The two processors line up just about as one might expect in this memory bandwidth benchmark. Like the other Athlon 64 processors, the EE SFF benefits from not having to traverse a front-side bus in order to reach main memory. Meanwhile, the difference between the Core 2 Duo E6300 and its equivalent with 4MB of L2 cache is virtually nil.
Our relatively unoptimized version of Linpack shows us the size and performance of the various parts of the memory hierarchy on each chip, out to 2MB matrix sizes. Because it can dynamically repartition its L2 cache between its two cores, the E6300's effective cache size in this test looks to be about 2MB. The EE SFF has individual 512K L2 caches dedicated to each core, so its effective L2 cache size here is much smaller.
The integrated memory controller on the Athlon 64 confers an advantage to the EE SFF, but it's not as great as expected because the Core 2 Duo includes a number of optimizations to mask memory access latencies, including out-of-order loads and stores and a sophisticated algorithm for recognizing access patterns and speculatively pre-fetching data into the L2 cache.
Let's have a look at how the memory access latencies look across a broader range of conditions.
The EE SFF gets to main memory faster, but it does have to go there more often due to its smaller L2 cache.