Quote Originally Posted by T_Flight View Post
Multiple dies on the same package is an entirely different matter, and one I support and is the reason I will never support this "double cheeseburger" arrangement. I wouldn't mind if it has 80-100 dies, but it better be under one IHS to put one cooler or block on top of. I wouldn't care if the package was 4 inches square.
I think that multiple separate gpus (whether they are on separate cards or the same one) will be around for a while - even after you get multiple dies under the same IHS. It's for the same reason that multi cpu still exist even though we have quad cores (well not the only reason). There is always someone that wants or needs more and an easy way to get that is to put together multiple of whatever is the best currently available.

With dual GPU's side by side all you get is double the heat, double the electric bill, but nowhere near double the performance. On top of that you get the microstutters. You double the price of your cooling, it puts more stress on the boards. It's a bad idea that should've been ditched after the first one of those blunders was unloaded on the market.
This is the real crux from my perspective. Software support needs to be better. Games and applications access the GPU through APIs. The GPU manufacturers need to make their drivers (or hardware) in a way that handles spreading the load across multiple gpus transparent to the API. Having to make a special profile for each game isn't going to work in the long run - it just puts more strain on the driver development team. Instead the GPUs (or drivers) need to have logic that will allow them to share memory and share the workload on a single frame: thus simultaneously ridding us of AFR and the doubled up memory.