Hello, my first post ere (although I was silently observing this forums for a while - creepy, I know)
I am currently working on new Render Node spec for a small office I am working for.
We currently own 3 years old i7 workstation (will post exact spec tomorrow morning, when I'm in the office) which renders a final frame in 3ds + Vray ~4-6 hours. As for modelling and postproduction, this spec is enough most of the time. It's slightly too slow for previewing, which tends to be very time consuming.
We mainly work on architectural visualisation we make for ourselves.
Since the workstation is enough to work with, but not enough to render, we are considering expanding the set with a new render node. My first shot was creating something similliar to Boxx Render PRO1. Something like this:
Single Xeon E51650v3 3.5 GHz, 15 MB cache (6core)
16GB DDR4-2133 SODIMM ECC
no graphic card/integrated
240GB SSD SATA
set up on our own though - I don't really believe in BOXX service working well in Poland, and price can go down something around 50% down BOXX vs. similliar custom build. Our budget varies, we aim in possible rendering a single final frame in under one hour time (preferably 15 mins ). Atm we are aiming at something between 2000$-3000$ but we can invest more if we can expect surge in working conditions for our graphic dude
I hope this illustrates a situation.
Now - I am no computer guru, we are about to consult with local service. But I would like to get some feedback from people who actually work in architecture and have some experience with hardware.
1./ Main issue I am having is - is it worth it to invest in Render node and leave Workstation significantly slower then Node or should we rather change old unit into Node and buy a new workstation (then we will upgrade the node eventually, when we think its worth it)? My logic is - if WS is enough for our workflow in every aspect except rendering, then its worth it to buy a node and save money (GPU, HDDs, optical drives, power(?) etc. ) and delegate all rendering tasks from WS to it (including bigger previews).
2./ It is important to us that this solution is scalable. Meaning, I would like to be able to replace processors, add GPU, eventually add more nodes of similliar type to the whole system later (in what case is it justified to set up a rack server instead of standalone units?)
3./ What local network solution would you recommend? Do we actually need separate LAN/disk space just for this purpose? What bandwidth can jobs of this scale require to work seamlessly?
We currently run a Windows Server on 2 mirrored RAIDs. Local network is built on LAN: WAN --- router --- switch --- clients (including server). This seves not only for 3ds but also for other CAD applications we're working with etc.
4. / Is it worth investing in Dual Core MoBo in price range 2000-3000$?
Any contribution will be welcome, hope I don't ask for too much information The idea is just forming its shape and I am really not sure, were it is going to lead us