Corsair TwinX 4000
Today I'll be using a Pentium-4 2.4C stepping L310A735-0101 Malay, the Epox 4PCA3+ with updated BIOS, Thermalright's venerable SLK900U, with Vantec Tornado (helicopter engine) 92mm/119CFM HSF, 1GB-kit of OCZ, and 1GB-kit of Corsair. Both are rated as DDR500 at a CAS-3-4-4-8 SPD. The HDD will be Maxtor Diamond Max Plus 9, 120GB SATA150 HDD, Tyan Tachyon G9700Pro, and perhaps most important PCPower&Cooling Silencer 400W PSU.
Windows XP is our OS, and we’re using ATI's Catalyst 3.6 drivers.
I'm going to make this a simple and to the point review. Sandra memory benchmarks will be used MAX-3 version and WCPUID for verification purposes.
First up; the Corsair. I went right to work, and benched the memory to its Maximum stable FSB at 1:1 ratio. Corsair was capable of reaching 266FSB, at SPD timings. No additional PAT features were enabled. At this point in time neither the Corsair, nor OCZ was able to run in another divider besides 1:1. Since this review BIOS revisions have ameliorated this anomaly.
Below is the Sandra screen shots, this was at 2.9VDIMM, as increasing the voltage beyond 2.9VDIMM had no effects in so far as an increase in performance above 266FSB 1:1.
I tried every conceivable setting to raise the performance, and you'll see from the next screen shot where I lower the CAS from CAS-3 to CAS-2.5, has a very surprising effect;
Yes, there was a slight bandwidth drop. I attribute this to "forcing" the memory to run at a setting lower then Corsair's intended SPD for these modules. This is where I'd like to reiterate the point of our misunderstanding where current DDR-I is headed by. Corsair engineers know what their doing.
They carefully scrutinize IC's prior to packaging them on the PCB. Although the following statement is going to draw criticism from the four corners of the (flat) world; reducing certain latency's does not by necessity, result in better performance. So while you'll criticize this statement, look again at the screen shots above.
Now onto the OCZ GolD4000...