Originally posted by: Rollo
Originally posted by: Amplifier
SLI 7 series *drool*
But...but....AMR gives you the ability to match your comparatively slow X800 card to a fast R520 card, and thereby slow down the R520card to X800 AMR level! And it has the long defunct
dongle!
Isnt the master/slave thing supposed to solve that problem? I know that when you run a parallel program on multiple machines of different speed, if you implement load balancing the slower one will not pull down the faster one with it. Unless Ati is really dumb, I'm sure they would have thought of that.
Also, another thing to worry about is the different feature sets of the cards, like mixing a sm3 card with a sm2 one. In this case, they would have a problem if they combined them in the traditional SLI way, but since AMR doesnt have to run on identical cards (supposedly, hence the master/slave model) you could have the newer card doing all the shaders with sm3, and the older one doing other tasks, such as vertex transformations, lighting calculations, and whetever else that's not dependent on sm3. Granted, such a setup will require an advanced driver to handle the situations, and it might just be too costly to implement, but it's not impossible.
Bottom line, if SLI could run 2 different cards at the same time, it would be a more viable option, since you're not limiting yourself to one generation of cards. Whether or not AMR can do this effectively remains to be seen.