NVIDIA Begins Funding Blender Development

[tildearrow]

That's a bit of an extreme case, but regardless, if they don't do anything to hinder AMD's performance, how is what you said a bad thing? Micro-optimizations (even if platform-specific) are welcome as far as I'm concerned, and this is coming from someone who has no intention on buying Nvidia hardware in the foreseeable future.
Nothing is stopping AMD from doing the same thing. If Nvidia spends the money and resources to improve the performance on their end (which again, does not suggest crippling competitors), I see nothing wrong with that.

It may not cripple the competition, but it may affect it.

[schmidtbag]

It does not finish there. This in turn will make NVIDIA look apparently faster than AMD, hurting AMD's business by reducing their demand.

That's a bit of an extreme case, but regardless, if they don't do anything to hinder AMD's performance, how is what you said a bad thing? Micro-optimizations (even if platform-specific) are welcome as far as I'm concerned, and this is coming from someone who has no intention on buying Nvidia hardware in the foreseeable future.
Nothing is stopping AMD from doing the same thing. If Nvidia spends the money and resources to improve the performance on their end (which again, does not suggest crippling competitors), I see nothing wrong with that.

[tildearrow]

People here are so negative. Worst-case scenario, Blender only gets better CUDA support. How is that a bad thing?

It does not finish there. This in turn will make NVIDIA look apparently faster than AMD, hurting AMD's business by reducing their demand.

[tului]

Absolutely, but, not everyone here can be described as cautious.
If Blender didn't have OpenCL support and Nvidia stepped in, that would be a reason to be cautious, because from what I've seen, most projects that support CUDA first don't bother with OpenCL.


Who said anything about cheering? Why does everything have to be so black and white? Blender already supports both OpenCL and CUDA; it's not like Nvidia is going to cram more proprietary stuff into Blender that is dysfunctional without their hardware. The gloom that people like you propose is unwarranted. Not everything Nvidia does is proprietary.
The newest piece of Nvidia hardware I own as a GT 630 and I have no intentions on buying an Nvidia GPU for myself in a long while, so obviously I'm not praising their involvement in Blender, but you don't hear me fearing or whining about it either.

So much drama over nothing.


You are naive if you seriously think Nvidia is going to somehow push out their competition in Blender. Name one realistic method where they could accomplish this, that does not go against Blender's license.

When was the last time Intel meddled with an open-source project where they inserted proprietary software that catered only to them?

I assume he meant their overall business strategies. Nvidia somehow forking over enough cash that despite both major consoles being AMD that games get their proprietary gameworks tripe. Intel is well known for both enticing and bullying OEMs et al to keep AMD out.

[schmidtbag]

No you are right. However it is always good to be cautious.

Absolutely, but, not everyone here can be described as cautious.
If Blender didn't have OpenCL support and Nvidia stepped in, that would be a reason to be cautious, because from what I've seen, most projects that support CUDA first don't bother with OpenCL.

Long story short; it promotes the use of proprietary APIs. Many (most?) people using Linux and open source systems are against proprietary protocols at all costs. And with very good reason, as we see over the years where it leads. So excuse some people (myself included) if we don't exactly cheer when a vigorously proprietary company attempts to stronghold its way into the market and minimize the use of free, open standards that all people can benefit from (aka, OpenCL).

Who said anything about cheering? Why does everything have to be so black and white? Blender already supports both OpenCL and CUDA; it's not like Nvidia is going to cram more proprietary stuff into Blender that is dysfunctional without their hardware. The gloom that people like you propose is unwarranted. Not everything Nvidia does is proprietary.
The newest piece of Nvidia hardware I own as a GT 630 and I have no intentions on buying an Nvidia GPU for myself in a long while, so obviously I'm not praising their involvement in Blender, but you don't hear me fearing or whining about it either.

So much drama over nothing.

To suggest that they'll somehow magically stop doing this is -- at least in my view -- naïve.

You are naive if you seriously think Nvidia is going to somehow push out their competition in Blender. Name one realistic method where they could accomplish this, that does not go against Blender's license.

Whether or not AMD does this is not entirely clear to me. I wouldn't be surprised if they did, though. And intel has a public record of doing this, so it's certainly not like this MO is unique to NVidia.

When was the last time Intel meddled with an open-source project where they inserted proprietary software that catered only to them?

[kpedersen]

People here are so negative. Worst-case scenario, Blender only gets better CUDA support.

No you are right. However it is always good to be cautious.

Besides the real enemy to Blender is the "Blender Cloud". I predict this is what is going to ultimately ruin it; not additional dependencies like CUDA.

[ermo]

One can hope that once Vulkan moves forward with its adoption/inclusion of OpenCL(-ish features), that particular barrier can more easily be scaled? Right now, the story around getting full-fat OpenCL drivers working w/AMD's FLOSS drivers seems ... convoluted?

Watching Red Hat's Dave Airlie explain the current state of Compute it seems clear to me that OpenCL simply isn't at the right level stack-wise to compete with CUDA.

The only obvious solution (read: targets both CUDA and AMD's HIP) seems to be hipSYCL which -- like CUDA and HIP -- is C++ single source and apparently strives to use OpenMP for CPU workloads and OpenCL/CUDA for GPU workloads.

Since my knowledge in this area is clearly shallow and superficial, I'm going to stop short of suggesting that it might be more beneficial to Blender to work on hipSYCL support instead of targeting closed source NVidia CUDA and RTX functionality.

[sa666666]

People here are so negative. Worst-case scenario, Blender only gets better CUDA support. How is that a bad thing?

Long story short; it promotes the use of proprietary APIs. Many (most?) people using Linux and open source systems are against proprietary protocols at all costs. And with very good reason, as we see over the years where it leads. So excuse some people (myself included) if we don't exactly cheer when a vigorously proprietary company attempts to stronghold its way into the market and minimize the use of free, open standards that all people can benefit from (aka, OpenCL).

[ermo]

There is so little of a chance Nvidia will be causing problems.

It's NVidia. Their whole business model is predicated on throwing their weight around, causing both problems to their competitors AND gaining NVidia any entrenched advantage possible.

To suggest that they'll somehow magically stop doing this is -- at least in my view -- naïve.

Whether or not AMD does this is not entirely clear to me. I wouldn't be surprised if they did, though. And intel has a public record of doing this, so it's certainly not like this MO is unique to NVidia.

[schmidtbag]

People here are so negative. Worst-case scenario, Blender only gets better CUDA support. How is that a bad thing? If you bought Nvidia hardware, you'll be directly benefiting from the money they put in. If you didn't buy their hardware, you won't see any differences at all. Sound perfectly reasonable to me.
Best-case scenario, Blender as a whole is improved, maybe even getting a nice kickstart on Vulkan, which everyone will benefit from.
There is so little of a chance Nvidia will be causing problems.