|
Page 3 of 5
RAM Testing Methodology
All benchmarks are conducted using the same system components and the same memory modules in the same DIMM slots. The memory slots sit directly between the main power connector and the CPU socket on the motherboard. Some systems work better when the memory modules are closer to the power source, and some perform better when the RAM is closer to the CPU. The manufacture of the motherboard used for testing in this review states their preference in the manual: "Install the RAM in the sockets closest to the power connector for better over-clocking capability."
Each benchmark begins after a complete system restart and is repeated five times; the high and low results are discarded, and the average of the three remaining results is calculated and reported in the text. To ensure system stability and the reliability of our results, each new memory configuration was fully tested with Memtest86 v4.00.
One of the advantages of an AMD based-system for memory testing is the wider availability of unlocked CPUs, allowing the FSB and CPU multiplier to be raised and lowered in tandem, so that the CPU clock remains the same while the FSB and base memory clock is increased.
One of the disadvantages of an AMD-based system is a problem that cropped up with the release of the Phenom II chips and the AM3 platform. Namely, there was a "Product Errata" called: #379, DDR3-1333 Configurations with Two DIMMs per Channel May Experience Unreliable Operation. The name is unfortunate, in that it doesn't exactly convey the fact that all memory clocks greater than DDR3-1066 were impacted, not just DDR3-1333. Luckily, time heals all wounds, and there are dozens of CPUs and motherboards today that do not experience any problems with these higher speeds.
Below is a table summarizing the hardware settings used in this review. I stuck with JEDEC standard frequencies, but experimented with tighter timings than the SPD values embedded in the modules. At all the standard frequencies I left the FSB base clock the same, at 200 MHz. When I started to overclock the ram, I had to raise the FSB, but I reduced the CPU multiplier to keep the CPU frequency at 3.6 GHz.
Memory Speed
|
Memory Timings
|
Voltage
|
System Bus
|
CPU Multiplier
|
CPU Speed
|
| 1066 MHz |
7-7-7-20 |
1.50V |
200 MHz |
18 |
3600 |
| 1066MHz |
6-6-6-20 |
1.50V |
200 MHz |
18 |
3600 |
| 1333 MHz |
7-8-7-24 |
1.50V |
200 MHz |
18 |
3600 |
| 1600 MHz |
8-8-8-24 |
1.64V |
200 MHz |
18 |
3600 |
| 1744 MHz |
8-8-8-24 |
1.64V |
218MHz |
16.5 |
3597 |
Test System
-
-
System Memory 1: 2X 2GB OCZ Reaper HPC DDR3 1600MHz (7-7-7-24)
-
System Memory 2: 2X 2GB G.Skill Ripjaws DDR3 1600MHz (7-8-7-24)
-
-
-
-
-
Optical Drive: Sony NEC Optiarc AD-7190A-OB 20X IDE DVD Burner
-
-
PSU: Corsair CMPSU-750TX ATX12V V2.2 750Watt
-
Monitor: SOYO 24" Widescreen LCD Monitor (DYLM24E6) 1920X1200
-
Benchmark Applications
-
Passmark Performance Test v7.0 Build 1011
-
EVEREST Ultimate Edition v5.30.1900
-
SiSoftware SANDRA v2009.9.15.124
-
Crysis v1.21 Benchmark Tool
-
Memtest86 v4.00
|
