This post is “From the Vault”, a collection of old writings. Enjoy this throwback!

This is huge. For many years now, I’ve been thinking about how we could leverage parallel processing and cloud computing to do intensive video processing, then send the rendered output back to a dumb terminal for things like high-definition video playback. My dreams have come true: OnLive appears to have perfected the idea in relation to games. Their architecture takes input from a keyboard/mouse/controller, which then arrives at a server. All of the video related to your inputs is then rendered and sent back to the client, all well-within (apparently) acceptable bounds for latency. Effectively, your PC becomes a dumb terminal that just sends controller commands and receives a video stream from a cluster of machines. Instead of playing a game running on your local machine, you’re playing it on a cluster of machines somewhere out on the internet.

The bandwidth requirements aren’t bad; you need a 1.5 megabit connection for 480i resolution (~640×480), and a 5 megabit connection to do 720p (~1280×720). The company also plans to release a set top box with HDMI, Ethernet, a few USB ports, and optical audio so that you can play titles on your television. The content model looks to be subscription-based (everyone seems to love annually recurring revenue these days), but few details are avaialble other than to expect the service in “Q4 2009.” If this works, the console as we know it is effectively obsolete, as is spending large amounts of money to keep a gaming PC up-to-date. We’ll see how it all plays out!

This would be a workable, viable product idea in another decade or so once NVIDIA figured out most of the kinks with GeForce Now (Google would make a good attempt with Stadia that flopped).