The results we discussed above were obtained with the processor bus of our test platform configured in nominal 266MHz mode. However, expensive high-frequency memory kits are often purchased by enthusiasts and hence they are often used in systems with overclocked processor and overclocked system bus. That is why we decided to add another set of tests to our article. These tests were conducted with FSB working at 400MHz. In other words, the next set of benchmarks was performed with Core 2 Extreme X6800 processor working at 3.2GHz speed obtained as 8 x 400MHz.
At first come the results of synthetic benchmarks and Everest:
By increasing the system bus frequency to 400MHz we managed to raise its peak bandwidth to 12.8GB/s. And in this case the use of high-speed memory is theoretically much more justified. The results of our memory subsystem bandwidth measurements differ from those obtained with 266MHz FSB. Besides the general bandwidths increase we can also point out that high-frequency memory is now performing much better against the background of the low latency solutions. This gives us some hope that it will prove more efficient in real applications than in the previous test session.
The results of the popular SuperPi benchmark are not that much different this time. High-speed memory performs neck and neck with low latency memory.