I have created an example design from the MIG IP. The MIG is configured with a 32-bit, 800 MHz DDR interface and a 256-bit, 200 MHz AXI interface. I am doing some tests where I just read a burst of 256 bytes from a row in memory. Then I read the next row along and so on. In the attached picture you can see that it doesn't seem that efficient to do this. Even though I am continuously requesting data I am only getting it around 50% of the time. Is there any way to improve this?