Specifications & Product Tour
The specifications for this kit are astronomical, the like we haven’t seen since the last ultra-high speed kit G.Skill sent for review (a DDR3-2400 Pi kit). This kit is even faster than that, guaranteed to run at DDR3-2666 (assuming a capable integrated memory controller, or IMC). Timings aren’t the tightest but to get this kind of MHz, slight timing sacrifices must be made.
|Multi-Channel Kit||Dual Channel kit|
|Tested Speed||DDR3-2666 MHz (PC3-21300)|
|Tested Latency||11-13-13-35 2N|
|SPD Speed||1600 MHz|
|Features||Intel XMP (Extreme Momery Profile)Ready
Removable Top Fin
As usual, G.Skill’s highly rated sticks come in the most nondescript box you can find. If you get a boring-looking box from G.Skill, it is virtually guaranteed that there is some good RAM in there. They come in these boxes because of the included dual fan assembly.
The fan assembly is the standard G.Skill dual fan assembly. My biggest gripe with these is not noise (they’re quiet) or even design (they work as intended). It’s that they don’t have a motherboard fan header on them. Who wants to run an extra, unsightly MOLEX cable when you could just plug these low-amperage fans into – and control them with – your motherboard?
Now on to G.Skill’s newly designed TridentX memory, which is quite nice looking actually. The heat spreaders are for more than just looks and do pull the heat away from the memory well. These sticks are begging to be run at higher-than-normal volts, so you’ll need them to do their job well.
Both people with tight motherboards (restricting use of large CPU coolers) and extreme overclockers that like to use dry ice or liquid nitrogen on their memory will be happy with this feature – a removable top fin!
The kit is doing well in the aesthetics department for sure, but do they perform as well as they look?
Our test system today consists of Intel’s new Ivy Bridge memory-clocking monster, the i7 3770K. To actually use that IMC to its full potential you need a good motherboard and we’ll be using the ASUS Maximus V GENE to fill that spot. For comparison, we benched a RipjawsX DDR3-2133 kit rated at tighter timings of 9-11-9-28.
|CPU||Intel i7 3770K|
|Motherboard||ASUS Maximus V GENE|
|RAM||G.Skill RipjawsX 8GB DDR3-2133 / 9-11-9-28
G.Skill TridentX 8GB DDR3-2666 / 11-13-13-35
|GPU||AMD HD 6970|
|OS||Windows 7 Professional x64|
The TridentX kit matches the red and black ASUS ROG color scheme perfectly.
You’ll notice I didn’t use the included fans. The first reason is there isn’t a MOLEX connector sticking out of the benching station. The second is that the MVG doesn’t have room between the DIMM slots and GPU. Most importantly, to overclock these with more voltage, it is wise to use an actual fan (like the pictured Yate Loon High) to cool them. The included fan is fine for looks and for normal use, but when it comes to actually pushing voltages, you’ll need something stronger.
Stability at Rated Speed
Before performance testing, it’s always good to look at stability, especially in a kit rated for such high speeds. Ultra-X has generously supplied their RAM stress tester, the R.S.T. Pro 3 PCI Express. There are only a couple of review sites that have one of these and we are proud to be one of them. It’s not your standard Memtest++ bootable CD. This is a standalone, bootable piece of hardware that plugs into a PCIe x1 slot and, man, does it ever beat on some RAM.
Several manufacturers use this to test their memory before it goes out the door to reduce RMAs. If it passes this test, it is stable and then some. Manufacturers generally run the full memory test for three passes. We’re overclockers though and tend to abuse sticks a little more than other people, so for our purposes five passes sounds good. Running this test on 8 GB of RAM ended up absolutely hammering the RAM for about three and a half hours with a load more that it will ever experience in every day use, so that should be sufficient to say whether or not it’s stable.
They passed with flying colors! If you prefer software testing, it was eight-threaded HyperPi 32M stable as well. As stressful as HyperPi is, it didn’t heat up the sticks nearly as much as the R.S.T. Pro 3.
At this point, we can deem these sticks stable and then some. This kit is rock solid at its rated speed and timings. The impressive part to me is that it took zero tweaking. Set the rated voltage of 1.65V, rated speed of DDR3-2666 and rated timings of 11-13-13-35 and you’re done! What’s even better is that it passed these stability tests with a 1T command rate.
Which brings us to our next topic…
Overclocking for 24/7 Use
Now, I usually don’t overclock for performance comparisons; but after seeing how easily it ran the rated speed and 1T command rate, I would have been remiss not to try stable overclocking. All that was necessary for this overclock was raising the memory multiplier. That’s it, no other changes from the settings above.
It doesn’t get much easier than that. Now, this was not HyperPi stable and I didn’t want to raise the voltage for 24/7 use. It was stable to run all benchmarks and Prime95 though, so results here are included in the performance graphs.
To test performance of this kit, we’ll run through our usual swath of 2D benchmarks. For pretty much all intents and purposes, 3D benching (except maybe 3DMark01 and 3DMark06, which are less GPU-dependent) and gaming aren’t going to benefit from this type of RAM. All DDR3-2133 and DDR3-2666 benchmarks were run three times each with the result average. Overclocked DDR3-2800 benchmarks were run once.
Note you can click the graphs below to make them larger, and thus more legible.
First up is what will show the largest difference – synthetic memory tests.
This is where you’ll see the biggest differences, none as large as the latency. While this memory will definitely move (read, write and copy) your data faster, it will get around to that action significantly faster, with latency dropping almost 17% from the DDR3-2133 kit.
Compression, Video Conversion and Rendering
In the real world, RAM speed doesn’t make a huge difference. It does make one, consistent and repeatable; but it’s not huge.
Across these benchmarks, the difference between DDR3-2133 and DDR3-2666 doesn’t amount to a hill of beans, accounting for less than a one percent difference across all benchmarks. Most of the runs favored the DDR3-2666 kit. The difference would be more pronounced if you were coming from DDR3-1600. Even moving to DDR3-2800 didn’t move things much farther.
This is where you’ll see the most pertinent differences people such as those on our benchmarking team will be looking for (SuperPi primarily).
WPrime doesn’t care one bit what RAM you use and that is borne out in testing. SuperPi shows its preference for faster RAM throughout, stepping up with each speed increase. If the timings were tighter, the difference would be even more palpable but it’s faster even with the loosened timings.
This is where things get interesting. When you feed this kit some voltage, it will show you what it’s capable of. Overclockers will just love this kit. Note you must have a strong IMC to take full advantage of it. Most Ivy Bridge chips have a capable IMC, but the occasional chip doesn’t.
First up, let’s see if you can tighten those timings at all. I went with an arbitrary voltage of 1.725V.
This kit tightened nicely, coming up with timings of 10-13-12-28 and 1T command rate. You need more voltage to tighten tRCD (the second timing), but the rest went down with only this small bump.
Next I chose just a little bump in vDIMM – to 1.67V, dropped the command rate to 2T and saw how fast the kit would run SuperPi 1M.
Wow, with just that little bump in voltage, they accomplished a very strong DDR3-2900 SuperPi 1M run!
Of course, never content to just stop at a round figure, maximum frequency was the next goal. I went higher this time, with 1.75V.
The kit’s (or the IMC’s, I’m not sure) maximum validated frequency ended up being DDR3-2945, very impressive by any measure!
Disclaimer – Running voltages like these run the risk of damaging your IMC and/or your RAM. If you choose to run these voltages, you do so at your own risk!
Final Thoughts & Conclusion
This kit is awesome. It is perfectly stable at its very highly rated speed, overclocks like mad and does so with stylish looks to boot. The only reason I couldn’t go higher or tighter seems to be the voltages I’m willing to push. There will be a stopping point of course; but I don’t think we’ve seen the ceiling for this kit.
Now, those who want a ‘normal’ system may find a good DDR3-1866 or DDR3-2133 kit with reasonably tight timings to be sufficient. You can see in the benchmarks your average computer user isn’t going to get a huge leap in performance with such powerful RAM.
No, this kit is tailor-made for overclockers, tweakers and benchmarkers – and it impresses; very much so. There’s a reason G.Skill is the most popular memory brand at HWBot and kits like this just add to that popularity.
So, what will this set you back? $199.99 at Newegg, which really isn’t bad. Frankly I thought they’d come at a higher premium. Again, Joe Stock System isn’t going to need these and would be perfectly fine with a slower kit. You could even save $100 and get their DDR3-2400 kit rated with slightly lower timings. What you’re paying for is the binning process and the rare chips within a kit that will run at blazing DDR3-2666 speeds, guaranteed for life (did I mention the lifetime warranty?). If you will use memory like this because you want the fastest RAM or you are a competitive benchmarker and need every second you can get or just because you like tweaking a superb set of RAM, this is the kit for you.