OTOY Forums

Goldorak wrote:

RobSteady wrote:

Goldorak wrote:Yes. Zero GPU means the GPU can be on another box over LAN, with no render GPU needed on the local machine running Max/Maya.

But you can do this already with any remote control software.
What's the difference in using TeamViewer for example? I can work perfectly from home with my iPad controlling my workstation in the office with Octane.
Or control another pc here in the office from my computer that runs Octane.

I don't get this feature. Is this the big secret thing everyone was working on so hard the last months?
Was this requested even once by any user? Is there any urgent need for this?

Not the same - you are taking about a live video stream of your desktop. But on your desktop there still needs to be a GPU to run Octane + Maya. But let's say you have MacPro with no CUDA GPUs (until we port CUDA to Metal). OE allows he Maya plugin to see any remote CUDA device over the local network and virtualize it efficiently as a stream so you can run Octane API entirely over a network, with the same efficiency as the old way using the c library plugin w/ local GPU

We have had more requests than any other to solve Apple dropping NVIDIA support on MacPro and newer machines.

I thought the cross-compiler that was talked about earlier was the solution for Macs, allowing a MacPro to render using their local AMD GPUs. Is that what you mean by "until we port CUDA to Metal"? Is OctaneEngine basically the name of the cross-compiler? Is is a stop-gap until the cross-compiler is finished? Is the cross-compiler just a subset of OctaneEngine's features? Are they two totally separate things? It's unclear.

As a Mac user who's had to move to PC for 3ds Max, I can certainly imagine OTOY has gotten a number of requests for AMD support on the MacPro. But is streaming to an NVIDIA GPU on the local network (which would need to sit in a PC that you already have) really the solution Mac users have been clamoring for? I would think Mac users want native, local, rendering ability on their AMD GPU without needing a LAN PC + NVIDIA GPU. I didn't see that specifically mentioned, which is odd if the number one request is for new MacPro support.

riggles wrote:

Goldorak wrote:
Not the same - you are taking about a live video stream of your desktop. But on your desktop there still needs to be a GPU to run Octane + Maya. But let's say you have MacPro with no CUDA GPUs (until we port CUDA to Metal). OE allows he Maya plugin to see any remote CUDA device over the local network and virtualize it efficiently as a stream so you can run Octane API entirely over a network, with the same efficiency as the old way using the c library plugin w/ local GPU

We have had more requests than any other to solve Apple dropping NVIDIA support on MacPro and newer machines.

I thought the cross-compiler that was talked about earlier was the solution for Macs, allowing a MacPro to render using their local AMD GPUs. Is that what you mean by "until we port CUDA to Metal"? Is OctaneEngine basically the name of the cross-compiler? Is is a stop-gap until the cross-compiler is finished? Is the cross-compiler just a subset of OctaneEngine's features? Are they two totally separate things? It's unclear.

As a Mac user who's had to move to PC for 3ds Max, I can certainly imagine OTOY has gotten a number of requests for AMD support on the MacPro. But is streaming to an NVIDIA GPU on the local network (which would need to sit in a PC that you already have) really the solution Mac users have been clamoring for? I would think Mac users want native, local, rendering ability on their AMD GPU without needing a LAN PC + NVIDIA GPU. I didn't see that specifically mentioned, which is odd if the number one request is for new MacPro support.

If OSX supported OpenCL (i.e. not the DOA 1.2 version they are now killing off), we'd be done already. The CUDA cross compiler works great since it compiles to AMD IL using the low level APIs exposed on Windows and Linux. But Apple does not allow that API on OSX.

Apple is very clearly forcing us all to Metal for any GPGPU code going forward on Apple's platforms. And Metal is not final on OSX (GPGPU is still pretty broken). So we have to wait for Metal to be finalized in OSX by Apple and then we can cross compile CUDA to it.

It's a mess. I know.

So, in the meantime, we have two options :

1) The 3.1 CPU fallback - which will be pretty slow, but at least will get Octane running on your MacPro.
2) The remote CUDA device feature with OctaneEngine (which could also support AMD devices once that backend is done).

will there be no Beta testing for the 3DS Max and Maya plugin of 3.0 with us the community? are you going to release a fully functional version?

So that's it? The big big big secret?? I was thinking about more features for the 3ds max plugin~ like a damn uv map channel would be so nice!(btw this is still a very basic feature that missing all the time!)
Also I've noticed that in that 77 pages you guys said that would be a 10$ per month with 1000+ mins rendering pts ,which will get 800 OB pts,this is exactly 20 K520 grid pts~ Does it means I can get a 800 OB pts GPU render farm for 1000 mins in a month? with 10$? and can pay month by month? no extra cost? and it can only get less than 20 gpu units right? what if I want more units say 40 GPUs ? Does it means I can only render a scene(or one camera in scene) by using ONLY 20 GPUs cuz standalone not allow 20+ more? That's a lot of confuse here and there...
Also I would like to know how much does it cost to upgrade the max plugin to 3.x ? Does it cost 49$ for the standalone and no cost for the plugin?(like before?)or I should pay more money for the plugin? If that so, I personally will choose not to upgrade and switch to c4d or other gpu engine like fsrender cuz max plugin sucks, sorry about that... better fix and add some of the damn baisc missing features instead of making a non-gpu pc workflow cuz we DO KNOW the hardware setup

Goldorak wrote:If OSX supported OpenCL (i.e. not the DOA 1.2 version they are now killing off), we'd be done already. The CUDA cross compiler works great since it compiles to AMD IL using the low level APIs exposed on Windows and Linux. But Apple does not allow that API on OSX.

Apple is very clearly forcing us all to Metal for any GPGPU code going forward on Apple's platforms. And Metal is not final on OSX (GPGPU is still pretty broken). So we have to wait for Metal to be finalized in OSX by Apple and then we can cross compile CUDA to it.

It's a mess. I know.

So, in the meantime, we have two options :

1) The 3.1 CPU fallback - which will be pretty slow, but at least will get Octane running on your MacPro.
2) The remote CUDA device feature with OctaneEngine (which could also support AMD devices once that backend is done).

Thanks for the clarification. Ah yes, I forgot about Metal being Apple's OpenCL layer replacement. Aside from my criticisms of the roadmaps and initiatives of OTOY, I realize I can't blame the dev team for the lack of results pursuing OpenCL (which is something many of us asked you to pursue) or the moving target of Apple in general.

c4d wrote:So that's it? The big big big secret?? I was thinking about more features for the 3ds max plugin~ like a damn uv map channel would be so nice!(btw this is still a very basic feature that missing all the time!)
Also I've noticed that in that 77 pages you guys said that would be a 10$ per month with 1000+ mins rendering pts ,which will get 800 OB pts,this is exactly 20 K520 grid pts~ Does it means I can get a 800 OB pts GPU render farm for 1000 mins in a month? with 10$? and can pay month by month? no extra cost? and it can only get less than 20 gpu units right? what if I want more units say 40 GPUs ? Does it means I can only render a scene(or one camera in scene) by using ONLY 20 GPUs cuz standalone not allow 20+ more? That's a lot of confuse here and there...
Also I would like to know how much does it cost to upgrade the max plugin to 3.x ? Does it cost 49$ for the standalone and no cost for the plugin?(like before?)or I should pay more money for the plugin? If that so, I personally will choose not to upgrade and switch to c4d or other gpu engine like fsrender cuz max plugin sucks, sorry about that... better fix and add some of the damn baisc missing features instead of making a non-gpu pc workflow cuz we DO KNOW the hardware setup

See my other post on ORC pricing and how it works. The short answer is yes, you can buy more render power - e.g up to 100x was enabled for high end customers (~80,000 OB/hr) during beta to get a one month local GPU job finished on ORC overnight.

The refreshed MAX plug-in will be entering beta shortly. We're planning to announce the plug-in upgrade path before May 15th. We're finalizing this now.

It's been a lot of time since I don't post in here, and only am doing to point a couple of things out.
I am totally with Oguzbir here.
We don't need more live stuff and I couldn't care less about mac users and the apple gpu aproach.
Our studio needs more robust features in the max plugin to rely on. as oguzbir said. A better more deep integration with max's materials and maps architecture too. A real LWF that actually works. a better more controllable exporter to octane standalone.
Stop the unnecesary feature development. ORC is nice but not update price nice.
I miss the freshness that octane brought to the table 6 years ago.

ttaberna wrote:It's been a lot of time since I don't post in here, and only am doing to point a couple of things out.
I am totally with Oguzbir here.
We don't need more live stuff and I couldn't care less about mac users and the apple gpu aproach.
Our studio needs more robust features in the max plugin to rely on. as oguzbir said. A better more deep integration with max's materials and maps architecture too. A real LWF that actually works. a better more controllable exporter to octane standalone.
Stop the unnecesary feature development. ORC is nice but not update price nice.
I miss the freshness that octane brought to the table 6 years ago.

We are working on getting more resources on Max development, just as we recently did with Maya.

Why people want to use Macintosh computers to make gpu renders?
The latest iMac versions are ok only for Adobe CC, Mac Pro are use useless also with Blackmagic Da Vinci or Fusion.
These two famous programs work better on WIndows with NVidia gpus.
Latest Macs don't have thunderbolt 3 ports and full functional and certified boxes to fit 4 Titans X
Also OpenCL doesn't work properly as Cuda.
Develop 3D render kernels for OpenCL is a waste of time except Otoy will do miracles....

c4d wrote:...like a damn uv map channel would be so nice!(btw this is still a very basic feature that missing all the time!

It has been implemented many versions ago, update your plugin/check the manual.