SDRAM Throughput problems using Arria 10 SOC
HI,
I was doing a test on Arria 10 soc board regarding fsdram0 and fsdram1 bus speed test using msgdma but I am facing the problem as follows:
For fpga fabric 100MHz clock I got a write speed of 8.88Gbps max for my current configurations on fsdram0 bus where as theoretical is 12.8Gbps for 128 bit fsdram bus which is good and around 70% efficiency in usage.
But when i changed fpga fabric clock to 200MHz also I got the same 8.88Gbps as max speed where as theoretical is 25.6Gbps which is very bad where as around 35% efficiency.
The picture shown below is the memory architecture of an Arria 10 SOC where as I used fsdram0 and fsdram2 bus with width of 128bit(port configuration 3)
I am supplying on fsdram0 bus using msgdma st-to-mm based ip config and all the configurations for both mentioned above tests are same where as only difference is fpga fabric clock supplied for the entire process is 100MHz and 200MHz respectively.
I checked the reports of Generate HDL also as I am using f_user_clk of mpu itself which is successfully generating 200MHz clock when changed the clock settings.
My Major doubt is,
Is it suppose to happen like that?
If not then where should I correct it to get the speed/throughput double the speed than I am getting now.