Overclocking Results:

3-4-3-9 5-5-5-15 memory speed: 872 MHz 1056 MHz If you're new to Overclocking and not sure what to do, check out these two excellent Guides for some pointers: Overclocking the CPU and Memory Overclocking the Videocard

With the Corsair Twin2X2048-6400C3 memory voltage set at 2.2V and the timings at 3-4-3-9, the motherboard clock speed was increased in 2-4 MHz increments. At stock settings, the Corsair Twin2X2048-6400C3 did not get very far before problems started to pop up. At 214 MHz the system simply wasn't stable enough to pass 3DMark05, so an increase of memory voltage to 2.3V was needed.

With 3-4-3-9 CAS Latency timings, the memory didn't go up much further, and 218 MHz was the highest the Corsair Twin2X2048-6400C3 memory would run at in the end. Unfortunately due to a situation with the CPU clock multiplier and memory divider, odd multipliers on Socket AM2 Athlon64's cause the memory to run at lower than DDR2-800 speeds.

This is why memory bandwidth numbers with the Athlon64 X2 5000+ based platform are considerably slower than with the FX-62. This means if I want the memory to run at DDR2-800 speeds, I'd have to use only even multipliers. In this case, we decided to leave the clock multiplier at its stock 14x setting. The CPU speed will affect the benchmarks slightly, but not enough to completely skew the results.

Loosening up the timings to JEDEC DDR2-800 settings of 5-5-5-15 allowed the memory to clock up much higher. Continuing from where the Corsair Twin2X2048-6400C3 left off with tight timings (218MHz), the memory was pushed further in 4-6 MHz steps.

With a voltage of 2.3V, the Corsair Twin2X2048-6400C3 memory maxed out at 264 MHz which is pretty good for DDR2-800 rated memory. We only benchmarked at the lower speed/lower latency setting however.

Prelude to Benchmarks

The details of how the Corsair Twin2X2048-6400 C3 memory test system was configured for benchmarking, including the specific hardware, software drivers, operating system and benchmark versions are indicated below. In the second column are the general specs for the reference platforms this pair of DDR-2 memory is to be compared against. Please take a moment to look over PCSTATS test system configurations before moving on to the individual benchmark results.

PCSTATS Test System Configurations

test system reference system processor: amd Athlon64 FX-62 Clock Speed: 14 x 200 MHz = 2.8 GHz 11 x 218 MHz = 3.05 GHz Motherboards: Foxconn C51XEM2AA (NF 590 SLI) Videocard: MSI NX7800GTX-VT2D256E Memory: 2x 1024MB Corsair CM2X1024-6400C3 2x 1024MB Super Talent T800UX2GC4 2x 1024MB Patriot PDC22G8000+XBLK 2x 1024MB Corsair CM2X1024-6400C4 Hard Drive: 74GB Western Digital Raptor WD740 CDROM: AOpen Combo 52x PowerSupply: PC Power & Cooling TurboCool 510 SLI Heatsink: Athlon64 FX-62 Reference Heatsink Software Setup: Microsoft WindowsXP SP2 nVIDIA Forceware NF4 9.34 nVIDIA Forceware 91.27 Benchmarks: Business Winstone 2004 Content Creation 2004 SiSoft Sandra 2005 PCMark05 3DMark05 Doom 3 Quake 4 Battlefield 2

Office Productivity	Source: Zdnet

Business Winstone 2004 runs real applications through a series of scripted activities and uses the time a PC takes to complete these activities to produce its performance scores.

Content Creation Winstone 2004 is a system-level, application-based benchmark that measures a PC's overall performance when running top, Windows-based, 32-bit, content creation applications in Windows XP.

Office Productivity
Business Winstone 2004:	Points	Ranking
2x 1024MB Super Talent T800UX2GC4 (200/800)	36.2
2x 1024MB Patriot PDC22G800+XBLK (200/800)	36.3
2x 1024MB Corsair CM2X1024-6400C4 (200/800)	36.3
2x 1024MB Corsair CM2X1024-6400C3 (200/800)	36.8
2x 1024MB Corsair CM2X1024-6400C3 (218/872)	37.3
Content Creation 2004:	Points	Ranking
2x 1024MB Super Talent T800UX2GC4 (200/800)	46.3
2x 1024MB Patriot PDC22G800+XBLK (200/800)	46.7
2x 1024MB Corsair CM2X1024-6400C4 (200/800)	46.2
2x 1024MB Corsair CM2X1024-6400C3 (200/800)	47.3
2x 1024MB Corsair CM2X1024-6400C3 (218/872)	48.6

The lower latency memory does not do much for the office benchmarks, but that shouldn't be a surprise. Here the system will be user limited anyway.