This is a repost of a previous idea (granular, slider-based hardware selection) as suggested by Moisey:
"..However, we do believe that there are different workloads that require different servers and hardware. For example caching services need a lot of RAM but not much else, DBs need very fast harddrives and CPU, and then there are cases when you just want a bunch of storage and not much else.
This is an area that we are exploring to see if we can begin provisioning different hardware types sometime in 2014 and testing to see if customers like the ability to select a different type of VM with different combinations of underlying specs based on their workload.
If anyone wants to startup a separate uservoice request for that we then upvote from there we can provide regular updates on those projects..."
This would be useful when building out a heterogenous system, and especially in cases (like hardware-accelerated decryption) where it might involve specialized hardware. This would also lay the groundwork for other specialized cases such as HPC (with GPUs) and mining.
I'll start it off with my +3 votes...