During another investigation I have noticed that the performance between SL and S are enormous when it comes to memory benchmarking.
I have run the same code and used the same blocks within the ASIC so it should be identical result. The SL and S has the same ARM CPU (996MHz) and PL310 cache and DDR controller. The only difference are the L2$ size that are half in SL compared with S.
My measurement shows that the difference are ~ 5 times better in S than in SL. Could you just elaborated with this result and also the expected benchmark for the DDR (400 MHz)?