Memory and single-threaded performance
In SiSoft SANDRA's buffered benchmark, the two AMD chipsets are better
able to harness the potential bandwidth available, unhampered by the
Intel's FSB bottleneck. Any additional bandwidth is welcome
when
the IGP can siphon off a considerable portion.
Memory-access latency results are rather surprising, with the Intel
platform having a lower access time despite the memory controller being
located on the northbridge, rather than integrated into the CPU like
the AMD systems. In addition the Intel is running at looser memory
timings. This highlights just how much of a benefit Intel's more
aggressive pre-fetchers can have.
A lot of the potential performance of the quad-core AMD Phenom X4
9850BE goes untapped in these applications, with the dual-core Intel
E8500, in the G45 based DG45ID, proving considerably quicker in both
the single-threaded HEXUS.PiFast number-crunching and dual-threaded
WAV-encoding tests.
