Robert Oschler of ExtremeTech sat down with some of the higher-ups at NVidia and Mental Images and got the scoop on the previously-announced RealityServer product. He was privy to a live demonstration (the same office-space used previously) and has some great comments on well it works.
Using shared meeting software to pump images from RealityServer to my screen, I saw in real time how a virtual room that existed only in silicon reacted to different combinations of lighting and materials. (See the video below for a demonstration.) The blinds were closed part way, then all the way. A different view of the room was chosen while the time of day was changed with a few mouse clicks, and then another click added a virtual table complete with chairs. After each change, the view changed in seconds to reflect the new choices and the resulting image looked so realistic, that when they showed me the first image at the start of the interview I told them I thought it was a photograph. Without the help of the remote server running RealityServer, each image would have taken my computer hours if not days to render.
Although, I think his article has a few errors (he says that the RealityServer Servers have a minimum 8 GPU’s per server, where I believe that’s a Maximum), it’s a good insight for anyone interested in the product.
via Ubiquitous 3D: Nvidia’s RealityServer. (Linked to the Print version, all on one page).
