SonK

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.

SESI is currently not in a easy situation for the development for the mac system. it is not only METAL it is now also ARM, that means all code must be portable cpu and gpu wise for the Karma render and Houdini.

Depending on the promise of ARM, it might be worth it in the long run. In about 30+ years of being involved in technology in some way or another, I have learned that the landscape is constantly shifting, and what might be undesirable at some point can become an absolute must-have at another point. The proof will ultimately be in the speed of the new CPU's which will determine just how appealing these new machines will be to companies looking to squeeze every last ounce of simulation and rendering juice from their hardware.

I will say that Apple is pretty aggressive in providing assistance and even programmers to facilitate adapting code to the OSX platform. They're doing that with Redshift, OTOY, and even Blender at the moment, so it wouldn't surprise me if they are actively collaborating with SideFX right now.

Lastly, don't underestimate the power of modern compilers. A lot of people seem to have the mistaken idea that simultaneously supporting Windows, Linux and OSX require huge amounts of resources and manpower, but in most cases that is no longer necessary as compilers will do most of the work automatically.

Midphase
I will say that Apple is pretty aggressive in providing assistance and even programmers to facilitate adapting code to the OSX platform. They're doing that with Redshift, OTOY, and even Blender at the moment, so it wouldn't surprise me if they are actively collaborating with SideFX right now.

Lastly, don't underestimate the power of modern compilers. A lot of people seem to have the mistaken idea that simultaneously supporting Windows, Linux and OSX require huge amounts of resources and manpower, but in most cases that is no longer necessary as compilers will do most of the work automatically.

History shows it is NOT a easy task:
Otoy
2016: https://home.otoy.com/octanerender-3-and-roadmap-update/ [home.otoy.com]
2018: http://www.cgchannel.com/2018/03/otoy-reveals-its-roadmap-for-octanerender-2018-and-2019/ [www.cgchannel.com]
“In Otoy’s early OctaneBench benchmark tests, the Vulkan and Metal backends are already as fast as CUDA, even when running on Nvidia GPUs.”

Unity
port to Windows: https://blogs.unity3d.com/2009/03/20/unity-25-for-mac-and-windows-now-available/ [blogs.unity3d.com]
port to Linux: https://blogs.unity3d.com/2015/07/01/the-state-of-unity-on-linux/ [blogs.unity3d.com]

as i said SESI and others are not in a easy situation.
the next 2 years will show what is happening.

Somewhat related: https://www.macrumors.com/2020/07/11/metal-developer-tools-windows/ [www.macrumors.com]

Midphase
Depending on the promise of ARM, it might be worth it in the long run. In about 30+ years of being involved in technology in some way or another, I have learned that the landscape is constantly shifting, and what might be undesirable at some point can become an absolute must-have at another point. The proof will ultimately be in the speed of the new CPU's which will determine just how appealing these new machines will be to companies looking to squeeze every last ounce of simulation and rendering juice from their hardware.

I will say that Apple is pretty aggressive in providing assistance and even programmers to facilitate adapting code to the OSX platform. They're doing that with Redshift, OTOY, and even Blender at the moment, so it wouldn't surprise me if they are actively collaborating with SideFX right now.

Lastly, don't underestimate the power of modern compilers. A lot of people seem to have the mistaken idea that simultaneously supporting Windows, Linux and OSX require huge amounts of resources and manpower, but in most cases that is no longer necessary as compilers will do most of the work automatically.

All that porting work for an ever shrinking platform which on top is heavily overpriced. In the end the porting very likely could cost more money than it may return…

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

Will the soon to be released 4000series Ryzen/Threadripper CPUs even be viable for Karma CPU rendering? Without AMD CPU CL support my impression is that Karma wouldn't run at all. Am I wrong or would Intel CPUs be the only alternative then? Not my preferred choice with Intel's current performance and pricing.

With most professional 3D artists having a decent GPU which will surpass the OpenCL processing power of a CPU this is/should be a non issue. As I've understood OpenCL isn't supported on AMD CPUs for years now anyway.

Thank you for your input, much appreciated. It doesn't answer my question though, if Karma CPU runs without problems on Ryzen/Threadripper.
Since we are specifically discussing Karma and not RS/Octane a maybe higher performing Karma GPU CL implementation down the road is currently irrelevant as only Karma CPU is available. And my feeling is that Karma GPU is still far away.

W.r.t to GPU render engines I'm not in the “they are the holy grail of rendering” camp. GPU renderers have admittedly made big improvements in the last years. And while I like the speedy feedback of RS, when it comes to rendering smoke/explosions, vdb's, caustics it can't match Mantra's or Arnold's quality. In addition, when scenes become really big RS's speed advantage gets nullified.

Hey E_Kay,
I am currently making some test with a threadripper 1950x, ad for now I don't have any issues while using Karma. Here are two render test:

Thanks ioness! Exactly the info I was looking for.

Any updates on this Karma GPU render?
I was really looking forward to this, maybe on H19?

I would dial back your anticipation. Karma is not a high prio, 3rd party renderers are.
My prediction is Karma will indeed be a great tool, but that is 2yrs away from full feature set and
competitive render speeds. Remember also, that there is no solid standpoint from sesi regarding is Karma
would be free, or a paid renderer.

What are GPUs? You mean Crypto Mining Processors?

tinyhawkus
Remember also, that there is no solid standpoint from sesi regarding is Karma
would be free, or a paid renderer.

Whoa, really? Going paid with it would be a disaster.

BrianHanke
Going paid with it would be a disaster.

Since they haven't said anything yet, no need to panic.

Also, I do hope y'all filled out your yearly questionnaires where these exact things were asked

I didn't get a questionnaire. If it had a section for "Should Karma be paid?" with a checkbox for "disaster" I would have been all over it.

BrianHanke
I didn't get a questionnaire.

I'm referring to the "Houdini Annual Survey". The latest edition of which closed in Sept. 2020.
Usually commercial customers should get it automatically without doing anything?

protozoan

BrianHanke
I didn't get a questionnaire.

I'm referring to the "Houdini Annual Survey". The latest edition of which closed in Sept. 2020.
Usually commercial customers should get it automatically without doing anything?

Hmm, pretty sure I didn't get it. I just have Indie so maybe that's why.

Indie is Commercial (Paid license, can be used for Commercial Projects, and includes support).

I got the Survey as well.

@protozoan but there was indeed talk initially upon announcement of Karma existing.
From memory, it was that "we are unsure if it will be a free product." Not quoting verbatim mind you.

A Disaster if it were not free? really? Have a look around at the bundled renderer that comes with a lot
of applications over the years. Toys, absolute toys. Mantra was always a labour of love, and it's deep integration
is something to still be marveled at. If Karma were to be as performant as a 3rd party engine, then why would
it not be possibly a purchased product?

Happy for it to be either, but saying it would be a disaster if it weren't free is a bit silly.



L