ScienceMark 2.0 and HEXUS Pifast
ScienceMark 2.0 - Memory Bandwidth
FX-57 returns a bandwidth figure in the same general area as the other high-end Athlon 64 and Athlon 64 FX products on test. The San Diego core's overall bandwidth is measurably, but only slightly, lower than Clawhammer's. You see a similar trend with Venice compared to older 130nm products, indicating that while tweaks have been made to the memory controller, bandwidth is reduced as a result of them.
ScienceMark 2.0 - Access Latency
In the memory access latency test, San Diego-powered FX-57 posts the lowest access latency numbers of any stock-clocked processor yet tested by us. The core's memory controller tweaks, specifically those in the data prefetcher and with regards to memory efficiency, are likely responsible. Whatever the reason, San Diego accesses memory supremely well.
HEXUS Pifast
Our HEXUS Pifast test, one which calculates the value of Pi to 10 million places, is a good overall test of memory controller and FPU performance in a processor. Changes to the memory subsystem have pronounced effects on the returned results.
FX-57 bests the previous Pifast record holder - FX-55 - by over two seconds. Comparatively, that's the expected boost from an extra 200MHz, so memory controller tweaks don't really show with this software benchmark. Hopefully we'll see the affects elsewhere in our tests.
