Originally posted by SunnyG.
What are the bottlenecks? Can you elaborate?
The memory cell architecture itself, controller clock speeds and number of bus lanes, the amount of crosstalk between individual channels, Thermal throttling*...It is better than IDE or SATA but there is still a lot going on with the PCIe interface.
* E.g Computer graphics cards equipped with GDDR6 can obtain a theoretical maximum of 128gb/s - however with a TDP [Thermal Design power] of over 350W [ for overclocked cards] all that heat has to be dumped somewhere otherwise the card will underclock itself, limiting it's performance.