More Development Activity Ticking Up Around Vulkan For Blender

[stormcrow]

Blender *LOVES* wasting resources in proprietary APIs and letting open source ones lag behind. Nothing new.

BS. No really. This is BS. Blender started out as a product targeting Linux for an in-house graphics development company. They eventually released it to the public. The wider public bought out the license in its entirety and released it as open source. Blender targets where its users are and the APIs that let them do so regardless of the licensing. I'm sorry but the AMD/Intel GPGPU isn't where the users are and for damned good reason in this case. Its users are mostly using Nvidia hardware which uses CUDA and the reason that's the case is because CUDA got their first, works well, easy to install, maintain and write for compared to AMD & Intel and generally has better raw performance for GPGPU use (especially compared to Intel, ARC cards not withstanding). Open source OpenCL libraries remind me of trying to use rabbit ears on TVs back in the day. Tilt your head wrong, breathe too hard, or some obnoxiously annoying reason and you'd suddenly lose reception. I've spent hours trying to figure out how to get a program to work on various OpenMP or OpenCL libraries reported to 'work' with some package only to give up. I never had such a problem with CUDA on an officially supported distribution which is why Nvidia GPUs dominate that market. Open source alternatives will only be widely used if they're useful. Most people couldn't give two shits for license purity if the software or hardware won't do what they need it for.

Luckily for me, I no longer need GPGPU support for anything I need to do, so the point is moot for me at this point. I can use AMD GPUs at home without caring about what state ROC/OpenCL support is in because all I do is use it for gaming. But if I needed proper GPGPU compute support I wouldn't be using either AMD nor Intel GPUs because Nvidia still has the better platform for that.

[Danielsan]

Somewhere they explained that Vulkan is too much low level for their needs, and that is more suited to make drivers rather than building rendering engines...

[mirmirmir]

BS. No really. This is BS. Blender started out as a product targeting Linux for an in-house graphics development company. They eventually released it to the public. The wider public bought out the license in its entirety and released it as open source. Blender targets where its users are and the APIs that let them do so regardless of the licensing. I'm sorry but the AMD/Intel GPGPU isn't where the users are and for damned good reason in this case. Its users are mostly using Nvidia hardware which uses CUDA and the reason that's the case is because CUDA got their first, works well, easy to install, maintain and write for compared to AMD & Intel and generally has better raw performance for GPGPU use (especially compared to Intel, ARC cards not withstanding). Open source OpenCL libraries remind me of trying to use rabbit ears on TVs back in the day. Tilt your head wrong, breathe too hard, or some obnoxiously annoying reason and you'd suddenly lose reception. I've spent hours trying to figure out how to get a program to work on various OpenMP or OpenCL libraries reported to 'work' with some package only to give up. I never had such a problem with CUDA on an officially supported distribution which is why Nvidia GPUs dominate that market. Open source alternatives will only be widely used if they're useful. Most people couldn't give two shits for license purity if the software or hardware won't do what they need it for.

Luckily for me, I no longer need GPGPU support for anything I need to do, so the point is moot for me at this point. I can use AMD GPUs at home without caring about what state ROC/OpenCL support is in because all I do is use it for gaming. But if I needed proper GPGPU compute support I wouldn't be using either AMD nor Intel GPUs because Nvidia still has the better platform for that.

This is bullshit, using the latest and the greatest, you can use and amd hip just fine. It was different several months ago, but hey.

Also, what's cuda? Cool kids use optix now.

[Quackdoc]

its a step in the right direction, AMD hip is a bloody joke, the only viable compute platform for AMD that anyone cares about is OCL and vulkan, and OCL is only recently viable things to rusticl. so any work towards vulkan is good work.

[luno]

Cool!
Imagine one day where we'll have both Vulkan and Wayland support and possibly 10bit+ colors and HDR support too!

problem is, there will always next big thing

[qarium]

AFAK, Blender has been on OpenGL since the very first release. And since Blender is licensed under GPLv2 and is multi-OS, I don't even think it's legally possible for them to ship proprietary Direct3D code.
Regarding "aged" code, in the production industry base things don't move too fast (nor should they). Autodesk's Maya, for instance, is basically the same program today as it was in 2008, and this is not an exaggeration - the program was already really mature all those years ago, and "if it ain't broke, don't fix it". Some big studios today are still on Maya 2012 and they're 100% fine with it.
IMO it's better that Blender stuck with "aged" OpenGL for all these years, while being stable, than not as many devs giving Vulkan the attention it deserves and bricking the program "just because" some people can't live with the latest and greatest.

to me it looks like those people on old 3dstudiomax and Maya will never upgrade and instead migrate to Blender...

OpenGL itself is death but there is maybe NextGL based on Zinc means high level openGL like language with modern features on the basis of zinc means the gpu driver only need to support vulkan.

also for the people who claim vulkan is to low-level it looks like WebGPU is a little more high-level than vulkan maybe these people target WebGPU instead of vulkan.

I find both options very good and elegant means NextGL based on Zinc and WebGPU. also WebGPU is supported by MacOS and iOS...

but one is clear to me no one will upgrade their 3D studio max and Maya if instead they can jump into Blender instead.

[qarium]

BS. No really. This is BS. Blender started out as a product targeting Linux for an in-house graphics development company. They eventually released it to the public. The wider public bought out the license in its entirety and released it as open source. Blender targets where its users are and the APIs that let them do so regardless of the licensing. I'm sorry but the AMD/Intel GPGPU isn't where the users are and for damned good reason in this case. Its users are mostly using Nvidia hardware which uses CUDA and the reason that's the case is because CUDA got their first, works well, easy to install, maintain and write for compared to AMD & Intel and generally has better raw performance for GPGPU use (especially compared to Intel, ARC cards not withstanding). Open source OpenCL libraries remind me of trying to use rabbit ears on TVs back in the day. Tilt your head wrong, breathe too hard, or some obnoxiously annoying reason and you'd suddenly lose reception. I've spent hours trying to figure out how to get a program to work on various OpenMP or OpenCL libraries reported to 'work' with some package only to give up. I never had such a problem with CUDA on an officially supported distribution which is why Nvidia GPUs dominate that market. Open source alternatives will only be widely used if they're useful. Most people couldn't give two shits for license purity if the software or hardware won't do what they need it for.
Luckily for me, I no longer need GPGPU support for anything I need to do, so the point is moot for me at this point. I can use AMD GPUs at home without caring about what state ROC/OpenCL support is in because all I do is use it for gaming. But if I needed proper GPGPU compute support I wouldn't be using either AMD nor Intel GPUs because Nvidia still has the better platform for that.

sorry i have to tell you that you are wrong... because people if we talk about Blender no longer use CUDA they use OptiX on Nvidia hardware.

also the professionals in my point of view tent to avoid nvidia gpus because of the lag of the amount of VRAM because if you compare the CUDA result to modern CPUs means 64 core EPIC and ryzen 7000 cpus and the upcoming ZEN4c cpus it looks like modern CPUs are at a similar performance level with the difference of 128-256GB ram is the new normal on even low-end systems. (i do not talk about optiX here because only CUDA has the mathematical same result than CPU rendering)
so why do professionals avoid Nvidia ? thats because even a RTX 4090 has only 24GB VRAM and even if you buy Super expensive professional version it is only 48GB VRAM.... and compared to the CPU solution 128-256GB ram is better.

and about support it is right that CUDA has superior support out of the box experience but if we talk about Blender even my 2017 dated Vega64 was Blender 3.3 support...

in my point of view if you dislike closed source you just buy a Threadripper system like i have and do Blender with the CPU with 128-256gb ram...

and i want also say something about the future if you see what a radeon rx 7900XTX is with 24GB VRAM

" AMD Radeon RX 7900 XTX - 24GB GDDR6 - Desktop
Chip Navi 31 XTX "RDNA 3", 96CU, 533mm²
Fertigung TSMC 5nm + TSMC 6nm
Chiptakt keine Angabe, Boost: keine Angabe
Speicher 24GB GDDR6, 2500MHz, 20Gbps (20000MHz effektiv), 384bit, 960GB/s
Shader-Einheiten/TMUs/ROPs 6144/384/192
TDP/TGP 355W (AMD)
Externe Stromversorgung 3x 8-Pin PCIe
Kühlung 3x Axial-Lüfter (110mm)
Gesamthöhe Quad-Slot (3.63 Slots)
Abmessungen 352.9x158.2x72.6mm
Besonderheiten Echtzeit-Raytracing, AMD Infinity Cache (96MB), HDCP 2.3, AMD FreeSync, AMD TrueAudio Next, AMD Eyefinity, AV1 Decode, 0dB-Zero-Fan-Modus, Backplate, LED-Beleuchtung (RGB)
"

i am pretty sure with 61 TFLOPs this 7900XTX for 999 dollars will be a very very very good option for Blender.

with the same limitations like the Nvidia cards have real professionals avoid it because 24GB vram is not what they need they need 256GB ram to render their complex szences.

[qarium]

Somewhere they explained that Vulkan is too much low level for their needs, and that is more suited to make drivers rather than building rendering engines...

NextGL based on Zinc does adress this problem...

also WenGPU is more high level than Vulkan maybe they should target WebGPU instead.

[qarium]

This is bullshit, using the latest and the greatest, you can use and amd hip just fine. It was different several months ago, but hey.
Also, what's cuda? Cool kids use optix now.

right this man is just outdated all he talks was true 6 month ago or even older.
these people are like old book did not get an upgrade to their database.

sadly with ROCm/HIP Blender support the same CUDA people claim OptiX is so much better.

but i am sure 1 TFLOPs 7900XTX with 24GB VRAM for 999 dollars will give Nvidia a hard time i am sure about this.

isn't it true that AMD has an OptiX alternative for the 7000 series RDNA3 cards ?

[qarium]

its a step in the right direction, AMD hip is a bloody joke, the only viable compute platform for AMD that anyone cares about is OCL and vulkan, and OCL is only recently viable things to rusticl. so any work towards vulkan is good work.

its the law what is the problem AMD could easily make a CUDA compatibility on the binary level but this is agaist the law.

ROCm/HIP is the next best legal think make it compatible on the source-code level.

on Blender OpenCL is death... and maybe Vulkan will not make it as soon as they Discover that NextGL based on Zinc and WebGPU is higher level and fits better to them because vulkan is to low-level...