Maximum OC table

Have a look at the overclocking results in our memory table with maximum overclocks now:



We did not expect the PC6400 "Urban Elite" to beat our king of the hill in this graph, the Team Group, but the added result does provide the speeds where the future PC6400 kits will have to measure up to. Also clear from this graph is the impressive showing at the cas 5 setting, at this price point of course, and the quite fast cas 4 performance for an 800Mhz part.

Performance numbers @ max. OC settings

Time to put our maximum overclock numbers from our previous chart to the test. Let's start again by looking at Sisoftware's SANDRA memory bandwidth test (longer bars are better):



As you can see from the chart, the cas 5 performance comes in the neighborhood of the more expensive kits. Cas 4 performance falls short a lot more, but it's far too soon to come to any conclusions. What is clear though is that, at cas 4 and 5, the memory wipes the floor with our basic (generic) module, offering far better bandwidth. You can also see that the cas 3 setting really falls short there, a combined issue of the rather slow 4:5 (667Mhz) memory strap and the limited performance of the memory at cas 3. Cas 3 is not the way to go with these modules...

Let's revisit Lavaly's Everest now (lower latency numbers = better):



In the latency test the results from the SANDRA bandwidth tests are confirmed, to say the least. You can see the cas 5 setting holding its own against much more expensive modules, the cas 4 performance falling behind a little, while the cas 3 performance has a really bad showing. Again, this is the difference between a 1:1 memory strap vs. a 4:5 memory strap, where the last one is penalized heavily on the 965 chipset. If you do want to run cas 3 with this speed grade of memory, you'll have to make sure to do it on a 1:1 (533Mhz) memory strap.

Onto the absolute performance numbers now: in the following tables you can find the test results for our benchmark suite on the different cas settings.




At this point a small reminder: our testing of the cas 3 setting was done using a 9x multiplier on the cpu, and the 667Mhz (4:5) memory strap in the bios. Because of the much lower FSB of the OCZ (a result of the limited OC), results differ greatly from the other kits. We'll have to wait for some more PC6400 kits to see how they scale...



Again the same reminder: our testing of the cas 4 and 5 setting was done using an 8x multiplier on the cpu, and the 800Mhz (2:3) memory strap in the bios.

This table is more interesting, as it provides some insight in the results from high front side busses and high speed memory. Please compare the PC6400 results with our generic modules now. The generic modules are tested at a CPU speed of 2.66Ghz, whereas the PC6400 in this case are tested at almost 100Mhz less, 2.565Ghz. Even with 100Mhz less cpu power, the PC6400 Urban Elite was able to secure 6 wins out of 11 tests, while coming really close in the other tests.



Here the results are rather difficult to compare again. It's nice to see though that the results from today's kit lean closer towards the PC8000/8500 parts than towards the generic kit.

Let's wrap it all up now on our final page ->