Test Setup

Performance and Overclocking

The first thing I did was to test the 'High Frequency' EPP Profile. Unfortunately you can only select an EPP profile on a motherboard with an NVIDIA chipset, so I had to enter the timings, voltages and frequencies myself. The settings worked a treat to my great surprise. I honestly wasn’t expecting the modules to reach 1 GHz with only 2.2V. I tested for stability with memtest86+, which I allowed to run for an hour without finding any errors.

In the first of the benchmark tests we evaluate how these modules do at their default settings, to establish a baseline performance value. Running @ 800MHz 4-4-4-12, these modules are already performing very well, so it will make the overclocking comparisons very interesting.

I immediately noticed that these modules get very warm, even at their stock 2.2V rating. If you plan to do any major overclocking with these modules, I strongly recommend that you have good airflow around the modules at all times. For testing, I used an 80mm fan, blowing cool air over them, which seemed to work well. I have spoken to Crucial about this and I was told the excess heat was due to increased voltage and frequency, and that they're ok for up to 85°c.

Unfortunately my motherboard will be the limiting factor in this review, since it can only reach a maximum of roughly 438MHz FSB without becoming unstable. I did one run @ 438MHz 1:1 to demonstrate the potential that these modules have. Since 438 MHz was nowhere near the limits of these modules, I tightened the timings slightly to 4-4-3-8. This test resulted in some extremely impressive benchmark results. 10,453MB/s Everest Read is nothing to snigger at! This was also the highest CPU frequency during any of the tests, so it may also skew some of the other benchmarks that are more CPU dependent, like Quake 3 for example.

The next test is to establish the maximum frequency possible on this memory at 2.4V. A fantastic result of 1140 MHz was achieved with 5-5-5-10 timings. Unfortunately, I had to use a 2:3 divider to reach this frequency, which will have an impact on performance during testing. Even the fastest motherboards would have a hard time maxing out these modules at 1:1, you would need nearly 600 MHz FSB! (The Asus P5B Deluxe (i965) or P5K Deluxe (P35) would currently be a good choice of motherboard to pair up to this RAM to get these extreme speeds.)

I repeated the process to find the maximum frequency with CL4 timings. A very impressive overclock of 508 MHz (DDR2-1016) was reached with 4-4-4-12 timings. This might be a better option for benchmarking, since it should offer a lower latency than a slightly higher frequency and looser timings.

Now it's time to find out how much the timings can be tightened. We found that at the timings could be tightened to 3-2-2-4, which are the tightest timings the motherboard will allow. Unfortunately to manage this, 2.4V is needed, and the memory will not clock any faster than 540 MHz without becoming unstable at these extremely tight timings.

For further comparison, the test "JEDEC DDR2-533" shows a generic DDR-2 module running at JEDEC standard timings.