Well, this might be interesting for NVidia GPU,couse they might perform even better than AMD.(we have to wait for sources if there will be some)
This could make system stronger, there is still no need of being afraid(increase is not so dramatical).
But it is still interesting if we will increace memory capacity to 0.5mb per core, this should make cpu faster, but they were some one saying that gpu wont suffer from it more than cpu, so it is still interesting to see the sources,because gpu memory even waving more bandwidth is still much higher latency than cpus cache, so it is interesting what is the way the kernel is made.