Corsair XMS 3200 XL Pro

LED’s and Performance to Match

by Josh Walrath

Initial Impressions

            Corsair is very well known for their build quality, as well as their lifetime warranty.  In the overclocking community, few others stand above Corsair.  The DIMMS themselves are quite heavy, and the aluminum shroud/heatsink is larger than expected.  These DIMMS are very distinct from other competing products, and in a windowed case other users will instantly know what kind of DIMMS are installed.

These are definitely some tall DIMMS, and they may have problems fitting into cases such as Shuttle's XPC series.

            The packaging that the DIMMS come in is quite non-descript, but it does protect the DIMMS as well as give the major features of the memory.  The stiff plastic does help to protect the DIMMS during shipping.  Nothing flashy, it just does its job.

            The DIMMS snap right into place, and I didn’t have to use any unnecessary pressure to ensure proper DIMM seating. 

Testbed 

AMD Athlon 64 3200+ Winchester Core

Gigabyte GA-K8NS Ultra 939

Sapphire Radeon 9800 Pro 256 MB

250 GB WD 7200 RPM Hard Drive

Toshiba DVD-ROM

Antec 480 Watt True Blue Power Supply

Windows XP Professional SP2 

            The new AMD Athlon 64 3200+ 939 is based on the 90 nm Winchester core.  The K8NS Ultra is nearly the exact same board as the higher priced K8NSNXP 939.  This board is based on the NVIDIA nForce 3 250 Ultra chipset.  The SATA drive was attached to the #1 SATA header (NVIDIA SATA controller).  Windows XP was then loaded and updated.

            I also verified that these DIMMs work on a variety of motherboards and processors.  These DIMMS worked successfully at the rated speeds and timings on the following testbeds:

AMD Athlon 64 3700+ on a MSI K8N Neo Platinum (nForce 3 250GB)

Intel 3.4 GHz Pentium 4 on a Gigabyte GA-8KNXP (Intel i875)

AMD Athlon 64 3400+ on a Asus K8V Deluxe (VIA K8T800)

AMD Athlon XP 3200+ on a Leadtek K7NCR18D-Pro II (nForce 2 Ultra)

            Overall compatibility appears excellent. 

Errata

            I found a very interesting phenomena when running the Athlon 64 3200+ at a lower multiplier from standard.  The 3200+ actually runs at 2.0 GHz, and has a 10x multiplier.  When turning the multiplier down below 10, I discovered that the memory divider stayed constant.  For example, when running the motherboard at 250 MHz HTT, with the 8x multiplier, the core ran at 2.0 GHz.  I expected the internal clock multiplier to utilize the 8x setting also, thus rendering a 250 MHz memory bus.  This was not the case.  It appears as though the memory divider could be more hardware bound than originally expected.  When the BIOS was set to a CPU/8 ratio (2000 MHz/8 = 250 MHz memory), the memory actually stayed at CPU/10, though programs such as CPU-ID reported it as CPU/8.

            To get around this problem I had to keep the CPU multiplier at 10, so accurate results could be gathered.  This leads to the problem of apples to apples comparisons.  As the CPU clockspeed was increased, it becomes impossible to determine if the faster memory or the faster core clock contributed to the majority of the performance increases.  As such I decided to only use bandwidth figures between the CPU and main memory. 

SiSoft Sandra 2005

Next: Results and Conclusion

Copyright 1999-2004 PenStar Systems, LLC.