Test Setup

Performance & Overclocking

After installing the memory and booting for the first time, CPU-Z was used to gain an insight to the SPD programming of the modules. At first sight, these do not make much sense, as the memory is set to handle much more than just 1220 MHz. The memory is rated to run at 1.8V to 2.0V and 933 MHz (1866 MHz).

Starting out we use the rated speed at the minimum voltage mentioned - in this case 1.8V - to check how the memory reacts at these settings. It booted right away without any problems and is completely stable. Using the ASUS P35 P5K3 Deluxe board, which has been the mainboard for all of our DDR3 tests up to this point, we raised the voltage to 2.0V and checked how far the memory would clock. The final stable speed was 1932 MHz or 486 MHz FSB. The memory seems to be able to run faster, but the mainboard cannot handle more than 486 MHz stable and freezes continuously at 489 MHz. New DDR3 boards like the nForce 790i or X38/48 should be able to handle 500 MHz FSB which translates to 2 GHz memory. Such speeds at CL8 are quite promising and a good basis for some very fast scores. The near future should mark the turning point for DDR3.

After reaching the limits of the mainboard, we concentrated on tightest timings and work our way up. First increasing the voltage from 1.5V to 2.0V in 0.1V steppings while keeping the same CL rating, then relaxing the timings and restarting this benching cycle.

The Cellshock Kit can manage 1333 MHz at 1.8V and over 1400 MHz at 2.0V at CL5. This is already a combination you will never be able to achieve with retail DDR2 modules. At CL6 we were able to break 1700 MHz and at CL7 we managed to run the memory at its rated speed - 1866 MHz. So it seems that Cellshock keeps the memory at relaxed timings, just to make sure it runs at the rated settings in any scenario.





The memory kit reacts very nicely to an increase in voltage. The increase is very consistent with every CL setting. This gives you the most possibilities to tweak your memory subsystem to run perfect in your system. Just remember, that these benchmarks only reflect our sample and your milage may vary.