Sandra gives us some nice scores across the board. As readily evident, the higher HTT and memory MHz raises the memory bandwidth significantly. Please keep in mind the benches at 285MHz are higher in part because of the higher CPU speed. I would guess the additional CPU frequency is responsible for approximately 100-150MB/sec of that higher score.
Cachemem is a nice benchmark that shows us the importance of tighter latencies over higher memory speeds. The 275MHz Memory running at 2.5 3:3:7 is still strong enough to overcome the scores from the higher frequency and higher speed scores at 285HTT and 3.135GHz respectively.
Sciencemark doesn't seem to be as sensitive as Cachemem when it comes to latencies. Again we see the 285MHz and higher CPU speeds take the lead.
Super Pi 1MB benchmark brings forth some interesting results. At first, it appears that 285MHz at Cas 2.5 4:4:8 and the scores at 275MHz at Cas 2.5 3:3:7 show identical performance, which in effect they do. However, when considering the added CPU speed at 285MHz HTT, the scores would probably be slower if we were able to bring the CPU speed back down to 3.035GHz as it is at 275MHz HTT/Memory.