OTOY Forums

Hey everyone,
I am sure this has been covered somewhere but after looking for some time now I just thought it may be easier to just ask. Currently my set up is 4 GTX titan blacks but looking to expand and need some info for the most optimal set up.
I was thinking about getting a case from http://www.cubixgpu.com/ and getting either an additional 4 titan black set up similar to the set up that fuse animation uses. With 2.0 coming out though I was wondering if there was a better option of if loading the GPUs on to one box is ideal? Last question I have would be in regards to how 2.0 handles distribution of work to multi machines. I some times need to work with larger engineering files so when rendering someones updating the scene takes longer then the actual render time... With 2.0 is it possible to have different machines work on different frames or will it all work on the same frame to before moving to the next? I know this is a lot of questions but any guidance would be very much appreciated!

Thanks,
Marc

hi, Marc.

Don't know where are You from, but one Guy in the forum from France was selling (maybe still is) an expansion box - one of netstor (http://www.netstor.com.tw/) models & three Titans =)

If You are going to do rendering using new network rendering feature I guess it's like doing one animation frame by frame..so You don't split loading time at all as same rig, main computer should loud & then send data over network.

The best way if You worry about loading times would be to build second maschine, anyway You need second set of licences (even for network rendering =) and if needed render the same scene, give first half of sequence for one rig, while second to other..- no distribute rendering simply divide the work for two machines(in this case You divide loading times for each mashine).

Alternativelly having two rigs You'll have a 'backup plan' if something would happen to one system, You'll still be able to work.

Hey thanks for the reply, this helps me a lot understand how things will work and allow me to plan accordingly. Making a 2 rig system seems pretty nice like you mentioned with the splitting of the project if needed as well as if something were to go wrong I could always use the one machine till I get the other back online. Also thank for sending the link about that set up from France if you could find out if they are still trying to sell it and how much they are wanting for it that would be great.

Thanks!
Marc

Drop a line for this Guy, maybe He's still selling his stuff:
http://render.otoy.com/forum/viewtopic. ... 93#p186393
talked with him earlier & I have to say fairly good price =)

I nearly bought his setup, but he was undecided and by the time he'd made his mind up I'd bought my own Neststor. I'll probably populate it with 780Tis cards, as they're way cheaper than Titans and just as fast [EDIT: they're actually faster!]. Also, with the distributed memory coming in CUDA, on-board memory isn't such an issue.

I have Octane Render 2.0, next we'll have Ahmet's plug-in, and in a few weeks my Netstor will turn up. Then I can start throwing GPUs in there - cannot wait!

Weezer wrote: Also, with the distributed memory coming in CUDA, on-board memory isn't such an issue.

I wouldn't be so sure about this feature & it's significance..memory is memory & if You don't have some proceses migh be imposible to do - like big resolutions =) 780 is slover than Ti, but for 6model 780 You would pay 600$ - that's 100 per 1Gb - hard to ignore =) compared to Titan it's a bargain =)

& hey, congrats with nice purchase =)