Hardware

Memory Performance Testing Part 1:

The first benchmarking section will deal with base memory bandwidth and performance, and utilizes SiSoft SANDRA 2008 XII SP2.

SiSoft SANDRA 2008 XII SP2 Performance

SiSoft SANDRA 2008 XII SP2 is the latest revision of this popular suite, and supplies a wide range of potential benchmarks. We're concentrating on memory performance, so are sticking with three basic tests - Integer Memory Bandwidth, Cache and Memory, and Memory Latency. All three of these are affected by the type of memory used, and should highlight the potential performance levels of the memory subsystem.

Our first chart covers the Memory Bandwidth testing, specifically the Integer scores. The FPU scores are quite close and follow the same rankings, so there is no need to include both. These results show a very nice jump for the OCZ PC2-8500 Titanium 2x1GB at 1066 MHz, and even a small increase when moving to 1111 MHz. This demonstrates that even compared to low-latency DDR2-800, high-end Intel platforms can still make use of DDR2-1066 or higher speeds.

The next benchmark is the Cache and Memory test, which shows the overall performance of the CPU cache and memory subsystem. This is highly dependant on the CPU cache design, so the processor must be consistent between tests. Here we find another noticeable performance jump between dual-channel DDR2-800 and DDR2-1066, but not so much in the jump to 1111 MHz.

Our final SANDRA benchmark is for Memory Latency, which supplies a rating (ns) of how the overall cache and memory subsystem reacts in terms of latency. Again, this takes the CPU and platform into account, so consistency is required. This also translates into the scores, as even at CL5 timings, the higher-speed/bandwidth scores at 1066 and 1111 MHz show noticeably lower memory latencies.