Test Setup

Test System
CPU:	Intel E6300 Conroe 1.86 GHz, 2 MB Cache
Motherboard:	ASUS P5Q Intel P45
Video Card:	PowerColor X800XL Pro 16 PCI-E
Hard disk:	Samsung P80 80 GB
Power Supply:	Ultra V-Power 450W
Software:	Windows XP SP2, Catalyst 8.3

As you can see, the ASUS P5Q was used in this review. The memory is stated to run on P35, P45, X38 and X48 as well as some high-end NVIDIA chipsets. But the ASUS P5KC did not manage to boot with the memory at 1.8 or 1.9V at the rated speed. It needed at least 2.3V, so the mainboard was exchanged with the mentioned Asus P5Q, which is a fairly cheap Intel P45 based board. It ran the memory at the rated settings right of the bat.

Performance & Overclocking

The memory in combination with the heat spreaders does not interfere with any components on the mainboard. You will also be able to fit four of these next to each other without any problems.


G.Skill has programmed the memory with two JEDEC speeds and the advertised EPP setting of 1100 MHz at CL5-5-5-15.

A starting base for each test run was a 333 MHz front side bus, with the CPU multiplier set at 7x. While pushing forward, the divider was kept constant to show how the memory scales with the CPU. While it is possible to start at a base of 400 MHz FSB with a 6x multiplier, that way the board could become the limiting factor, forcing us to change the divider, which also has an effect on the overall result.

The first setting benchmarked, once the memory has been installed, is the advertised one. The G.Skill Pi Series kit manages the 1066 and 1100 MHz at 1.8V easily. Considering the fact that the company is offering the memory at default voltage already hints at the lack of scaling with additional voltage. Other memory manufacturers offer such high speeds with a base of 2.1-2.2V and at times even more.

The next step was to figure out how far the memory would go with the minimal 1.8V applied to the kit. It still managed 1174 MHz, so you still have a healthy overclocking potential out of the box with this setting at CL5-5-5-15. Then the voltage was raised slowly up to 2.4V with active cooling. The maximum result of 1210 MHz underlines the fact that this memory scales well with a higher CL rating, but is not very dependent on the voltage applied. It should be mentioned that the kit managed 1200 MHz at a mere 2.0V, which is quite impressive. Remember, we are looking at a pair of 2 GB modules.

Now, the CL setting was dropped down to 3-3-3-8. The memory did not manage to boot at 800 MHz at this setting, no matter how much voltage was applied. As the benchmark results show, the kit maxes out at 730 MHz with this setting. Relaxing the CL to 4-4-4-12 enables us to run the memory at 930 MHz instead.

As you can see in the results, the difference in results between 1.8V and 2.4V is certainly not as much as some other kits. The use of lower voltage could also mean a longer lifetime of the memory, but most of use would never use a kit more than for a few months or a year at most anyways.

G.Skill Pi Series 1100 MHz CL5-5-5-15 4GB Kit
CPU Clock Memory Ratio	Memory Speed	Memory Timings	Everest Read	Everest Write	Everest Latency	Quake 3 Timedemo	3DMark 2001SE	SuperPi Mod 1M
7 x 344 1:1	344 MHz	3-3-3-8 1.8V	6302 MB/s	6261 MB/s	84.5 ns	643.3 fps	26045	23.08 s
7 x 365 1:1	365 MHz	3-3-3-8 2.4V	6502 MB/s	6641 MB/s	82.2 ns	676.0 fps	26529	21.79 s
7 x 333 5:6	400 MHz	4-4-4-12 1.8V	6255 MB/s	6073 MB/s	86.6 ns	626.2 fps	25489	23.73 s
7 x 357 4:5	447 MHz	4-4-4-12 1.8V	6854 MB/s	6491 MB/s	78.3 ns	667.0 fps	26554	22.09 s
7 x 372 4:5	465 MHz	4-4-4-12 2.4V	7116 MB/s	6747 MB/s	75.2 ns	697.7 fps	27079	21.20 s
7 x 333 5:8	533 MHz	5-5-5-15 1.8V	7135 MB/s	6061 MB/s	74.2 ns	611.1 fps	26096	23.43 s
7 x 344 5:8	550 MHz	5-5-5-15 1.8V	7368 MB/s	6260 MB/s	71.8 ns	655.9 fps	26564	22.68 s
7 x 366 5:8	587 MHz	5-5-5-15 1.8V	7829 MB/s	6670 MB/s	67.4 ns	703.9 fps	27406	21.26 s
7 x 375 5:8	600 MHz	5-5-5-15 2.0V	8054 MB/s	6828 MB/s	65.8 ns	720.4 fps	27671	20.78 s
7 x 377 5:8	605 MHz	5-5-5-15 2.4V	8118 MB/s	6848 MB/s	65.4 ns	725.4 fps	27626	20.65 s

The scale graph shows quite nicely, how independent the G.Skill memory is to the applied voltage. This is great for boards, which can handle high dividers or front side bus, but do not offer the option to push the voltage of memory any higher. Remember, that these results only represent what the sample is capable of and your success may vary.