Continuing on from our Athlon 64 X2 4200+ and Sempron 3400+ Linux benchmarks, today we have looked at AMD's DDR2 performance with their Socket AM2. Using the system we used to deliver our initial AMD AM2 performance results, we have run a matched pair of Corsair XMS2-5400UL DDR2 through its paces at different frequencies. Used for the memory benchmarking under Linux was RAMspeed, which is an open-source cache and memory benchmark, and has been part of our Phoronix Linux benchmarking suite for some time now.
|Processor:||AMD Athlon 64 X2 4200+ (2.20GHz)|
|Motherboard:||ASRock AM2NF4G-SATA2 (nForce 410)|
|Memory:||2 x 512MB Corsair XMS2 DDR2-667 (5-5-5-15)|
|Graphics Card:||NVIDIA GeForce 6800GT 256MB|
|Hard Drives:||Seagate Barracuda 7200.9 160GB SATA2|
|Optical Drives:||Lite-On 16x DVD-ROM|
|Power Supply:||Sytrin Nextherm 460W|
|Operating System:||Fedora Core 5|
|Linux Kernel:||2.6.16-1.2122_FC5 SMP (x86_64)|
|Graphics Driver:||NVIDIA 1.0-8762|
The DDR2-667/533/400 system memory frequency was controlled through the AM2NF4G-SATA2 BIOS. Originally, we had intended on delivering DDR2-800 results as well, seeing as the AM2 memory controller supports this speed, but we had experienced some memory compatibility problems when using DDR2-1000/800 memory with the nForce410-powered AM2NF4G-SATA2. With RAMspeed, while five paces are generally used for our Linux memory reviews, we had run each of these memory frequencies for twelve times to ensure accuracy. Both integer and floating-point tests were used. These tests can be replicated by using ramspeed -b 3 -l 12 and ramspeed -b 6 -l 12. RAMspeed was compiled in our tests for 64-bit Linux, and the version we used was the latest at time of testing. We hope to complete additional AM2 memory performance benchmarks in the very near future, as well as looking at the memory latencies, and specifically the DDR2-800 performance, but until then we will save on drawing any tentative conclusions to AMD's DDR2 adoption.