AMD since GCN was ahead of Nvidia in parallel tasks. An example of this was OpenCL. I don't know if this is due to the ACE in the gpu architecture, but since they were designed for parallel processing I think part of the large performance increases over Nvidia in this area could be due to them. I remember a 7970 boost could be on par with the orginal Titan. Just seems to me AMD have looked further ahead with their architecture for the advancement in APIs in parallel processing. It would make sense because they developed Mantle which was capable of taking advantage of this. There's nothing really to compare against Nvidia atm with something similar to the ACE so from the info we have seems AMD has better positioned themselves for the future. We'll see though if devs can really get a grip on the parallel processing these cores provide and get much more out of them.