Bij Computer Games Online is een interessant interview te vinden met Nympha Lee, product manager chick van ATi voor de ATi Rage Fury MAXX. Hieronder vast een paar vraagies over hun SLI design:
CG: The idea of using two chips at once to improve performance was popularized back with the Voodoo2 from 3dfx. Now they're doing it again in a single-card solution with the Voodoo5. Your multi-chip technology works in a different way. Could you explain how it differs?
Lee: In the SLI technique, one chip would draw all the even lines and one would draw all the odd lines. The key with that technique is that both of the chips work on the same frame, and both would process the same triangle setup on the one frame. ATI knew of that technique and felt that it was inefficient in that both of the chips would have to wait for the other one to finish it's stuff before moving on to the next frame. With our technique, Alternate Frame Rendering, one chip renders even frames and one renders odd frames. So we've taken the same concept but instead of working on the same frame, we've separated it so the chips don't have to wait for the other chip to move on to the next frame. Each chip is responsible for it's own set of frames.
CG: What about latency? Is it possible that one chip could get too far ahead of the other one, causing a potential split-second difference in the time between input is made and the frame that is created from that input actually shows up on screen?
Lee: I'll have to admit that the latency concern is something that our competitor has brought up to sort of discredit the whole technique. But in talking with our engineers on it, the whole latency level of the Rage Fury MAXX is the same in double buffer as with a single-chip doing triple buffering. It's really not much of a concernólatency isn't ever even brought up as a concern when talking about a single chip doing triple buffering. We're building a white paper that will explain it to everyone.