kind of a farm. s1150 will be the more complicated, s2011-3 1P, s2011 or 2011-3 2P etc.

You will have to balance hardware, management & software costs go your needs. I don't know if this is doable with less than $20K tho.



*I leave HT threads out, if you are keeping the comparison Intel i7/Xeon to Intel i7/Xeon, its the same

** Many of the operations taking part before the actual raytracing is happening, remain single threaded, so performance is not 100% scalable as the numbers above suggest. Even with an infinite amount of cores in a render farm, you will hit a wall due to the above bottleneck, but lets leave this aside.

Edited by dtolios
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Create New...