G.Skill Pi Series 17600 4GB Kit
As you could see from the first page specifications the amount of motherboards this is officially supported by is quite small. Luckily here at the OC3D bunker we've still got the Maximus III Extreme to hand and so it is with that we will be testing the G.Skill Pi. Our other test kit is as follows :
Motherboard : ASUS Maximus III Extreme
CPU : Intel Core i7 870
Graphics : ASUS GTX275
Cooler : Cooler Master Hyper 212 Plus with Arctic Cooling MX3 Thermal Compound
PSU : Cougar 1000CM
OS : Windows 7 64
RAM : G.Skill PIS 17600CL7
The RAM will be tested using our standard memory benchmarks and a selection of games to see if there is a benefit to having such blazing speed.
To begin with we ran the memory using the XMP settings, which gave us the following timings :
Luckily as we're using an 870 it enables us to keep the processor running at the same speed all the time, so any benefits in performance will be purely RAM speed related. Once in Windows 7, CPUz confirmed that the rated 2200MHz @ CL7 is definitely accurate. It might seem an obvious thing to say, but we've tested some high-speed kits that weren't too happy running at their rated timings, so thumbs up to G.Skill for obtaining a blazing speed at rated timings. Of course 7-10-10-28 is a good CAS but very lax on the other timings.
As if to demonstrate how right at the very edge of performance these chips are, over-clocking was a frustrating exercise. It was impossible to tighten the timings and remain remotely stable in the time scale we have, and I'm not sure if given an age it would be possible to get them any tighter.
Further more, the best stable overclock we managed was from 2200MHz to 2322MHz and this forced us to use timings so slack that they should be a college student. Still, nonetheless it's an overclock, which is more than we expected considering the already insane speeds this G.Skill kit runs at "out of the box" and that we remained using the recommended 1.65v.
Let's see if the extra cost for these speeds is justified. Benchmarking time.