|Test System "AA8XE"|
|CPU:||Intel Pentium 4 3.0F (S775; 2MB; Prescott)|
|Motherboard:||ABIT Fatal1ty AA8XE, Bios 1.4|
|Memory:||2x 512MB OCZ PC2-5400 EB 4-2-2-8|
|Video Card:||ATI X800 Non-Pro PCI-E|
|Harddisk:||Maxtor DiamondMax 160GB|
|Power Supply:||HEC Power475|
|Software:||Windows XP SP2, Catalyst 5.4|
PerformanceThe first test we did, is how the memory performs at a stock frequency of 200 FSB with timings recommended by OCZ.
Next, we increased the memory clock step by step to find out what the maximum speeds are, this memory can run at a standard of 1.8V. We reached 281 MHz, which is still a good deal away from the 333 MHz required for the DDR667 operation, OCZ promises.
Once we increased voltage to 2.0V we hit exactly that speed during overclocking tests. Personally I would say, that it is better to run this memory at 2.1V if you are going for DDR667, just to have a little bit of overclocking headroom and make sure stability is right, even when it's hot in summer.
The maximum clock test at 2.4V resulted in an impressive 363 MHz. Please note that OCZ's Extended Voltage Protection covers only voltages up to 2.31V (2.2V + 5%).
The last test "JEDEC DDR2-400" is for comparison with a generic DDR2 module running at JEDEC standard timings.
|OCZ DDR2 PC2-5400 Platinum EB|
|CPU Clock &|
|15 x 200 1:1||200 MHz||4-2-2-8 1.8V||5721 MB/s||1793 MB/s||97.1 ns||302.7||20319||43.27|
|15 x 211 3:4||281 MHz||4-2-2-8 1.8V||6336 MB/s||2256 MB/s||86.3 ns||323.6||21004||40.49|
|15 x 250 3:4||333 MHz||4-2-2-8 2.0V||7531 MB/s||2651 MB/s||74.1 ns||385.2||23322||34.25|
|15 x 272 3:4||363 MHz||4-2-2-8 2.4V||8083 MB/s||2893 MB/s||67.0 ns||416.3||24435||31.35|
|15 x 200 1:1||JEDEC DDR2-400||4-4-4-12 1.8V||5325 MB/s||1585 MB/s||102.0 ns||299.0||19884||44.28|
It was to be expected that the synthetic benchmarks which Everest runs, are to gain most from overclocking. Over 8000 MB/s in Everest Read is awesome.
Application benchmarks clearly show, that there is a good deal of performance to be gained from overclocking.
DDR562 @ 1.8V
DDR666 @ 2.0V
DDR726 @ 2.4V
Performance increase in ScienceMark is consistent as well.
We also tried increasing and decreasing the latencies. Tighter latencies did either not work or not considerably improve performance. With loose timings of 5-5-5-15 @ 2.10 V we could run the memory at an impressive 375 MHz. This shows that even if you reach the memory's clock limits, you can still reduce latencies to bump that up.
Letting a fan blow air on the memory did not help more than 1-3 MHz FSB, so it is not worth it for the added noise.