| Corsair XMS2 TWIN2X1024-4300C3PRO 1GB DC Kit |
|
|
| Written by John Chen | ||||||||
| Saturday, 02 April 2005 | ||||||||
Page 3 of 6
Test System and Overclocking Test System:
The proper method of testing DDR2 requires a CPU with low multiplier options. Unfortunately Intel locks their CPU's, so the only processor that allows low multipliers would be their Extreme Editions. Because the Extreme Edition processors are outrageously priced, I can't afford one for my testbed. Thus the lowest multiplier available is 14x, from the Pentium 4 520 2.8GHZ processor. The Pentium 4 560 3.6GHZ processor I'll be using today provides a range of multipliers from 14x-18x. For testing, the 14x multiplier is used for all benchmarks and tests. The Abit Fatal1ty AA8XE is one of the best overclocking motherboards available. It can reach extremely high FSB without any modifications. The motherboard also allows high VDIMM options of up to 2.5v. The stock voltage of DDR2 modules is 1.8v. For those who are not familiar with voltage scaling of DDR2, 2.5v to DDR2 would be the same as pumping 3.2v to DDR. All overclocking tests will be conducted with 2.3v, which is the highest VDIMM the majority of quality LGA775 motherboards provide. Anything beyond 2.3v will be too much for DDR2 memory anyway. Overclocking: Keep in mind that during overclocking, the latencies are as follows: CAS-TRCD-TRP-TRAS. A memory ratio of 3:4 is used throughout all tests. The first thing that I aimed for was the lowest latencies allowed at a stock speed of 266MHZ. Knowing that the XMS2 PC2-4300 Pro uses the "fatty" chips, I immediately shot for timings of 3-2-2-5. Sure enough, the memory ran at those timings. The memory overclocked so well that it even reached 260FSB, resulting in a memory frequency of 346MHZ. That's the highest overclock I've ever reached with these low latencies. The low latency overclock surpassed PC2-5400 speeds! Perhaps Corsair's new low latency PC2-5400 memory uses these chips? I loosened the TRCD and TRP latencies to 3 and was unable to go higher; a 5FSB increase resulted in an immediate non POST. With even looser timings of 4-3-3-10, I reached 275FSB, resulting in 367MHZ. At 14x275, my CPU was at a limit. The engineering sample I have is one of the early productions and doesn't seem to allow any good CPU overclocking headroom. The board's FSB and memory frequency were able to go higher, but the CPU would throttle so much that all benchmarks would run poorly, resulting in poor scores. The memory is definitely capable of reaching higher speeds. As much as I hate to provide bottlenecked overclocking results, it can't really be helped. Using a regular 2.8GHZ processor with the same 14x multiplier will yield the same results. Anyone want to donate an LGA775 processor with low multipliers so I can provide you all with good results? Overclocking Results:
|
||||||||