LuxCore Open-Source Renderer v2.4 Released With CUDA Support, Better Windows Scaling

[phoronix]

Phoronix: LuxCore Open-Source Renderer v2.4 Released With CUDA Support, Better Windows Scaling

Debuting this weekend was LuxCoreRender 2.4, the newest version of this impressive open-source physically based renderer...

http://www.phoronix.com/scan.php?page=news_item&px=LuxCore-Renderer-2.4

[Imout0]

A first step into dropping OpenCL altogether and it is no wonder considering the absolutely pathetic job The Khronos Group and specially AMD have been doing promoting OpenCL.

[Auzy]

A first step into dropping OpenCL altogether and it is no wonder considering the absolutely pathetic job The Khronos Group and specially AMD have been doing promoting OpenCL.

Huh? You do realise CUDA is NVIDIA only (totally proprietary right?). Not sure why anyone other than NVIDIA (and Apple) would want to drop OpenCL support. Apple depreciated OpenCL for their proprietary Metal API, and that was probably a monopoly thing (to make it harder for developers to develop cross platform apps to PC/Android).

OpenCL is supported by AMD, Nvidia, Intel and many other Embedded GPU's. Cuda, only NVIDIA. Metal only Apple..

Also, developers shouldn't care about how well an API is promoted.. They care about results. The only advantage of CUDA, is apparently the speed is better (however, this could simply be attributed to Nvidia's compiler optimising it better, as it only needs to compile for 1 platform anyway).

As an enduser, widespread CUDA adoption = vendor lock-in. It's not good for us.. Also, this announcement also stated that OpenCL had some improvements too. Not sure how you managed to interpret it as a sign that OpenCL is getting nuked..

[Imout0]

Huh? You do realise CUDA is NVIDIA only (totally proprietary right?). Not sure why anyone other than NVIDIA (and Apple) would want to drop OpenCL support. Apple depreciated OpenCL for their proprietary Metal API, and that was probably a monopoly thing (to make it harder for developers to develop cross platform apps to PC/Android).

OpenCL is supported by AMD, Nvidia, Intel and many other Embedded GPU's. Cuda, only NVIDIA. Metal only Apple..

Also, developers shouldn't care about how well an API is promoted.. They care about results. The only advantage of CUDA, is apparently the speed is better (however, this could simply be attributed to Nvidia's compiler optimising it better, as it only needs to compile for 1 platform anyway).

As an enduser, widespread CUDA adoption = vendor lock-in. It's not good for us.. Also, this announcement also stated that OpenCL had some improvements too. Not sure how you managed to interpret it as a sign that OpenCL is getting nuked..

Of course it is proprietary and it is heaps and bounds better which is objectively what anyone working in the 3D design field wants. So you are on the market for a graphics card and between Nvidia and AMD you gotta be nuts to pick AMD because OpenCL sucks so hard that even Luxcore, which was a huge OpenCL endorser and the only place where AMD had huge wins, is shifting to CUDA. Even its own raytracer, Radeon ProRender, works better on Nvidia cards. Blender Cycles had no OpenCL support at all when they implemented GPU acceleration and it never worked as well as CUDA once it got baked in back in the day. Today it is totally a no show for AMD. And don't get me started with the vast majority of 3D design suites, OpenCL is missing in most of them. All of this is the byproduct of AMD not contributing nor promoting OpenCL nearly enough because it was the pretty close to the only party interested in it getting traction.

Developers care about results and CUDA is more widely adopted, better performing and supported than OpenCL. So yeah, it is an egg and chicken situation. Most of the market is Nvidia so most of the acceleration is coded for CUDA. Maybe if Intel launched a somewhat competitive GPU things would change for OpenCL but with AMD as its main target it is pretty much dead.

I've been a Blenderhead for quite a while and Luxrender developers were adamant about it being only OpenCL and the latter being mostly ignored even by AMD (check their last OpenCL driver date) that is fragmenting it with its ROCm chimera. Maybe we will have another chance with Vulkan, but OpenCL si effectively dead.

[starshipeleven]

Huh? You do realise CUDA is NVIDIA only (totally proprietary right?). Not sure why anyone other than NVIDIA (and Apple) would want to drop OpenCL support.

The real question isn't who would want to drop OpenCL, but who would want to add support for it.

[Auzy]

Of course it is proprietary and it is heaps and bounds better which is objectively what anyone working in the 3D design field wants. So you are on the market for a graphics card and between Nvidia and AMD you gotta be nuts to pick AMD because OpenCL sucks so hard that even Luxcore, which was a huge OpenCL endorser and the only place where AMD had huge wins, is shifting to CUDA. Even its own raytracer, Radeon ProRender, works better on Nvidia cards. Blender Cycles had no OpenCL support at all when they implemented GPU acceleration and it never worked as well as CUDA once it got baked in back in the day. Today it is totally a no show for AMD. And don't get me started with the vast majority of 3D design suites, OpenCL is missing in most of them. All of this is the byproduct of AMD not contributing nor promoting OpenCL nearly enough because it was the pretty close to the only party interested in it getting traction.

Developers care about results and CUDA is more widely adopted, better performing and supported than OpenCL. So yeah, it is an egg and chicken situation. Most of the market is Nvidia so most of the acceleration is coded for CUDA. Maybe if Intel launched a somewhat competitive GPU things would change for OpenCL but with AMD as its main target it is pretty much dead.

I've been a Blenderhead for quite a while and Luxrender developers were adamant about it being only OpenCL and the latter being mostly ignored even by AMD (check their last OpenCL driver date) that is fragmenting it with its ROCm chimera. Maybe we will have another chance with Vulkan, but OpenCL si effectively dead.

Adding support != shifting... You're leaping a few assumptions here... Just cause a product adds osx support for instance doesn't mean they're shifting to osx..

Also, don't forget that both the new xbox and ps5 use amd. New rdna2 cards will be released for pc soon. AMD was behind for a while, but don't be surprised if they bounce back (everyone thought AMD CPUs were dead in the past, and then they released the athlon 64). AMD are incredibly resilient..

The market changes quickly. Amd just needs something either really fast, or fast+cheap. And the extra income from consoles will really push them along I feel

[starshipeleven]

Also, don't forget that both the new xbox and ps5 use amd.

As were the current gen consoles. Neither is relevant for OpenCL.

New rdna2 cards will be released for pc soon.

That's just hardware. The main issue of OpenCL is that the software ecosystem around it sucks and has sucked for a long while.

The market changes quickly.

Not for the type of applications that do computing.

Amd just needs something either really fast, or fast+cheap.

And overhauling OpenCL software ecosystem.

[zeealpal]

Firstly: It's great to see a engine I used back in the LuxRender 0.8 days before it was reborn as LuxCoreRender. The project has gone through enormous ongoing change and improvements that keep it relevant for a number of users and use cases.

The whole move to compiling the engine once for OpenCL/CUDA (which has dramatically improved workflow) and using a material/texture runtime interpretation came about when adding CUDA support.

Next up RTX support is already in v2.5 code and Optix denoising will be available for viewport denoising as well.

It will be interesting to see how AMD's hardware raytracing is used, through Vulkan? OpenCL? If it's OpenCL I can see it being added to LuxCore quickly, otherwise another compute platform would have to be added.

To note: I believe the addition of CUDA did allow for easier addition of other compute backends, so adding Metal or something new from AMD would not be impossible.

[Imout0]

Adding support != shifting... You're leaping a few assumptions here... Just cause a product adds osx support for instance doesn't mean they're shifting to osx..

Once 99.9% of the market belongs to Nvidia because AMD can't compete I wonder what would be the point of developing it for OpenCL. I assume you haven't seen how current RTX cards steamroll AMD cards on all ray tracers. That steamrolling comes from three sides: hardware (14nm vs 7nm no less), software and engines. AMD can't compete in any of the three because as I have said, Nvidia is currently better than AMD even at its own renderer.

Also, don't forget that both the new xbox and ps5 use amd.

Yeah, as starshipeleven said, it helped it a lot having the current ones too. Sheesh.

New rdna2 cards will be released for pc soon. AMD was behind for a while, but don't be surprised if they bounce back (everyone thought AMD CPUs were dead in the past, and then they released the athlon 64). AMD are incredibly resilient..

I will be thoroughly surprised if AMD can beat Nvidia at anything. Right now they can't even fix their drivers and RDNA gave them quite the negative reputation. Then again, the CPU situation isn't so much because of AMD management but Intel mismanagement. Nvidia won't give AMD the slightest chance.

The analogy would be the Bulldozer era, when you had to be out of your mind to buy an AMD chip. Right now buying an AMD graphics card for raytracing rendering is beyond moronic.

The market changes quickly. Amd just needs something either really fast, or fast+cheap. And the extra income from consoles will really push them along I feel

AMD does not. They took years to enable OpenCL for Cycles and it was pretty much broken and missing most features. When it reached feature parity and AMD seemed like an alternative Nvidia crushed them again with OptiX.