Karma w/GPU

   29649   57   4
User Avatar
Member
4189 posts
Joined: 6月 2012
Offline
Thanks for the Solaris presentation and the Karma GPU sneak peak.

First question would be is it planned to be multi-GPU aware?

Thanks!
User Avatar
Member
899 posts
Joined: 2月 2016
Offline
… and AMD GPUs compatible?
User Avatar
Member
147 posts
Joined: 3月 2014
Offline
Just run it on a 32 core AMD
Apprentice Attribute / Houdini 17.0.381 / GTX 970 - driver 411.63
User Avatar
Member
27 posts
Joined: 5月 2015
Offline
RobW
Just run it on a 32 core AMD
I'm pretty sure you are talking about CPU
User Avatar
スタッフ
5202 posts
Joined: 7月 2005
Offline
We try to remain as device-agnostic as possible by using OpenCL. Currently the denoiser is locked to Nvidia GPUs (as it's their denoiser), but we have plans to support other denoisers as well. As RobW mentioned, you should be able to run karma on both a GPU-less farm machine via CPU CL and a local workstation with GPU CL, without any user intervention. And if you're lucky enough to have farm machines with GPUs, you can use the GPU on both.
When the Karma GPU variant is released, of course
User Avatar
Member
146 posts
Joined: 1月 2018
Offline
twod
We try to remain as device-agnostic as possible by using OpenCL. Currently the denoiser is locked to Nvidia GPUs (as it's their denoiser), but we have plans to support other denoisers as well. As RobW mentioned, you should be able to run karma on both a GPU-less farm machine via CPU CL and a local workstation with GPU CL, without any user intervention. And if you're lucky enough to have farm machines with GPUs, you can use the GPU on both.
When the Karma GPU variant is released, of course

Thanks for the detailed answer. I'm very happy you're opting for a device-agnostic path.

Are you planning on releasing any more info regarding features and technical details of Karma before the H18 release? such as feature parity with Mantra on release, missing features, shader development, plug-in architecture mentioned by Scott, rendering technology used (I suppose it's based on the current unidirectional PBR Mantra engine but of course no inside info here)….

And congratulations for the Solaris development, you guys are on fire and amazingly focused on the important.
Edited by jarenas - 2019年7月31日 13:38:11
User Avatar
Member
638 posts
Joined: 7月 2013
Offline
Yeah, the Intel Denoiser is device-agnostic and works alot better than the Optix Denoiser - so, I hope the Intel Denoiser is on the short list for consideration.

I'm hearing that the Renderman Denoiser is “magically good” - but requires alot more AOVs than its competitors.
Houdini Indie
Karma/Redshift 3D
User Avatar
Member
638 posts
Joined: 7月 2013
Offline
jarenas
Are you planning on releasing any more info regarding features and technical details of Karma before the H18 release? such as feature parity with Mantra on release, missing features, shader development, plug-in architecture mentioned by Scott, rendering technology used (I suppose it's based on the current unidirectional PBR Mantra engine but of course no inside info here)….

Mostly all of those types of questions where answered at the end of the Solaris reveal video during the Q&A session.
Houdini Indie
Karma/Redshift 3D
User Avatar
Member
146 posts
Joined: 1月 2018
Offline
Daryl Dunlap
jarenas
Are you planning on releasing any more info regarding features and technical details of Karma before the H18 release? such as feature parity with Mantra on release, missing features, shader development, plug-in architecture mentioned by Scott, rendering technology used (I suppose it's based on the current unidirectional PBR Mantra engine but of course no inside info here)….

Mostly all of those types of questions where answered at the end of the Solaris reveal video during the Q&A session.

I know I know, I mean in a more detailed and extended way
Edited by jarenas - 2019年7月31日 13:43:09
User Avatar
スタッフ
5202 posts
Joined: 7月 2005
Offline
jarenas
Daryl Dunlap
jarenas
Are you planning on releasing any more info regarding features and technical details of Karma before the H18 release? such as feature parity with Mantra on release, missing features, shader development, plug-in architecture mentioned by Scott, rendering technology used (I suppose it's based on the current unidirectional PBR Mantra engine but of course no inside info here)….

Mostly all of those types of questions where answered at the end of the Solaris reveal video during the Q&A session.

I know I know, I mean in a more detailed and extended way

Probably a lot closer to release once things such as procedurals and features are more finalized. It's still in development now, which is where a lot of the vagueness in the details comes from.
User Avatar
Member
146 posts
Joined: 1月 2018
Offline
twod
jarenas
Daryl Dunlap
jarenas
Are you planning on releasing any more info regarding features and technical details of Karma before the H18 release? such as feature parity with Mantra on release, missing features, shader development, plug-in architecture mentioned by Scott, rendering technology used (I suppose it's based on the current unidirectional PBR Mantra engine but of course no inside info here)….

Mostly all of those types of questions where answered at the end of the Solaris reveal video during the Q&A session.

I know I know, I mean in a more detailed and extended way

Probably a lot closer to release once things such as procedurals and features are more finalized. It's still in development now, which is where a lot of the vagueness in the details comes from.
Thanks for the informative answers!
User Avatar
Member
899 posts
Joined: 2月 2016
Offline
thanks for clarifying!

sorry if already addressed, I might have missed this information:
will be hybrid rendering possible like combining karma cpu and karma gpu for the same image?

cheers
User Avatar
Member
4189 posts
Joined: 6月 2012
Offline
I guess I missed if OpenCL Karma will scale across GPUs, without more information I'm guessing it won't as Houdini's OpenCL currently doesn't.
User Avatar
Member
398 posts
Joined: 7月 2005
Offline
twod
We try to remain as device-agnostic as possible by using OpenCL. Currently the denoiser is locked to Nvidia GPUs (as it's their denoiser), but we have plans to support other denoisers as well. As RobW mentioned, you should be able to run karma on both a GPU-less farm machine via CPU CL and a local workstation with GPU CL, without any user intervention. And if you're lucky enough to have farm machines with GPUs, you can use the GPU on both.
When the Karma GPU variant is released, of course

As far as I'm aware AMD has dropped the support of CPU CL. Does it mean that Karma would not work on AMD CPUs? Or maybe the third party OCL implementations like POCL would be a viable solution?
f = conserve . diffuse . advect . add

fx td @ the mill
User Avatar
Member
4 posts
Joined: 8月 2017
Offline
I'm planing to switch from redshift to karma if it supports multiple gpus, just because it's native renderer. Does anyone know more info?
Houdini Indie
Windows 11 Pro / RHEL 9
AMD 3970X - 256GB RAM
User Avatar
Member
83 posts
Joined: 2月 2016
Offline
+1 for Multi GPU support including Nvlink as well as a Hybrid of CPU+GPU ( why leave that multi-core Cpu idling ?)
User Avatar
Member
1 posts
Joined: 10月 2017
Offline
twod
We try to remain as device-agnostic as possible by using OpenCL. Currently the denoiser is locked to Nvidia GPUs (as it's their denoiser), but we have plans to support other denoisers as well. As RobW mentioned, you should be able to run karma on both a GPU-less farm machine via CPU CL and a local workstation with GPU CL, without any user intervention. And if you're lucky enough to have farm machines with GPUs, you can use the GPU on both.
When the Karma GPU variant is released, of course

Very good the device-agnostic using OpenCL but what happens on Mac with Mojave that support only Metal and no more OpenCL?

Is the support on AMD with Metal planned?

Thanx and cheers
User Avatar
Member
4189 posts
Joined: 6月 2012
Offline
OpenCL works on Mac currently but at some point will be removed. That's not good for future proofing the Mac investment.
User Avatar
Member
210 posts
Joined: 12月 2016
Offline
I'm not saying ditch OpenCL and transitioning to Vulcan.
But ditch OpenCL and transitioning to Vulcan.
User Avatar
Member
7 posts
Joined: 7月 2006
Offline
NicTanghe
I'm not saying ditch OpenCL and transitioning to Vulcan.
But ditch OpenCL and transitioning to Vulcan.

Worse case they might have to do a Metal version for the Mac and CUDA/Optix/Vulkan version for Windows. OpenCL is just too slow.
  • Quick Links