Memory Benchmarks
Pifast and Sciencemark let us get a picture of how the memory subsystem, of a given complete system, contributes to overall performance.
It takes nearly 18% longer to complete Pifast when we're running our artificial load on the second station. Sciencemark will show us how the artificial load effects overall memory bandwidth and memory access latency.
86% of the memory bandwidth available when the operating system has no other loads is consistent with what we'd expect, given the articial load outlined on the previous page.
It takes 6% longer to access main memory with the load on the second MagicTwin station, confirming that although the load on the second station is typical of a regular user's workload, it still leaves plenty of processing power for the other station to do meaningful, intensive work.