We are still assessing the impact of the acceleration on the end date of the HCC1 project. Our rough estimate with 'known' work based on this acceleration is about 100-130 days until end of project. BUT my use of the word 'known' simply means the number of batches we are aware to run. What we are are is not equal to what the researchers want to run so we are discussing this further with them at this time.
However, we do not yet have any research project beyond HCC1 that we know can run on a GPU. While this has definitely shown the power of GPU working well, for an application that clearly fits the profile of being suitable for GPU, the researchers are still the ones that provide the code and they will have to determine if GPU would work for their project and then make the modifications.
It is important to us to have GPU capable projects, but we cannot promise them. Even then, we could find ourselves with a project that was written for CUDA and not be able to run on AMD graphics cards (or written for AMD but not CUDA). As a result, we ask that people not make purchasing decision without understand these caveats.