The system we used for testing today contained the following components:
|Processor:||AMD Athlon 64 3000+ (Winchester)|
|Motherboard:||Tyan Tomcat K8E (S2865AG2NRF)|
|Graphics Card:||Leadtek WinFast PX6600GT TDH|
|Hard Drives:||160GB Western Digital SATA 7200RPM|
|Optical Drives:||Lite-On 16x DVD-ROM & Lite-On 52x CD-RW|
|Add-On Devices:||NetGear WAG311 802.11g (Atheros)|
|Case:||Sytrin Nextherm ICS 8200|
|Power Supply:||Sytrin 460W EPS12V|
|GCC (GNU Compiler):||3.4.3|
|Graphics Driver:||NVIDIA 1.0-7174|
The nForce4 motherboard we chose to use for the testing today is a Tyan server system motherboard. Overclocking was strictly limited as it’s designed for servers and workstations rather than an enthusiast or gamer. Nevertheless, we were able to increase the FSB up to 222MHz (222 x 9 = 1998MHz) while the timings remained un-touched. In all of the benchmarks, the OCZ EL DDR PC3200 Titanium stayed at 2-3-2-5, and also while the CPU was overclocked to its 222MHz FSB. During the entire benchmarking process, AMD’s Cool ‘n’ Quiet technology was disabled from the K8E v1.01 BIOS. For comparison purposes, due to the lack of overclocking and timing options available on this server/workstation system, we also benchmarked 2 x 512MB Mushkin PC4000 (3-4-4-8), 2 x 512MB Corsair XMS PC4400 (3-4-4-8), and 2 x 512MB Ultra Products PC3200 (2-3-2-5).
The benchmarks used in this review were Doom 3 (1.3.1302), LAME (3.96.1), time make, RAMspeed (2.3.1), and Unreal Tournament 2004 (3334). In the Doom 3 benchmarking, the standard time demo 1 was used while running 640 x 480 with Low Quality and then again at 1280 x 1024 Ultra Quality. With LAME, we measured the time it took to encode an 81.3MB song as an MP3 file. For time make, we measured the time it took to compile LAME 3.96.1 using GCC. To break apart the different memory modules we have here for testing from the other real-world benchmarks we used RAMspeed. With RAMspeed we used the options of ramspeed -b 6 -l 5 and ramspeed -b 3 -l 5. The first set of parameters specified to run the FLOATmem (floating point) benchmarks while the –l specified to run in LongRun mode for five times, for most accurate results. The second set of parameters was identical to that of the first set except for INTmem (integer) testing. In Unreal Tournament 2004, we used the maps dm-rankin and ons-torlan with 12 bots for 80 seconds.