OTOY Forums

Hi All...

So now i wanna start using all the PC i can in my studio...currently using ver 2 - 1.9 of octane for max...
right now I plan to use my msi laptop with 980m (8gb) as the master to 2 or 3 Desktop on the same network..one of them has 4gtx780 (6gb) the other has one gtx780 (3gb) and possibly a third one aslo has a different board.

1) is it necessary to have the same octane for max plugin on the slaves? or just the same std alone suffices?

2) I know when u have 2 or 3 different cards on the same pc, octane will recognize the card with the least RAM, but what happens if the card with the least RAM is on another (slave) computer? is this also an issue or will it take only the cuda cores from that Card's pc and will recognize the 8GB from my laptop?

3) If rendering animation..the frames rendered by the slave get stored in the master or in that particular slave?

thanks

Hi,
1) you so not need plugin on the slaves. On the slaves you need standalone which has same Octane version as your master (plugin or standalone).
2) have not tried that... But it seems logical to me that the gpu's on your master are running not influenced by gpu's on
slaves, and that gpu with less ram on slave only influence usability of gpu's on that slave (when scene bigger then lowest ram)
3) all rendered images are saved on slaves. The slaves do load the scene, and add renderspeed on master.
Cheers,

Edited: mistaken: all rendered imges are saved on the master (not slaves)!

Hi Rappet,

thanks alot for your reply...

nildoe wrote:Hi All...

So now i wanna start using all the PC i can in my studio...currently using ver 2 - 1.9 of octane for max...
right now I plan to use my msi laptop with 980m (8gb) as the master to 2 or 3 Desktop on the same network..one of them has 4gtx780 (6gb) the other has one gtx780 (3gb) and possibly a third one aslo has a different board.

1) is it necessary to have the same octane for max plugin on the slaves? or just the same std alone suffices?

2) I know when u have 2 or 3 different cards on the same pc, octane will recognize the card with the least RAM, but what happens if the card with the least RAM is on another (slave) computer? is this also an issue or will it take only the cuda cores from that Card's pc and will recognize the 8GB from my laptop?

3) If rendering animation..the frames rendered by the slave get stored in the master or in that particular slave?

thanks

Hi,

1) Standalone licenses on the slave machines will suffice. You would only want to install plugin licenses on the slaves if you intended on using 3ds Max native Network Rendering - it would be far more cost-effective to stick with just to standalone licenses though.

2) The purpose of the VRAM limitation that you described, is to ensure that each GPU can fit it's own copy of the scene data. If a GPU cannot fit the scene data in it's VRAM, it cannot be used in the rendering - this applies to network rendering as well as rendering entirely on the local machine. Typically, the GPU with the lowest VRAM will be on the Master machine (also commonplace, all GPUs are the same model). You would be best advised not to install your weaker GPUs on any of the slaves, as Octane cannot automatically enforce the VRAM limitation across a network the way it can on the local machine.

3) That is correct. Each slave will store the results of the rendering workload supplied to them from master.


Hopefully this has clarified things better

azen wrote:...

2) The purpose of the VRAM limitation that you described, is to ensure that each GPU can fit it's own copy of the scene data. If a GPU cannot fit the scene data in it's VRAM, it cannot be used in the rendering - this applies to network rendering as well as rendering entirely on the local machine. Typically, the GPU with the lowest VRAM will be on the Master machine (also commonplace, all GPUs are the same model). You would be best advised not to install your weaker GPUs on any of the slaves, as Octane cannot automatically enforce the VRAM limitation across a network the way it can on the local machine.
...
Hopefully this has clarified things better

Hi Azen,

I cannot follow you and I would like to comment that I prefer having my strongest gpu's on the master and weaker (read lower vram) on the slaves.
That seems more logical, because I want to be sure that my biggest scene can be rendered on the master for sure. It would be a waste of renderpower to have my 12gb Titan X installed on a slave when needing to render a large scene. If the weaker gpu's are on the slaves (the way I have) and they cannot handle the big scene they just do not participate and/or I can disable the slave(s) that cannot handle the scene.
And before rendering final images, slaves are not always needed to participate, and then the strongest renderpower is preferable on the master.
If the weaker gpu (i.e. 3gb) would be on the master I Could not even load a large scenes on the master to begin with, and therfor not on the slaves.
Sorry, but I think you are mistaken here.

Cheers,

thanks for the reply Azen...but yes i tend to agree with rappet..seems more logical to have master with the strongest gpu.

rappet wrote:

azen wrote:...

2) The purpose of the VRAM limitation that you described, is to ensure that each GPU can fit it's own copy of the scene data. If a GPU cannot fit the scene data in it's VRAM, it cannot be used in the rendering - this applies to network rendering as well as rendering entirely on the local machine. Typically, the GPU with the lowest VRAM will be on the Master machine (also commonplace, all GPUs are the same model). You would be best advised not to install your weaker GPUs on any of the slaves, as Octane cannot automatically enforce the VRAM limitation across a network the way it can on the local machine.
...
Hopefully this has clarified things better

Hi Azen,

I cannot follow you and I would like to comment that I prefer having my strongest gpu's on the master and weaker (read lower vram) on the slaves.
That seems more logical, because I want to be sure that my biggest scene can be rendered on the master for sure. It would be a waste of renderpower to have my 12gb Titan X installed on a slave when needing to render a large scene. If the weaker gpu's are on the slaves (the way I have) and they cannot handle the big scene they just do not participate and/or I can disable the slave(s) that cannot handle the scene.
And before rendering final images, slaves are not always needed to participate, and then the strongest renderpower is preferable on the master.
If the weaker gpu (i.e. 3gb) would be on the master I Could not even load a large scenes on the master to begin with, and therfor not on the slaves.
Sorry, but I think you are mistaken here.

Cheers,

Hi,

That is a very good point. I was approaching it from the standpoint of wanting to ensure that all slaves could handle the scene (rather than having to cut out a slave due to one low card). If there is a considerable disparity between the largest and smallest VRAM value, then you would want to take that approach. Ideally, you would have the same model of GPU installed across your network render setup, for better stability (assuming that you are building a network rendering setup from scratch rather than making use of what you happen to have on hand). Apologies for any confusion caused, I was indeed mistaken in my thinking.

Dear Friends!

Are you sure images are saved on slaves in network rendering?
When I render using 3ds Max with two slaves, images are saved ONLY at master, since it's the only machine with 3ds Max.

It's different situation when I render with 3 machines via Backburner, then on each one 3ds Max is launched, and then images are saved on particular machines.

Am I misunderstanding something?

kcpr-raffaEl wrote:Dear Friends!

Are you sure images are saved on slaves in network rendering?
When I render using 3ds Max with two slaves, images are saved ONLY at master, since it's the only machine with 3ds Max.

It's different situation when I render with 3 machines via Backburner, then on each one 3ds Max is launched, and then images are saved on particular machines.

Am I misunderstanding something?

Oops... Got a brainlse there:
Indeed: Edited: mistaken: all rendered imges are saved on the master (not slaves)!