Please note: When using P965 motherboards like the ASUS P5B with the OCZ PC2-6400 Crossfire memory, make sure to set Write to Precharge to 11, 12 or 13. The manual timing settings in the BIOS default to 10 which is too fast for this memory. When I tried the memory with Write to Precharge at 10 I saw very low clock rates, also the maximum stable frequency got lower after 2.05V. However, the SPD "Auto" setting in the BIOS worked correctly.
Performance
As always, the first test is how the memory performs at the standard frequency of 400 MHz, with standard voltage of 1.8V. The second and third tests are at 400 MHz too, but with different timings of CL5 and CL3. Here you can easily see how much difference changing the timings makes.
Now we tried to relax the timings to CL5, to find out what the maximum frequency is at these timings. DDR2-960 is quite impressive for a memory which is advertised as DDR2-800.
The next four tests show how well the memory performs at standard CL4 timings over a wide range of voltages from 1.8V to 2.1V. I think it is important to note here that the minimum voltage required to run DDR2-800 at CL4 is more than the standard voltage of 1.8V.
What if you don't need extremely high MHz, but want good timings? We tested how far the timings could be tightened. We could reach 3-3-2-4 which is almost the fastest setting the chipset allows. While the clocks are not very high, they can still be useful for a Pentium 4 rig that is running at 1:1 memory ratio.
OCZ PC2-6400 ATI Crossfire Certified
CPU Clock & Memory Ratio
Memory Speed
Memory Timings
Everest Read
Everest Write
Everest Latency
Quake 3 Timedemo
3DMark 2001SE
SuperPi Mod 1M
7 x 266 2:3
400 MHz
5-5-5-15 1.8V
6497 MB/s
4805 MB/s
74.5 ns
363.6 fps
26297
28.92 s
7 x 266 2:3
400 MHz
4-4-4-12 1.8V
6544 MB/s
4824 MB/s
72.5 ns
364.8 fps
26490
28.81 s
7 x 266 1:1
266 MHz
3-3-2-4 2.1V
5325 MB/s
4765 MB/s
92.9 ns
349.7 fps
24753
30.10 s
7 x 298 2:3
447 MHz
5-5-5-15 1.8V
7278 MB/s
5336 MB/s
66.4 ns
405.7 fps
28433
25.88 s
7 x 306 2:3
459 MHz
5-5-5-15 1.9V
7472 MB/s
5537 MB/s
65.0 ns
416.4 fps
29165
25.23 s
7 x 314 2:3
471 MHz
5-5-5-15 2.0V
7670 MB/s
5683 MB/s
63.3 ns
427.7 fps
29529
24.54 s
7 x 320 2:3
480 MHz
5-5-5-15 2.1V
7818 MB/s
5786 MB/s
62.0 ns
435.1 fps
29841
24.10 s
7 x 250 2:3
375 MHz
4-4-4-12 1.8V
6128 MB/s
4471 MB/s
77.6 ns
341.5 fps
25343
30.80 s
7 x 269 2:3
403 MHz
4-4-4-12 1.9V
6592 MB/s
4867 MB/s
72.0 ns
368.5 fps
26628
28.54 s
7 x 287 2:3
431 MHz
4-4-4-12 2.0V
7052 MB/s
5150 MB/s
67.5 ns
392.4 fps
27947
26.78 s
7 x 301 2:3
452 MHz
4-4-4-12 2.1V
7414 MB/s
5427 MB/s
64.2 ns
412.3 fps
28814
25.61 s
7 x 226 2:3
301 MHz
3-3-2-4 2.0V
5308 MB/s
4061 MB/s
90.7 ns
303.5 fps
22473
34.66 s
7 x 233 2:3
311 MHz
3-3-2-4 2.1V
5472 MB/s
4171 MB/s
88.1 ns
312.4 fps
22657
33.63 s
With this new graph we are showing how well the memory maximum operating frequency responds to increased voltages. CL3 was not tested at below 2.0V, because the clock rates would be too low to make any sense.