Corsair XMS2 5400UL 1Gb Kit DDR2 Review

Memory by KeithSuppe @ 2005-11-12

Corsair originally designed their XMS2 5400UL for optimum performance on the nVIDIA C19 chipset. Unfortunately there have been some problems innate to the C19 and it seems there?s a Pink Elephant in nVIDIA?s living room. Today we not only test Corsair?s Twin2X-5400UL,, but in the process discuss the Pachyderm stepping on motherboard maker?s toes.

  • next

Memory for the C19 chipset

Madshrimps (c)


Corsair originally designed their Twin2X1024-5400UL for optimal performance in conjunction with the nVIDIA C19 chipset as this TweakTown article exemplifies. Among motherboards sporting this SLI chipset, the Asus P5ND2-SLI Deluxe was particularly well suited for Corsair's 5400UL due to a BIOS feature allowing adjustment of Command Rates. Purportedly the C19 is the first chipset designed for Intel DDR2 platforms with the ability to run memory at CMD 1T. As I began testing on the Asus P5ND2-SLI I quickly discovered an anomaly with utilities such as WCPUUID, CPU-Z, SiSoftware's Sandra, Everest, etc. had been reading memory speed at 1/4 or 1/2 the frequency set in he BIOS. Had this been an isolated incident I'd say something was wrong with me or my system. However, after testing across three C19 based boards the results were the same. I tried every BIOS available for the Asus including 0605, 0708, 0802 and did the same on a Gigabyte GA-N8-SLI Royal. To exemplify the anomaly I've taken a photo of the P5ND2-SLi BIOS screen after stabilizing the system (Pentium 630/P5ND2-SLI Dlx, BIOS 0605, 5400UL) at 15x270=4050MHz, and a memory speed of 850MHz DDR2.

Madshrimps (c)


After booting into Windows XP under the settings seen above I ran and snapped two instances of CPU-Z (ver. 1.30) which lists the memory frequency at 213.1MHz (1/4 actual speed) on a Divider ratio of 14:11.

Madshrimps (c)


Running Sandra Memory Bandwidth benchmark gives us a bandwidth score of 6740MB/s. In the windows just below the bars where Sandra displays chipset model and logical chipset, values are displayed as "4x 271MHz" (FSB), next we see the anomaly reading: "2x 212MHz (424MHz data rate)". Another indicator something was amiss was the absence of CL rates which Sandra usually displays. Of course the score itself is most telling at 6740MB/s which would be miraculous for the indicated memory speed especially given the latencies as shown by CPU-Z? Multiplying the (1/4) value 212MHz x 4 = 848MHz (850MHz) provides the correct frequency.

Madshrimps (c)


Everest Home Edition memory benchmark usually indicates memory frequency in real-time. In the screenshot below Everest interprets that speed as "Dual DDR2-427" 1/2 the actual 850MHz speed. Also included in the screenshot is Asus AIBooster which indicates a value of 0 for FSB and CPU Utilization.

Madshrimps (c)


Certainly a conundrum since the divider as indicated in CPU-Z reads 14:11. This would make the memory speed as indicated in both utilities incorrect at 427MHz (x2) or 213MHz (x4)? Where then does the divider fit into this formula if the memory speed remains unaffected? I've inquired into this issue for several weeks unable to get a definitive answer, perhaps the Dual 16X motherboards will fare better. We will be doing a review on the Asus P5ND2-SLI Deluxe very soon and hope to compare that board to its Dual 16X counterpart. Perhaps we'll delve more into the issue at that time.

Test Speeds

I used two motherboards to test this memory; Asus P5AD2-E Premium (BIOS 1005) and the Asus P5ND2-SLI Deluxe (BIOS 0605). In order all benchmarks and the CPU-Z speed, Latency, Ratio screenshots below occur in the following order: P5AD2-E (925XE) 254FSB 679MHz 3:4 CL3-2-2-8, P5AD2-E (925XE) 265FSB 704MHz 3:4 3-2-2-8, P5ND2-SLI (C19) 1066FSB (QDR) 672MHz 11:7 CL3-2-2-8, P5ND2-SLI (C19) 1085FSB (QDR) 850MHz 14:11 4-4-4-14. Clicking on the thumbnails below exemplifies each formula after running all benchmarks.

Madshrimps (c) Madshrimps (c) Madshrimps (c) Madshrimps (c)

  • next