Announcing OctaneRender Cloud Edition

[mbetke]

@pixelrush:
Radiance made his exit stragegy. I guess he even doesn't care anymore with all his money he made and he was so proud of in his last posting.
His last seen online: "Tue May 22, 2012 7:59 am"

He deserved the payout of course and made a good base for Octane.

[kavorka]

For the beta testing in June, is that open or closed?

[MOSFET]

For the beta testing in June, is that open or closed?

The service will initially be available in the US only during the beta period (starting in July) with international support planned for later this year.

[pixelrush]

@mbetke
Well @radiance started out with the best of intentions and really I think it all became too much. Give him some credit for getting Octane going. I think on a personal level he would be happy to see his baby settled and maturing like this.

[Refracty]

Well he has choosen a good time to retire.
I am glad that there is still a good team of developers at Otoy and no Giant has swallowed it

[JuM]

can i build my own gpu farm? working in local network?

[ROUBAL]

Maybe I have missed something, and probably I am going to say something stupid, but I have done some fast calculus and I don't understand the interest of Cloud computing from the distant client point of view :

I explain : I have just tested my internet upload speed, and I get 0.83 Mb per second, which is rather good compared to most users in my area. I must say that it is 4:00 in the night, and that the speed can be much lower at some hours. Added to that, the measured bitrate is the real bitrate, including the control data. So, the bitrate of useful data is actually slower (maybe 20% slower).

My current scene with every required files (ocs, ocm, obj, mtl, textures, scatter files) represents around 1.7 GB = 1700 MB = 13600 Mb. So, in best conditions, I would load my scene on the cloud in 4,55 hours instead of 4 minutes on my machine.

I rarely render more than 4 hours on one image on just two GTX 580.
What is the interest of rendering an image in few seconds on 128 distant GPUs when you have to upload during more than 4 hours ?

Maybe just to save some hours on very large resolutions ?

Uploading during 4 hours also means that the internet connection is busy for 4 hours. So if you have something else important to do in the mean time, you are stuck or in best case you have to share the badwidth with the main loading and increase the transfer time for both tasks...
If you consider that so long transfer can also be aborted at random by unexpected service interruptions (not rare here)...

The power of 128 GPUs is very interesting for an animation if you have them at home, but if each frame requires 4 hours (I don't count texture which could be loaded only one time. They represent only 300 MB on my 1.7GB current scene) of loading through internet, what is the advantage ? I really don't understand who is supposed to need this kind of service.

Is the cloud reserved to softwares with an integrated plugin ?

I perfectly understand the interest of usual render farms (for Blender for example), because the animation is provided in the project file (Blender file) itself and each frame is processed from the provided project file, so the project file is loaded only one time on the render farm.

Animation is already currently a problem for Blender users because of export time for each frame, if added to that there is a transfer time to the cloud of 4.5 hours per frame ... So, animation seems still out of question, even on the cloud. Right ? At least for Blender users ?

[emihich]

Perfect News.
Can't wait to have the chance to use the cloud service for animations.
Perhaps, New updated Octane tutorials , in order to comemorate.
Alembic or FBX import and Export capabilities , for the amazing
Octane Standalone , is still a far away dream, yes ?
And this resumes all my wishes for Octane Render to be really, even more awesome.

[gabrielefx]

When the scene is final there aren't problems to send it via we-transfer, it's fast.
I think that Otoy should authorize many franchising renderfarms all over the world.
I noticed that the Nvidia barebone it's a Tyan server with 8 gpus, nothing special.

Today to buy an 8 Titan barebone costs about 15.000€
The investment is 300.000€ except the power and the office lease were to install the systems.
300.000€ for servers, refrigeration, racks, switch and wiring.

I think it's a good business
I remember that Mental Images developed 2 years ago a similar technology.

[kavorka]

My current scene with every required files (ocs, ocm, obj, mtl, textures, scatter files) represents around 1.7 GB = 1700 MB = 13600 Mb. So, in best conditions, I would load my scene on the cloud in 4,55 hours instead of 4 minutes on my machine.

For me, my scenes are soo much smaller, but sometimes still require 30 min - 1 hour to render. Some of the interiors take 3-4 hours or more depending. For me, I could see transferring the files over in a couple of min. This will not only get my image back faster, but allow me to start working on my next project.

For animation, hopefully we get a file structure that will hold the animation data so you only have to transfer it once. we will see.
Also, there is the possibility for you to work in the cloud, where all of your data is held on their servers and you are accessing it remotely, therefore no internet transfer times. That service will probably be more expensive, so might not be something you would want to use.

What kind of internet connection do you have? I just upgraded to verizon FIOS and downloaded a 2 GB file in about 10 min last night. I didn't test upload speeds yet though.