During the GPU Technology Conference, a presentation was made by Morgan McGuire, associate professor of computer science at Williams College, and soon an Nvidia research scientist. McGuire spoke on what the future held for graphics.
McGuire spoke about the need to be able to process 100,000 megapixels per second, which is up a considerable amount from the 450 megapixels per second that modern graphics systems are currently capable of. The aim is to approach the perceptual capabilities of the human eye, to make things as real as possible. “We’re about five or six magnitudes of order [between] modern VR systems and what we want,” McGuire said during the talk, “That’s not an incremental increase.”
There is a known trade-off between latency and image quality. Reducing latency often results in lower image quality, so a new approach will be needed to maximise image quality while still maintaining the kinds of low latency needed for virtual reality (VR) content. Nvidia are working on a number of approaches to this problem, such as removing the Path Tracer rendering tool and instead using a combination of rasterisation and GPU power.
Interestingly, McGuire also spoke on the subject of text within the VR environment. McGuire explained that as VR devices become gateways to other content and a headset becomes a default for computer use, the keyboard would no longer be a practical interface device, so the problem of entering text becomes something companies involved with VR will need to consider for the future.
“Text is actually the killer app,” McGuire said.
The changing face of the desktop PC, with the requirements for running high-quality VR experiences, will likely mean that the prices of those rigs will increase, McGuire acknowledged, though he believes that the increase will not be as great as some anticipate and that consumers and the market will find a balance point.
VRFocus will continue to bring you news on Nvidia’s VR projects and research.