AIDA64 Memory Test

Benchmark Reviews has used AIDA64 and its predecessor Lavalys Everest for years, mainly for its processor benchmarks and CPU stress-testing features. But it also includes a "Cache and Memory Benchmark" that performs read, write, and copy bandwidth tests on a system's installed memory.

The Kingston HyperX memory beats all the other memory in this test, although the "Write" results are very close (about half a percent difference between the lowest and highest scores). Quad-channel memory brings a whole new level of memory bandwidth performance to desktop systems, as you can see in the scores above. In the Read test, the Kingston HyperX memory at 2133MHz beats the generic 1333MHz memory by 20%.

NOTE: In my original tests C states were enabled because the test machine operated with the default BIOS settings. I re-ran the AIDA64 tests on the same motherboard and CPU used in this review with Kingston HyperX Genesis DDR3-2133 memory at enthusiast-level speeds.

C states enabled: read 18924, write 14995, copy 16914, latency 48.8ns
C states disabled: read 18897, write 15082, copy 17101, latency 48.9ns

The difference in write performance is a fraction over 1%, well within the margin of error of these tests. I conclude that at least with this motherboard and CPU, whether or not C states are enabled makes no difference in memory performance as measured by AIDA64's tests.

SiSoft Sandra Memory Test

SiSoft's Sandra Lite is a free version of SiSoftware's "Sandra" benchmarking utility. Its comprehensive memory benchmark tests report a number of items, but we're interested memory throughput tests. Unlike some other benchmarks that merely perform straight sequential reads and writes, Sandra reads and writes different areas of memory, using integer as well as floating point data, all in SMP mode.

The Kingston HyperX memory wins again here, with 2133MHz scores that are 38% better than 1333MHz memory and 15% better than 1600MHz memory.

Euler 3D

Euler 3D is a Computational Fluid Dynamics (CFD) program that is multi-threaded and computationally very intensive. The benchmark version by Case Labs is built with the Intel FORTRAN compiler and uses 8-byte double precision floating point math. The test case simulates a Mach 0.5 airflow over a NACA 65A004 airfoil section. The benchmark score is the CFD cycle frequency, with higher scores being better.

This benchmark kind of straddles with "synthetic" and "application" benchmark worlds. While the DDR3-2133 HyperX memory still wins, it's by a much narrower margin than on the pure synthetic tests.

Join me in the next section as I run the application benchmarks.

<< Previous Page - Next Page >>