Overclocked testing
We also tested the memory at the highest stable overclock we could achieve with extra voltage. We tested using "optimal" SDP settings that were selected by the Inno3D 680i motherboard, as well as limiting the timings to 5-5-5-15. This is because the motherboard will increase the latency beyond any perceivable overclocking benefit.
Comparing the maximum overclocks at 5-5-5-15 timings of the PC6400 memory to some more expensive OCZ PC2 9200 FlexXLC, which is underclocked to 1066MHz front side bus, but its timings are also reduced to the lowest we could find. This shows the difference between buying extremely fast but expensive memory and that of the more budget DDR2-800 overclocked.
This is an example that with a little effort you can do better than a motherboards "optimal" timing. Only the Corsair does more raw MHz when we didn't limit the timing to 5-5-5-15, the rest just achieve the same speed and voltage, but the motherboard reduces the timings to play it safe
There's actually a drop in performance from the stock 800MHz, even though we're seeing on average a 20% increase in clock speeds. This is mostly because of the necessary increment in memory timings.
Want to comment? Please log in.