SPI is the fastest bus on AVR. For Arduino the max speed is 8 MHz and it is programmed I/O so the The best speed to an SD is about 820 microseconds to write a 512 byte block. I don't see how your shield will improve much on this. The current SdFat can do this.
Other modern micro-controllers have fine SDIO controllers so a shield makes no sense.
The problem is not so much hardware but how to handle file systems on SD flash. This is what I am working on.
There are at least two type of apps that one can optimize. One is data logging where latency is the problem so you don't lose data. The other is where the average rate is important. These require different approaches.
What are you trying to accomplish? What is your application?