AMR Tiling

[gerbiaNem]

Might some of the websites be wrong with ATI's tiling AMR method? Could both cards always have the same section tiled and just send out the signal to a converter on the outside of the case that sets the tiles together? That would make the SLI bridge useless and save bandwidth. Could this be done? And would it make the video ram requirement halved?

 

[Hate_Bot]

ATI ha sbeen using this AMR for a few years, for military uses and sucj

 

[HighTest]

Might some of the websites be wrong with ATI's tiling AMR method? Could both cards always have the same section tiled and just send out the signal to a converter on the outside of the case that sets the tiles together? That would make the SLI bridge useless and save bandwidth. Could this be done? And would it make the video ram requirement halved?

Don't think that there is anything "wrong" with doing AMR, it's just a different approach than SLI.

Since it's the comparison of the textures and framebuffers that is happening over the PCI-E bus in AMR and across the "Bridge" in SLI, an external device connected to the monitor connections would be useless for reducing "bandwidth" impacts. This way the "tiles" stich seamlessly without having artifacts and other errors that would occurr without both cards knowing exactly what the other is doing.

The only debate is if the PCI-E bus won't be throttled with the amount of data going back and forth at 1600 res screens with all details and full AA and AF running. nVIDIA stated they needed the bridge to ensure sufficient bandwidth and think it gives them an edge over ATI's approach.

Until "shipping" product is benchmarked and compared it's all just hype and speculation, but my opinion is that there isn't anything "wrong" with AMR's reported information to date.

 

[Erasmus354]

a cynic would say nvidia uses the bridge so they could charge even more money for their SLi solutions....