Test Setup

Test System
CPU:	P4 3.0E 1MB Prescott
Motherboard:	ABIT Fatal1ty AA8XE i925XE
Memory:	2x 512 MB OCZ EL DDR2 PC2-6400 Platinum XTC
Video Card:	ATI Radeon X850 Pro PCI-E
Harddisk:	Maxtor Diamondmax 160GB
Power Supply:	HEC PurePower 475
Software:	Windows XP SP2, Catalyst 5.13

Performance

As first test we ran the memory at 200 MHz, at the standard voltage of 1.8V, to see how it performs at everything set to stock. The next test determines the maximum overclocking at 1.8V, which is something you would experience if your motherboard does not allow memory voltage adjustments at all.
For the next three tests we raised voltage to 2.1V, 2.2V and 2.3V. As you can see there is no difference in maximum clock here.
In the next two tests we reduced memory timings to 3-3-2-4. The first test at CL3 allows you to compare performance of CL3 vs. CL4 (first test).
The last two tests are running at the maximum CPU and memory clock which is a real-world scenario. You want to maximize both CPU and memory speeds. The first test here is at memory 1:1 in sync to FSB, the second one uses the 3:4 multiplier. These two tests are to find out if and how much the Pentium 4 profits from running at higher memory speed than FSB. As you can see the gains are there, but not so big, except for the Everest Write Benchmark. Depending on the application you will see a 1%-3% performance increase.

During testing I realized that the i925XE's memory controller seems to become unstable in the 400 MHz memory range. I would expect that there is some more overclocking possible with this memory.

For further comparison, the test "JEDEC DDR2-400" shows a generic DDR module running at JEDEC standard timings.

OCZ EL DDR2 PC2-6400 Platinum XTC
CPU Clock & Memory Ratio	Memory Speed	Memory Timings	Everest Read	Everest Write	Everest Latency	Quake 3 Timedemo	3DMark 2001SE	SuperPi Mod 1M
15 x 200 1:1	200 MHz	4-5-4-15 1.8V	5228 MB/s	1524 MB/s	103.4 ns	284.2 fps	20900	45.08s
15 x 193 1:2	387 MHz	4-5-4-15 1.8V	5552 MB/s	2139 MB/s	81.9 ns	287.3 fps	20434	44.00s
15 x 200 1:2	400 MHz	4-5-4-15 2.1V	6029 MB/s	2324 MB/s	78.9 ns	292.9 fps	20958	42.69s
15 x 199 1:2	400 MHz	4-5-4-15 2.2V	5950 MB/s	2293 MB/s	79.4 ns	291.2 fps	20875	42.95s
15 x 198 1:2	400 MHz	4-5-4-15 2.3V	5952 MB/s	2294 MB/s	80.0 ns	290.1 fps	20769	43.13s
15 x 200 3:4	267 MHz	3-3-2-4 2.1V	5954 MB/s	2141 MB/s	88.2 ns	290.3 fps	20919	42.70s
15 x 216 3:4	288 MHz	3-3-2-4 2.1V	5954 MB/s	2141 MB/s	81.6 ns	313.7 fps	22168	39.53s
15 x 291 1:1	291 MHz	4-5-4-15 2.1V	7600 MB/s	2223 MB/s	71.0 ns	412.0 fps	27175	30.89s
15 x 291 3:4	388 MHz	4-5-4-15 2.1V	8127 MB/s	2691 MB/s	66.0 ns	417.9 fps	27692	30.06s
JEDEC DDR2-400	200 MHz	4-4-4-12 1.8V	5326 MB/s	1585 MB/s	102.0 ns	284.5 fps	20159	44.45s

For an easier comparison with other modules, we set a maximum voltage of 2.1V and tested until we found the highest clock frequency and fastest timings for this memory. The benchmarks Everest Read, Everest Write and Quake 3 were run. We then calculated the performance increase in percent compared to a generic DDR2-400 memory running at JEDEC DDR2-400 (4-4-4-12). The average percentage of the three benchmarks is listed in following table: