Friday, July 20, 2012

Gathering feedback: 5 Major Challenges in Real-Time Rendering (2012)

For SIGGRAPH 2012 now soon in August I'll be doing a new talk about "5 Major Challenges in Real-Time Rendering" in the Beyond Programmable Shading course. 2 years ago I also did a talk on the same topic at SIGGRAPH 2010 so this will be an new version based on where we as an industry are today and list important challenges to try and solve in the next 2-5 years, or even longer in some cases. 

The original 2010 talk about the challenges I talked about then can be found on SlideShare or as a PowerPoint.

For this new talk I want to do it a little bit differently and try and focus even more on the concrete big use cases, why they are important and what needs to be solved to achieve them. Hopefully this will be a good inspiration, motivation and prioritization for future research and development within the different areas of real-time rendering.

Just like last time, I would appreciate any feedback from fellow developers & researchers about what you think are the most important issues and use cases that you would like to see solved real-time rendering in the next 5+ years and why those use cases are important + if you have ideas on how to solve them. 

Gathering broad feedback from people in the industry will be very useful to get a wide understanding of how we as an industry would like to go forward and what problems are the most important to solve. 

If you have any feedback/comments/examples please do post it here and I'll try and include it. Can also message me on twitter (use #rend5 hashtag), post in the Google+ thread or mail me at repi (at)

Thanks in advance for any feedback!



Are you going to stream this event?

repi said...

No, it is at a conference. But SIGGRAPH typically records the talks and puts them up on afterwards

Sam Martin said...

Your 5 from last year are still quite applicable :)

However, I would add that finding more power-economical ways to do graphics is an important challenge. For example, deferred rendering is perhaps not the most power-efficient of architectures we could have produced :)

And, although this is a repeat, I still think that too little emphasis is being put on the art production side, particularly modelling and texturing. ZBrush was about the last major step forward recently, where "recently" was a good number of years ago. I believe we should be seeing a lot more innovation in this area than we are. More efficient artists, and tools that allow them to be more expressive, will both directly result in better graphics.

Cinematic image quality will not go away for some time :)

I'd like programmability to still be on there, particularly a virtual ISA for graphics, primarily for developer productivity reasons and to expose new architectural coding possibilities, but perhaps this involves more politics than innovation. And I would consider the art productivity problem to be more important at this point.