Test Setup

Performance & Overclocking

Base Settings

This section is new and illustrates what settings were used to attain the advertised speed of the memory sample in this review:

Base Clock: 200 MHz
CPU Multiplier: 20x
CPU Speed: 4000 MHz
CPU Voltage: 1.7 V
CPU VTT (UnCore) Voltage: +400 mV
X58 I0H Core Voltage: 1.34 V
Memory Multiplier: 10x

These are just what our Foxconn Bloodrage board booted easily with, to reach the Base Clock and resulting 2000 MHz memory speed. We should mention at this point, that some of these values are well above something you would use for a 24/7 overclock. We are not responsible for what you do with your system and our settings are by no means what you should use for a normal workstation of yours!

Performance

First off, we started up CPU-Z to check the programming of the modules. It looks like G.Skill has a correct XMP profile embedded in the DIMMs. This should make it easier to run them at the advertised speed right out of the box. It is interesting to see the three levels of lower speeds stored in the modules. This means that the memory should boot successfully on most boards, so that you can easily set them up properly. Now that we know that the memory can run at the rated speed, we dropped the voltage down to 1.5 V, which is the default for DDR3. The Base Clock was also lowered to 133 MHz and we tried lowering the timings as much as possible. The memory did not boot at any speed in combination with CL5, but we had no problems to run the kit at CL6-6-6. At this setting the memory managed to run 720 MHz with 1.65 V. This is a bit less than the Mushkin Redlines, which worked with up to 725 MHz at this setting. Relaxing the timings to CL7 meant that we were able to easily break the 800 MHz mark. The memory booted and had some head room at 1.5 V and even went a bit higher with 1.65V. Once again, the kit trails the Mushkin memory by quite a bit at these tight timings. Moving unto CL8, this is where the G.Skill Trident manages to pass the Redlines and push a very respectable 962 MHz while the Mushkin kit called it quits at 911 MHz. The Trident even manages the advertised 1000 MHz mark at CL9 with a mere 1.5 V, but raising the voltage to 1.65 V did not give us a lot more headroom to overclock.

It should be mentioned, that at this point, it is possible that our test system has been pushed to the limit, as we were only able to tickle around 4.1 GHz out of the CPU. On the other hand, overclockability suffered greatly when pushing the voltage above 1.65 V. While Intel explicitly states that anything over 1.65 V is not good for the entire platform and may kill the parts, we ignored all the warning signs and pushed the memory to 2.0 V to check if this would yield higher performance. Using such insane settings may be of interest for extreme overclockers. The higher we raised the voltage the worse our overclock became, so the G.Skill memory is not intended for such scenarios. Seems like the memory ICs which manage more speed with high voltage are still being used for dual channel kits, which makes sense from a business perspective.




As we mentioned before, raising the voltage severely slowed down the memory. Our kit did not like anything above 1.65/1.7 V and netted much lower speeds. This phenomenon became worse, as the voltage was raised higher. At CL9 we have an almost flat line, which seems to suggest, that the memory may be capable of just a tad bit more than our test system can handle, as the drop towards 2.0V is not there, much unlike CL6 through 8. Remember - your mileage may vary!