Back in September of last year, G.Skill provided a Ripjaws4 DDR4-3000 MHz memory kit to compliment our Haswell-E launch day coverage. This time around, G.Skill sent along a Ripjaws4 DDR4-2666 MHz 16 GB kit for us to have a look at. This particular kit carries a part# F4-2666C15Q-16GRR and is priced substantially less than their DDR4-3000 MHz and above offerings. DDR4 memory is by nature a very expensive proposition, so we’re excited to see if G.Skill can provide a cost efficient DDR4 memory kit that still provides the performance we’ve come to expect from their memory products. Let’s go see how they did, shall we?
The specifications below are provided by G.Skill and provide detailed information on the kit at hand. You’ll notice the timings are set identical to the DDR4-3000 MHz kit we spoke of earlier, but the operating voltage is much lower at a mere 1.2 V. Full support for Intel’s XMP 2.0 standard, a lifetime warranty, and a very respectable memory height round out the highlights here.
G.Skill Ripjaws 4 DDR4-2666 4X4 GB Specifications
|DDR4 Unbuffered Non-ECC|
|2666 MHz (PC4-21300)|
|16 GB Total (4X4 GB Modules)|
||Compatible with Intel XMP 2.0|
|Intel X99 Chipset/Haswell-E (LGA 2011-3) CPU|
|40 mm/1.58 in.|
Once the memory is up and running with the XMP Profile set, here is what the UEFI BIOS and CPU-Z show.
The high level features are listed below and are also courtesy G.Skill.
- Extreme performance DDR4 memory designed for gaming and PC enthusiasts
- Optimized compatibility with Intel X99 system
- Stylish aluminum heat spreader design to keep the memory cool
- 1.2V ultra low voltage
- Intel XMP 2.0 Support
For a deeper look into the features the Ripjaws4 DDR4 memory line offers, the G.Skill marketing folks have several key points they’d like you to know about.
G.Skill has a long history of extreme overclocking and world records to prove it. I don’t think their current record was accomplished with the kit we are reviewing today, but it gives you an idea of the Ripjaws4 overclocking prowess.
One of the advantages the new DDR4 memory platform offers is its low power consumption. It’s really pretty astounding that memory running at these speeds can do so at such low voltage.
Optimizing CPU cooler compatibility and still being able to keep the modules cool is something G.Skill thinks they’ve accomplished with their new heatspreader design.
The memory kit comes packaged in a clear plastic clam shell with a cardboard insert. The insert has a few high-level features and specifications printed on it, and they stuff a nifty little case badge inside too.
Obviously, the kit we were sent has red heatspreaders. As we mentioned before, you can see the heatspreaders are not very tall and should accommodate a wide variety of air coolers. G.Skill’s logo and branding are on both sides of the modules, but one side has a sticker applied with basic specifications printed on it. The heatspreaders feel very well made, and you definitely get a feeling of quality when handling the memory modules. Enjoy the pictures below!
Testing for Stability at Rated Speed/Timings
Once the memory is installed in our test bed, we begin with a stability check at advertised speed/timings/voltage. To perform the stability check, we use our new Ultra-X R.S.T. Premium USB memory tester. The fine folks at Ultra-X provided us with the tester, and we’re proud to be one of only a few review sites that have one. The Ultra-X R.S.T. Premium is a bootable USB device that’s a relentless, memory pounding beast and will definitely find any instabilities a memory kit may have. Depending on the capacity of the memory kit, it can take several hours or more to complete the five passes we prefer to run. As you can see by the picture below, it took well over six hours to complete the five passes. No errors were recorded, so we’re confident we have a perfectly stable set of memory.
Testing for stability in-OS is performed with a 32M run of HyperPi, which is a pretty stout stability test in it’s own right. If you don’t posses a tester like the Ultra-X R.S.T Premium, it’s the next best thing. After passing the torture test that is the Ultra-X R.S.T. Premium, we didn’t expect a problem when running HyperPi. As you can see, no problems encountered here either.
Here are the components used in our test bed. All testing was performed with the CPU at its stock setting and the memory set to its XMP Profile.
|Test System Components|
|Motherboard||ASUS Rampave V Extreme|
|CPU||Intel i7 5960X|
|Memory||Various – See Table Below|
|Video Card||EVGA GTX 780 Ti Classified|
|Storage||Samsung 840 Pro SSD 256 GB|
|Operating System||Windows 7 Pro x64|
The four memory kits in our comparison group include today’s review sample and the previously reviewed G.Skill Ripjaws4, Corsair Vengeance LPX, and ADATA XPG Z1. The previously reviewed G.Skill and Corsair kits are 16 GB and run at DDR4-3000 and DDR4-2800 speeds respectively. The ADATA kit is 32 GB in capacity and has the slowest MHz speed of all the comparison samples at 2400 MHz. Both G.Skill kits were tested on the ASUS Rampage V Extreme, while the ADATA and Corsair kits were tested on the ASUS X99 Deluxe. Both motherboards used the memory’s XMP Profile with the CPU at it stock settings for all tests. Both of the motherboards also feature the ASUS OC Socket, so results should be consistent between the two motherboards used for testing.
|G.Skill||Ripjaws4||DDR4 3000||Quad||4×4 GB||15-15-15-35-2T||1.35 V|
|Corsair||Vengeance LPX||DDR4-2800||Quad||4X4 GB||16-18-18-36-2T||1.2 V|
|G.Skill||Ripjaws4||DDR4-2666||Quad||4X4 GB||15-15-15-35-2T||1.2 V|
|ADATA||XPG Z1||DDR4-2400||Quad||4X8 GB||16-16-16-39-2T||1.2 V|
The graphs below show percentage values with the G.Skill DDR4-2800 kit (today’s review sample) being the basis, and therefore always 100%. In the scored benchmarks, a higher percentage is better while timed benchmarks a lower percentage is better. Below each chart is a table with the raw data used to compile the chart.
The AIDA64 suite of memory tests show the G.Skill Ripjaws4 DDR4-2666 MHz kit landing right where it should. The latency test shows our review sample beating all but the G.Skill DDR4-3000 kit (lower percentage is better). The copy tests show our kit again beating all but the G.Skill DDR4-3000, while the read and write test results show very little difference between all the test samples.
|AIDA64 Memory Benchmarks – Raw Data|
|G.Skill Ripjaws4 DDR4-3000||61874||46994||67752||64.0|
|Corsair Vengeance LPX DDR4-2800||59046||47684||56632||69.8|
|G.Skill Ripjaws4 DDR4-2666||60378||46997||62783||68.9|
|ADATA XPG Z1 DDR4-24000||59793||47921||61091||69.8|
The SuperPi and wPrime testing show a similar pattern of all the samples being very close, but even more so here. All four of these test results show less than a 2% difference between the samples.
|SuperPi & wPrime Benchmarks – Raw Data|
||SuperPi 32M||wPrime 32M
|G.Skill Ripjaws4 DDR4-3000||10.358||535.128||3.588||103.693|
|Corsair Vengeance LPX DDR4-2800||10.281||529.356||3.527||101.728|
|G.Skill Ripjaws4 DDR4-2666||10.342||538.528||3.526||103.554|
|ADATA XPG Z1 DDR4-2400||10.296||543.115||3.525||103.586|
Rendering and Compression Testing
The three Cinebench tests again show minor differences between the sample kits with around 2% being the biggest difference. The 7zip test run shows the same behavior with less than a 2% difference between the test samples.
|Cinebench & 7zip Benchmarks – Raw Data|
|G.Skill Ripjaws4 DDR4-3000||44800||15.17||1412||43532|
|Corsair Vengeance LPX DDR4-2800||44206||15.49||1412||43640|
|G.Skill Ripjaws4 DDR4-2666||44333||15.26||1407||42945|
|ADATA XPG Z1 DDR4 2400||43761||15.26||1407||42945|
Video Conversion Benchmarks
Video conversion performance is tested with x264 and PoV Ray 3.73. Once again, around 2% separated all the comparison samples in these test runs.
|x264 & PoV Ray 3.73 Benchmarks – Raw Data|
|Kit||PoV Ray 3.73
||x264 Pass 1||x264 Pass 2|
|G.Skill Ripjaws4 DDR4-3000||2847.19||205.88||83.42|
|Corsair Vengeance LPX DDR4-2800||2903.74||208.17||84.80|
|G.Skill Ripjaws4 DDR4-2666||2845.25||205.88||83.43|
|ADATA XPG Z1 DDR4-2400||2849.11||204.98||83.20|
From top to bottom, the test results make one believe memory speed doesn’t play as big a part in total system performance as it might have in previous generation platforms. With that said, the G.Skill Ripjaws4 DDR4-2666 kit becomes quite a value when you consider it performs right on par with the DDR4-3000 and DDR4-2800 kits in our comparison samples. Hardly anything to complain about there.
When I overclock memory, I like to make sure it’s stable and something you can actually use day to day. Sure, you can probably ramp it up higher and complete a quick benchmark, but I prefer settings that can actually be used reliably. By setting the memory speed in the UEFI BIOS to 2750 MHz and then adjusting the BCLK to 127.3, we landed at 2800 MHz for the memory. This only took a slight bump in voltage to 1.3 V and we were able to keep the timings at their default 15-15-15-35. It passed a full 32M run of HyperPi at this setting, so we’re off to a good start!
Getting the memory to play nice at 3000 MHz required both an increase in voltage to 1.35 V and lowering the timings to 16-16-16-36, which isn’t too much of a sacrifice. HyperPi 32M completed yet again at these settings. Compared to the original run of HyperPi 32M when we were checking for stability at stock speeds, we see over a 45 second improvement in completion time.
Once past this point, the timings had to be raised to a point where they negated any increase in speed. Not to mention the voltage required was getting well over the 1.4 V ~1.45 V range. I did manage to get to the desktop at 3333 MHz using 17-22-22-36 timings, but it was anything but stable. Still, in the end, this kit runs happily at DDR4-3000 MHz with an acceptable set of timings and voltage. Let’s move on and see what we can do about tightening up the timings.
After dropping the memory back to its default speed of 2666 MHz, we set the timings to 13-14-14-35 and tried HyperPi 32M again. It took right at 1.4 V to get this stable, but we got through the SuperPi 32M run. I tried lowering the timings further and applying even more voltage, but unfortunately it wasn’t having any part of it. Even so, a CL13 isn’t a bad result.
G.Skill has done a very nice job with their entire Ripjaws4 line of memory, and this kit held true to form. For the average overclocker, you can easily get this kit to run at 2800 MHz or 3000 MHz. Taking it past those speeds is doable if you’re willing to throw a lot of voltage at it and have the ability to keep it cool. In the end, it’s a kit that’s has a substantial amount of headroom and is fun to play with. Even if you simply use the kit at its stock XMP profile settings, you’ll enjoy performance right on par with kits costing substantially more.
Aesthetically, the Ripjaws4 DDR4-2666 is a good looking affair, and we really like that G.Skill kept the memory height at a level conducive to using a wide variety of air cooling solutions.
The Ripjaws4 DDR4-2666 kit is currently available at Newegg for $249, which is the least expensive DDR4-2666 kit available there with 15-15-15-35 timings. Given how easy it is to overclock this kit to the 2800/3000 MHz range, the value is certainly pretty impressive.
A good looking kit that is priced right, overclocks very well, and offers enthusiast level performance… Overclockers approved!