Testing Method & Test Setup

To give the overclocking oriented memory a run for it’s money one would obviously need a highly overclockable platform. To make sure that our testing rig is capable of running memory at least at its rated speed, prior to this review we have binned over twenty i5-3570K and i7-3770K processors based on their memory clocking capabilities. While we found one or two that did well above 1300MHz, vast majority of chips were starting to struggle at 1200 which raised a question whether assigning specifications above what an average platform would do regardless of skill involved could be classified as pure marketing-caused cheating. Seeing how more and more memory makers are catching on with this trend, we might soon come to a point where memory spec will be just a collection of random unachievable numbers.
In the meantime, we will proceed with the overclock testing using our best CPU that is fully stable up to 1333MHz or DDR3-2666.

Motherboard	ASUS Maximus V Gene (BIOS 0086)
CPU	Intel Core i7-3770K
Graphic card	XFX 8600 GT
Memory	G.Skill TridentX F3-2600C10D-8GTXD
HDD	Samsung 40 GB
PSU	Silverstone OP1000
OS	Windows 7, 64 bit SP1

Results

We will verify stability of every setting with eight 750MB instances of HCI Memtest which is one of the toughest tests around. If we see a single error before we achieve 150% test completion then we call things not stable, if not – we raise the memory frequency by 10MHz and start the test all over again until we reach the final result.



As with most of modern ICs, first timing called CAS Latency is a linear function of frequency and voltage. The higher voltage you run, the higher chances you have of lowering CAS at given frequency. Second and third primary timings, tRCD and tRP, are not a function of voltage – certain values either run at certain frequency, or they don’t. Most of the time, you would want to run them equal, but in border areas it is sometimes possible to run tRP one value below tRCD.
Looking at the situation in general, we see that there isn’t much frequency or timing headroom above specification so, provided your memory controller can do 1300MHz, the best you can do to gain speed is fiddle with tRAS and subtimings. If luck is not on your side and you hit a bad CPU, you can opt out for 1200MHz using 9-11-11-28 timings, which are still reasonably tight.

Page 1 - Introduction	Page 4 - Results
Page 2 - Closer Look	Page 5 - Some SuperPi 32M
Page 3 - Photo Gallery	Page 6 - Conclusion

Discuss this article in the forums