OTOY Forums

Hi

We are thinking of setting up a render service for Octane and were wondering if it would interest anyone?

Our idea is to set up a system where you would rent a machine per hour. The machines would have 3 high spec GPU cards in them (12Gb of ram) and you would have full use of one or more. The system would allow you to upload files to the machine then screen share/remote work until you have finished, or your hour/hours run out.

Do you guys think it is a good idea? Would you use it? How much would you be willing to pay for a service like this?

Cheers

Neil

Its a good idea but you should seriously wait for a command line version of Octane Render

This is a little bit soon to start thinking about renderfarms

I think this might be a good idea for people doing large animations or people who want to render super complex / high res poster format images and don't have a GPU with multiple gigs of memory.
It's up to our customers/artists to give feedback on it though, i'm not the right person to advise the commercial prospect.

Radiance

Cheers Radiance, I agree, at present this is just an idea, I really don't want to invest too much if no one is interested.

Sam, I am not sure why we need to wait for a command line version, the whole idea is that it would be visual, seeing as how visual Octane is. I do agree however that it is early stages, hence the reason I am asking now.

Cheers

Neil

Sam, I am not sure why we need to wait for a command line version, the whole idea is that it would be visual, seeing as how visual Octane is

Sorry I didn't read your message correctly. I understand what you mean now
Its a really good idea I think.

I have 100+ boxes at disposal.
Still I would think abit about doing a service, need alot of jobs or they have to cost a decent amount to reflect maintainaince, powerbil, sallery/man hours and so forth. Be that for any kind of service, offline as well. Ive done several jobs for people but only because it was panic and I helped them for free.
Ive not had to go out to look for farms as I have a decent sized one for my type of projects, perhaps someone would like it.
im pritty sure

you should charge 100 to 150 sterling per hour of render time. Thats the going rate. Work out the cost of hardware and when you will reach the break even point.

SO you need around 30 to 40 projects at 3 hours min per animation to make it worth your while. The longer the render the cheaper it is per hour.

Something along those lines.

Look up competition in google. "render farm service"
Add all your costs up, hardware/ staff /electric (special high wattage consumer panel needed with high amperage)
See how many clients needed to break even
HOw much are they willing to pay per hour. 125 sterling per hour is a safe bet.
And how good is the output of the renders.. thats up to you though. (number of bounces/samples etc)

Gonna need some serious hardware for path tracing... it really stresses out the gpus.

50 gpus would do it for commercial use. (about 24 frames per minute) gtx 480x 4 per case with 3gb to start off. Or go for fermi 6gb if you have odddles of cash. (at least its shared ram for bigger projects)

Invest in lots of SSD drives in RAID as well to shift that amount of data.

My 2 cents (or yours after buying all this hardware) :0)

Cubix has (or, what will soon be) the most cost effective means of deploying this type of service, when the software is ready for it. Maximize the number of GPUs per workstation, even using your existing hardware. 19" rackmount versions will be available shortly, and these units would be ideal for what you have in mind.

Eric

Its an interesting idea. Big thing is finding out whether or no there is enough interest to invest in some decent hardware that will lower your hourly costs without needing to raise costs too high to make it unappealing to an average user. Costs overall shouldn't bee too bad as gpgpu is cheaper per core then standard cpu based solutions. Imagine a unit, and I am not even sure this is possible, with 4 pci-e slots, each with a pci-e expander unit attached to them with 2-4 slots in each expander. that would be roughly 8-16 gpu's. figure putting a gtx480 in each one it works out to roughly 3584 - 7168 cores per system costing something around 10-15 thousand dollars per system. Granted these are really rough estimates on both cos and technology, but even with realistic figures the cost benefit is way over that of a standard cpu based render farm, thus enabling you to charge less and still make more with less overhead. Good luck. I'd like to see you do well.

that would be roughly 8-16 gpu's. figure putting a gtx480 in each one it works out to roughly 3584 - 7168 cores per system

He's gonna loose money because the render would be done in like 2 minutes
Its a good idea, and it can get really damn fast with some extenders... Im dreaming already