Previous | Next --- Slide 6 of 81
Back to Lecture Thumbnails
taz

On average, a single GPU equals the rendering performance of 5 to 20 high-end CPUs. This allows artists to quickly iterate on their work without having to wait hours to see the effect of a tweak, thus having a positive effect on creativity and the final result.

Source: http://www.cgsociety.org/news/article/2180/five-reasons-to-adopt-gpu-rendering-in-2016

vasua

Modern high end computers tend to use some combination of CPUs and GPUs. However, at some point in the past, a few companies tried to push a third type of chip, a PPU (Physics Processing Unit), for high throughput physics computations. Unfortunately, a number of factors, including limited developer support, lead to the death of these chips in favor of something like NVIDIA PhysX, which is now integrated directly into GPUs. I wonder if, as we tend towards more and more physically accurate computations for video games / simulations, if something like this might make a comeback.

https://en.wikipedia.org/wiki/Physics_processing_unit

1_1

Why is there an additional component for graphics? There are algorithms suited for drawing pictures by taking a 3D description of the scene and simulating how a real camera would see it. GPUs were originally designed to do such rendering.

Brandon

@vasua I feel that VR might make a great case for this as true life simulation didn't need to be computed on the fly before but now with users interacting with things in unpredictable ways in VR they expect a high level of realism and computation - take a look at NVIDIA's VR fun house - tons of physically based particle simulation leads to a fun time. Though I'm not sure what the reason for a separation between gpu and ppu would be