Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
CPU Intel P4 840 D P4 820 D P4 630 P4 640 P4 650 P4 660 P4 670 AMD Athlon64 3500+ 3700+ 3800+ 4000+ X2-3800+ X2-4200+ X2-4400+ X2-4600+ X2-4800+ 1-Way Opteron Opteron 144 Opteron 146 Opteron 148 Opteron 150 Opteron 152 2-Way Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Opteron 248 Opteron 250 Opteron 252 2-Way Dual Core Opteron Opteron 270 Opteron 275 nVidia GF 7800GT GF 6800GT GF 6600GT ATI R X850 XT PE R X850 XT R X800 XT PE R X800 XT R X800 XL Memory Corsair Crucial Kingston Mushkin OCZ What are you shopping for?

From theory to practice

Let’s start with the simple stuff, that is the performance delta between different timing settings without altering the bus speed. There are different utilities or benchmarks that can be used for this purpose. One commonly used benchmark is SiSoft Sandra’s memory benchmark that measures memory bandwidth from the main memory to either the integer or else the floating point unit of the CPU. SiSoft Sandra memory benchmark encompasses all different cache levels that are in between the CPU core and the memory, therefore, the differences in numbers do not exclusively reflect a performance delta of the system memory path alone.

SiSoft Sandra memory benchmark scores for the ASUS CUSL2 (i815 Solano chipset) and an FC-PGA PIII Coppermine running at 1 GHz (133 MHz x 7.5). The memory bus is in both cases running at 100 MHz. The lower columns show the bandwidth in MByte/sec at CAS delay, RAS-to-CAS delay and RAS precharge set to 3 cycles each (CAS-3). The upper column pair shows the bandwidth in MByte/sec at CAS delay, RAS-to-CAS delay and RAS precharge set to 2 cycles each (CAS-2). Both categories measured show an increase in performance of 21% and 24%, respectively.

What does this mean for real world applications?

Usually, business (office) applications are only marginally affected by memory timing parameters since in most cases, the relevant data are small enough to be stored in the level 2 cache. Therefore, the main memory is not even accessed in most office applications after the initial read-out of the data and writing them back to the cache.

The situation is quite different if one looks at 3D applications. By definition, video memory is excluded from being cached and, thus, most of the data traffic occurs directly between CPU, main memory and AGP. One benchmark famous for its sensitivity regarding latencies and bandwidth is Expendable "go.exe". In addition to being very sensitive to memory issues, Expendable also poses substantial demands on the CPU, whereas the graphics adapter is relatively unchallenged.

Expendable frame rates, running at 1 GHz, (133MHz x 7.5, memory bus at 100 MHz) to compare 3:3:3 latency timing settings to 2:2:2 settings. Expendable is a rather extreme case showing 4% performance delta caused by the different latencies.

next page:    => what about 133 MHz? =>