Can someone try to explain to a normie CPU render engines, why for example, were considered to be more suitable / able in achieving photo-realistic results and why XPU capable engines like Karma and Renderman have some features that are only available in CPU-only mode?

Why would the GPU not be able to handle certain tasks reliably / efficiently / at all?

have found an article that might answer your question CPU vs. GPU [www.cgdirector.com].

Another one (shorter) [academyofanimatedart.com]

XPU renderers are relatively new while CPU renderers have been around for decades. Leveraging GPU's effectively is also typically quite a bit more complicated than CPU's. Utilizing both CPU and GPU at the same time is harder still. I guess it's a bit like hybrid vehicles. According to the internet the first hybrid car was created in 1899 [www.carsdirect.com] yet hybrids only become competitive fairly recently.

Hello this is Gulshan Negi
Well, depending on the project's specific requirements, CPU-based or GPU-based rendering engines should be considered. For photo-realistic 3D graphics, CPU-based rendering engines are better suited than GPU-based rendering engines for video game real-time rendering. However, XPU-capable engines may offer features that are only available in CPU-only mode for some rendering tasks that require more complex calculations that are better suited for the CPU. Understanding the qualities and shortcomings of each kind of rendering engine can help in choosing the right device for a specific undertaking.
I hope it will help you.
Thanks

GPU render engines are relatively new so they have to fight their way into big studios. GPUs are very expensive, use a lot of power, are memory constrained and much less versatile than CPUs (farms also do a lot of other stuff than just rendering) . In a big production it doesn't matter all that much I find. As long as renders finish over night and don't crash or waste time on weird problems it's fine. At home or in smaller studios it's a different thing and it's easier to get a fast gpu and do it all on there. Because of these things perhaps development of code that runs on gpu is taking less priority.

GPU's be design are inherently flawed. The code you write for a GPU is not the same as CPU, it's a pain, period.
Below is a link to a very succinct explanation why GPU's will/have hit walls that they simply cannot get past.


GPU reality [news.ycombinator.com]

The primary reason for attempting to get exactly the same pixels with CPU, GPU or CPU+GPU is so that the strengths and weaknesses of both can be utilized where it makes sense. Artists workstations usually have a GPU available whereas renderfarms less so. Artists usually need interactive renders for cheaper scenes e.g asset look dev or lighting blocking. Higher quality full complexity renders are normally done over night because it doesn’t make sense to sit there watching it. The greater the complexity the more difficult it is for GPU, but that is somewhat aligned with interactive tasks needing to be faster and less complex anyway.

lewis_T
GPU's be design are inherently flawed. The code you write for a GPU is not the same as CPU, it's a pain, period.
Below is a link to a very succinct explanation why GPU's will/have hit walls that they simply cannot get past.


GPU reality [news.ycombinator.com]

thank you for posting this link to that great explanation! do you happen to know other resources that further explain things like this potentially aimed at beginners for such topics? im guessing this is all in the field of electrical/chemical engineering? thanks again for the link!

That's about as basic an overview as you can get.
As you shrink things down, and increase speed, signal noise becomes a huge issue. Electrical engineering
text books would shed some light on this, but it's not a simple topic.


L

Thank you very much! It is much appreciated!