It has been a while since we had a look at some RAM, so today we’re happy to bring you a solid set from G.Skill, their Ripjaws-X DDR3-2133 8 GB kit, model # F3-17000CL9D-4GBXL.
Packaging, Specifications and First Impressions
Like all G.Skill kits that include a fan, these come in a nondescript cardboard box. The top displays their design intent (to be used on LGA1155 systems) and the side shows the model number and specifications.
Opening the box, we are met with the first view of the rather sizable fan assembly. Everything is packaged well and will survive its journey to you without issue.
Exploring the fan assembly, it is very well made with a metal frame housing dual 50 mm, 0.8 W fans.
The fans are not absolutely silent at full tilt but don’t qualify as loud. Merely audible would be the best descriptor, and this is on an open bench station. Inside a case you’d never hear them.
The wire is sleeved, which is a nice touch. I’d have preferred a 3-pin header to the MOLEX connector though. It’s much easier to use a motherboard header for these things than run an extra MOLEX connector to them.
Pulling the RAM out, we are met with a nice looking set of shiny red sticks.
These aren’t the giant heatspreaders of recent memory. That’s ok, because in use (and even under heavy R.S.T. Pro 3 testing) they didn’t get hot. The fan is almost overkill with these guys. It doesn’t hurt at all but if you want a low profile set of RAM and can’t fit the fan, it’s nothing to stress.
All in all a great looking set. Two thumbs up for appearance.
Test Setup and Methodology
Due to a series of events that led to the initial baseline kit being passed on to another of our editors, there are two systems compared in this review. They are both based off the P67 chipset and differences between the two systems when CPU and RAM are set in identical configurations amount to a rounding error at best. The systems, side by side, were both used in our recent review of the ASRock P67 Extreme 6.
|Processor||i7 2600K||i7 2600K|
|Stock / Overclocked Speeds (GHz)||3.4 / 4.3||3.4 / 4.6|
|Motherboard||Intel DP67BG||ASRock P67 Extreme 6|
|RAM||Patriot DDR3-2400||G.Skill DDR3-2133|
|GPU (for total 3DMark Score Only)||ASUS Matrix 5870 Platinum||ASUS Matrix 5870 Platinum|
|Operating System||Windows 7 x64||Windows 7 x64|
The two boards are separated by a lot more than benchmarks at identical settings with their sizable feature set differences, but for this comparison testing against each other works well. Due to these similarities with the P67 Extreme 6 review, you will see some of the same numbers used here.
The results graphed in the performance section were obtained by operating the CPU at its stock speed (3.4 GHz) and the RAM at the speeds referenced above. Benchmarks were run three times and the results were averaged.
The graphs you will see are all normalized to the baseline DDR3-1600 / 8-8-8-24 results. The DDR3-1600 set is always 100% and results from the G.Skill Ripjaws kit are expressed and graphed as percentages relative to its results. The actual benchmark results are in parenthesis below the percentage.
Before performance testing, it’s always good to look at stability. 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. By the time the tests finish, the sticks can get quite toasty.
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. Usually I go with six passes, but with 8GB worth of RAM, five already took three hours, and three hours of absolutely hammering RAM more that it will ever experience in every day use should be sufficient.
They passed with flying colors. The only items changed in UEFI were manually setting the recommended voltage (1.65 V), rated speed (DDR3-2133) and first four timings (9-11-9-28). No other tweaking was necessary and they were solid as a rock. This is both a testament to the strength of these sticks and the strength of the Sandy Bridge IMC. We couldn’t ask for more!
We’ll start with 3D Benchmarking, then go to real-world comparison (compression & rendering) and then end with some 2D benchmarking.
Scores in 3DMark06 and 3DMark Vantage are both quite dependent on scores generated by benchmarking the CPU. RAM certainly will make a difference in these subtests; but which do they prefer – timings or MHz? Both is obviously preferable, but what if you only have one or the other?
Interestingly, the largest score difference on the graph – 3DMark06 total score – prefers the faster MHz with loosened timings; but the CPU test itself seems to prefer timings to MHz. Vantage prefers the faster kit in both instances. It seems this kit would be well suited to gamers trying to squeeze out every last FPS.
Rendering & Compression
These results are important to every computer user, especially 7zip. Everyone compresses or decompresses files at some point and anything that goes toward making that faster is a win. For anyone that renders regularly, they know every second shaved off those tasks means increased productivity overall.
Both 7zip and Cinebench R11.5 show a gain with the Ripjaws kit. Cinebench R10 shows a slight loss, but take that with a grain of salt. In such tight competition, R10’s well known score variations can account for any differences in either direction. R11.5 and 7zip are both more consistent with their results and both tip their hat to the Ripjaws.
Wprime historically doesn’t really care about RAM. I know several benchmarkers that run it with just one stick installed and (within reason) it doesn’t seem to care how fast that stick runs.
It does seem to make a slight difference here, especially in WPrime 32M, one of the shortest benchmarks in our suite today; it seems to prefer timings to Mhz. The much longer test, WPrime 1024M, definitely doesn’t seem to mind either way.
SuperPi and PiFast are where we see the biggest benefit to changing RAM most of the time. SuperPi 32M is especially attuned to those changes.
Across the board in these RAM-sensitive benchmarks, the faster memory shows its muscle even with loosened timings. While 2.72% doesn’t seem like much in SuperPi 32M, with such a long benchmark you’re talking about 15 seconds, which can easily mean the difference between first place and 50th.
Synthetic Bandwidth and Latency
Of course, what memory review is complete without showing latency and bandwidth numbers?
Again, the higher MHz kit shows strong gains across the board, increasing write bandwith by 877 MB/s, copy bandwith by 1852 MB/s and read bandwidth by a whopping 2570 MB/s. Latency was also reduced by 4.2 ns, an 8% improvement.
Final Thoughts & Conclusion
In past memory reviews, there hasn’t been a massive difference in performance from one set to the next. You need a combination of higher MHz and lower timings to equate to a large performance increase. That said, we’re dealing with 8GB of memory here. Before recently it was unheard of to have high density sticks operating at such blazing speeds with solid timings.
You probably noticed there was no overclocking section. Unfotunately the motherboard I had does not have a UEFI that plays well with the combination of the DDR3-2133 memory multiplier and BCLK increases. It’s not the RAM either; I tried it with a kit of DDR3-2400 (the G.Skill Pi kit reviewed here) to no avail. I chalk this up to a new platform with a still relatively young UEFI.
Even if there were an overclocking section, it couldn’t go all that far. The highest I’ve been able to push this CPU was to 106 BCLK (different UEFI with a lower memory multiplier), so you’re looking at roughly DDR3-2225’ish max. Regardless, straight-up benchmarkers that push memory’s limits aren’t going to be using a 2 x 4GB kit to show off high frequencies.
These sticks are going to be used by ‘normal’ folks on a Sandy Bridge system. They will either run the system stock or push its overclock by simply raising the multiplier. This kit will run at its rated speed very well and have certainly shown an advantage over a 2 x 2GB kit running at DDR3-1600 / 8-8-8-24. Add that to the fact that you’re looking at a sizable 8GB of system memory and this is a great product for every day use.
The G.Skill Ripjaws DDR3-2133 / 9-11-9-28 8GB kit retails for $199.99 at Newegg, which isn’t too bad considering the speed combined with a two-stick, eight-gigabyte kit. The only other 8GB kit listed with these speeds is one from Kingston that comes in a 4 x 2G configuarion (which can be harder to stabilize) and it’s $8.01 more expensive, making this G.Skill kit the clear price winner.
The absolute, unquestionable stability at such fast speeds for an 8GB kit, plus the performance gains seen here, when combined with such a strong overclocking platform in Sandy Bridge definitely adds up to make this kit Overclockers Approved.
-Jeremy Vaughan (hokiealumnus)