In 2.75. That’s last year.
As far as I know Nvidia is worse for Linux than AMD. At least for me, AMD GPU never worked in Linux on my old machine but with my current machine I tried really hard - meaning I did pretty much everything the net threw at me, and nothing worked… by which I mean that every time I installed the drivers it either didn’t sense my Nvidia GPU or flat out broke the display. I reinstalled Linux more than 50 times in a few months during that time, and now, with the generic drivers selected the GPU option is not even present in the properties panel.
No, this only means that you are doing something wrong not that Nvida is worse than AMD for Linux … As a matter a fact with Ubuntu 16.04 fglrx is not supported at all yet, so you are stick to open source drivers. And all that because of AMD ‘flawless’ support of Linux drivers.
We can all talk entire day here but facts are that for GPU rendering Nvidia is better choice, numbers are on their side. Even though they are more expensive still users prefer it over AMD for GPU rendering, both users and developers. Nvidia obviously did their job better.
Like I said, I tried everything I could find on the internet, sometimes multiple times, and trials took months.
As a matter a fact with Ubuntu 16.04 fglrx is not supported at all yet
Except that I was not talking about version 16. It was 14 LTS, and some tests on 11-13.
I’m not arguing that Nvidia is worse in general, I spoke of my experience. I went from AMD to Nvidia because of CUDA and it works fine - on Windows. But my experience trying to get it to run on Linux was a nightmare. And like I said, I researched every type of installation and configuration I could find. The one explanation I am left with is that the card issue itself was a byproduct and not worth Nvidia’s time. Doesn’t change the facts.
I’m looking for an update about the current state of AMD Cycles dev. Basically, I’m on the verge of buying Nvidia hardware for rendering work (or maybe even a rack of used Opterons) because I’m finding myself using volumes more and more. I’m assuming there’s a thread somewhere talking about this, so I’m hoping someone can send me to it.
On a more overall view of things, AMD is focused on the gaming industry, so that they don’t have to compete with Intel.
Intel makes high-end CPUs for intense computing like rendering and whatnot, while AMD has found their niche to be gaming.
If I’m not mistaken, but I might be, AMD doesn’t have any high-end, engineering-oriented GPUs like NVIDIA has Tesla. Also, NVIDIA owns mental ray and iray, and they’re clearly in the 3DCG business, while AMD isn’t. There’s a lot more examples like this, NVIDIA does a lot of research for GPU programming and such things, while AMD lags behind, probably purposefully just so they can focus on gaming.
So in a more general sense without getting into the whole OpenCL and CUDA technical debate, that’s why.
But if you wanna taste some crazy-fast GPU rendering, go try LuxRender with their new amazing LuxCore API and OpenCL integrators.
Than they are doing it wrong too. So much focus on gaming and still not much better than Nvidia…What is point than?
I can only see that bigger companies like otoy are capable of diverting enough
resources and time in solving what AMD should already have done.
Maybe AMD should buy OTOYS CUDA cross-compiler?
AMD is already working on a cross-compiler: http://www.amd.com/en-us/press-releases/Pages/boltzmann-initiative-2015nov16.aspx
Will this require additional hardware chip on GPU’s or will be software based ?
Maybe I’m misunderstanding you. But just in case I clarify some things.
OpenCL ‘Kernel Split’ recently has been officially launched with Blender, that’s true. But as I have understood, Cycles always had compatibility with OpenCL. The problem is that AMD/ATI has never been able to compile the big kernel without errors. On the other hand, for years that Intel (CPU) and nVidia (GPU) were able to compile the big OpenCL kernel without errors.