Rendering Server -- Input?


I'm the C.T.O. of this studio that does VFX work on movies. We're trying to put together a render server or three that will allow these users to render on the server an not on their desktop (causing them downtime). We do have this program called Smedge3 that allows for network rendering and the use of specified CPU's from each engine on the network, which works OK, but we're looking for more.

Im thinking:

2 TYAN S2927A2NRF-E Dual 1207(F) NVIDIA nForce Professional 3600 ATX Server Motherboard - Retail

4 AMD Opteron 2352 Barcelona 2.1GHz Socket F 75W Quad-Core Server Processor Model OS2352WAL4BGHWOF - Retail

2 Kingston 4GB (2 x 2GB) 240-Pin DDR2 FB-DIMM DDR2 667 (PC2 5300) ECC Fully Buffered Dual Channel Kit Server Memory Model KVR667D2D4F5K2/4G - Retail

4 Dynatron A5MG 77mm 2 Ball CPU Cooler for 4.1" mounting pitch - Retail

2 Western Digital Caviar SE WD1600AAJS 160GB 7200 RPM SATA 3.0Gb/s Hard Drive - OEM

A couple 2U cases with 500Watters

Its pretty cost efficient considering that 1 dell Power Edge server is MORE for 1 than a two server setup from newegg.

Lemme know what you think,
9 answers Last reply
More about rendering server input
  1. Can you run your rendering software in CUDA?

    If you can, screw the multi-cpu and multi server crap. Run a few video cards that will flat out smoke the anything to do your rendering.

    Example of what Im talking about:
    Whether you can use this is unknown to me.
  2. No CUDA.
    You're thinking that Video cards could render hi-res frames better than a CPU with hyper-transport and a 1.8ghz mem controller? I'll pass on that idea.
  3. rotojon1320 said:
    No CUDA.
    You're thinking that Video cards could render hi-res frames better than a CPU with hyper-transport and a 1.8ghz mem controller? I'll pass on that idea.

    Think what you want, but GPU > CPU for rendering. This is something Nvidia actually has their **** together on.

    Its at least 10 times faster than a cpu for ALOT less money. Thats why FASTRA was built on GPUs, not CPUs. Its a classic $5000 machine with GPUs is just as fast as $50000 server.
  4. Agree. If you can use CUDA, a couple graphics cards can outperform many, many CPUs in parallel. The GPU is far more adept at this kind of stuff.
  5. From what I see smedge3 is not rendering itself but allows distributed rendering over the network and has modules for many rendering programs like Maya, 3d Studio Max and a lot of others. You need take a look what programs Is actually used for rendering and check if they can be run in Cuda. Or if they are open source and can be ported to run in Cuda.
  6. Check out Nvidia's Gelato. It is FREE rendering program from Nvidia which runs on GPU in Cuda. It also has Maya plug in. However pro version will only run on Quadro cards. But even then it will be more cost effective then running on CPU. Also don't forget that many Geforce cards can be softmodded to Quadro if You need pro functionality. Anyway check out : . If it is something that could work I seriously recommend considering building around GPU rather than CPU.
  7. This is all very interesting information, thanks
  8. An application must be highly parallel to take full advantage of the GPU(s).
  9. My out put of these renders need to be movie files. This system would also need to support multiple users trying to render simultaneously without delays (other than physical cpu ...delays)
    AMD OR Intel on the CPU rendering?
Ask a new question

Read More

Rendering Servers Components