G.Skill Ripjaws 4 F4-3000C15Q-GRR 16GB Memory Kit Review

Memory by leeghoofd @ 2015-03-02

With the release of Intel's high end X99 platform aka Haswell-E last year DDR4 became available for the big masses. Besides the massive available bandwidth these DDR4 modules can also pack a far higher density per stick. All this goodness is available at your disposal while requiring a mere 1.2-1.35Volts. Only drawback, as with each new technology at launch is the high retail price, Things have settled a bit and prices are slowly plummeting. Today we have a look at one of G.Skill enthusiast kits: the Ripjaws 4 F4-3000C15Q quad channel kit. While they have even faster kits in their lineup, ranging up to a massive 3400MHz out of the box. Today's reviewed 3000C15 kit has been heralded by many enthusiasts worldwide. Time to take a closer look.


Unboxing

The G.Skill F4-3000C15Q 16GRR Quad Channel kit comes in a solid blister package, no fancy cardboard box and thus also no extra memory cooler included. No frills this time, not even with this high end specced kit. This has a positive effect on the price as this particular kit can be yours for around 285 Euros. Not bad as this is around the same price of the Corsair Vengeance LPX 2800C16 memory kit we reviewed end of last year, though G.Skill adds another 200MHz and tighter timings than the similar priced Corsair offering.

 

 

 

Take note that G.Skill has several 16GB Quad channel 3000C15 offerings with only a small differentiation in the naming:

  • F4-3000C15Q-16GRBB (Samsung IC based) C15-16-16-35 1.35V - Blue Heat spreader
  • F4-3000C15Q-16GRK (Hynix IC based) C15-15-15-35 1.35V - Black Heat spreader
  • F4-3000C15Q-16GRR (Hynix IC based) C15-15-15-35 1.35V - Red Heat spreader

 

 

No surprise to discover rebranded Hynix MFR ICs under the heat spreader. These were not so popular with the OCing community in the DDR3 variant, besides when hunting down massive clock speeds. In the DDR4 version they allow tighter timings and very good efficiency; behavior similar when overclocking to the oh so popular Samsung ICs on LGA1150 platform. Removal of the heat spreader requires some patience. A thin piece of adhesive tape holds the read heat spreader firmly in place.

 

XMP Profile and Gallery

The G.Skill F4-3000C15Q 16GRR features one XMP 2.0 profile:

 

  • 3000C15-15-15-35 Command Rate 2T at 1.35Volts.

 

Below is the extract of what is read out on the ASUS Rampage V Extreme motherboard after enabling the XMP profile in the BIOS

 

 

 

Looking at them timings there seems to be plenty of possibilities for manual tweaking :)

 

 

Test Setup and Methodology

 

 

MADSHRIMPS DDR4 TEST SETUP on LGA 2011-v3:

  • Intel Core i7-5960X ES with 3500 Uncore speed iso 3000MHz
  • Cooled by EK Waterblocks L360 kit
  • ASUS Rampage V Extreme motherboard 0801 bios
  • 16GB G.Skill F4-3000C15Q-GRR
  • AMD 7970HD & Catalyst 14.4 driver
  • Western Digital HDD 1TB Green Caviar
  • Corsair AX1200 PSU
  • Micrcool Banchetto 101
  • Windows 7 Professional 64Bit SP1 fully patched.

 

 

To compare the out of the box performance of these G.Skill Ripjaws 4, we included the most used speeds by the memory vendors: being 2133, 2400, 2666 and 3000MHz.

Just to give you a quick hands on, here's a small snippet from the Haswell-E launch article of how the DDR4 speeds relate to the DDR3 memory used with the previous Intel generations; of course this is comparing apples with oranges even though both the i7-4930K and i7-5960X were running with 6 six cores at 4000MHz. Nevertheless architectural improvements have their impact, though after quickly wading through the results one doesn't need to be fooled by the looser timings on the DDR4 memory kits. 2133C15 DDR4 can easily keep up with 2400C10 DDR3 memory speeds.

 

 


The Importance of Processor Uncore

While testing the new Haswell-E platform we observed that the high memory speeds were being bottle-necked by the 3000Mhz uncore processor speeds. This was clear while fetching the data for the Corsair Vengeance LPX kit, thus we decided to retest with an increased uncore speed from the stock 3000 to 3500MHz.

At the moment only two retail ASUS boards, both being equipped with their patented OCing socket, are able to push the processor uncore easily over 4000Mhz. All other boards are maxing out on air at 3500-3700ish speeds. For a daily user the latter are no real limitation, though for the efficiency tweakers at the moment there are only two viable options: the ASUS X99 Deluxe and the Republic of Gamers Rampage V Extreme. Gigabyte has introduced their brand new X99 SOC Champion board at CES 2015, though the board hasn't hit the retail shelves yet.

 

 

 

Back to the testing: We used the i7-5960X with the full 8 cores enabled. The DDR4 memory speed was limited at 2666Mhz C15-15-15-35 with a 2T Command Rate set.

 

 

We tested Uncore speed at 3000MHz, 3500MHz, 3750MHz and 4000MHz speeds. Any X99 board can handle the first two uncore speeds, however from 3600ish speeds the two aforementioned ASUS boards will have the edge and will even go above +4000MHz on air with a moderate voltage increase. (depending on the processor limit)

 

 

In SuperPi 32M we see a massive gain in performance by gradually increasing the uncore speed of the processor. Higher uncore speeds of 4500MHz were also possible though required upping the cache voltage. Fine for benchers though for those that want to max out their daily setup with safe voltages can easily achieve 4000MHz on the two former mentioned ASUS boards. Even with the memory just running at 2666MHz it beats the scores of the 3000MHz DDR4 setup with over 4 seconds, by just increasing the Uncore frequency.

Conclusion: when running high memory speeds, one needs a high uncore speed to lift the bottleneck and to maximize the performance of the hardware of your setup.

 

 

 

The AIDA64 Bandwidth test shows what really happens: not only do Read and Copy progress with higher uncore speeds, also the Write performance receives a significant boost, confirming our theory that indeed boosting the uncore speed is a must to achieve the best performance of your favorite Haswell-E processor.

 

 

 

While Cinebench R15 scaled very mildly with the increased memory speeds and not touching the uncore, boosting the uncore has a drastic effect on the performance in this benchmark; even in the synthetic 3D benchmarks a small gain is each time noticeable once we increase the uncore speeds.

 

 

Gaming wise there is hardly any influence with the standard uncore speed of 3000MHz or the overclocked one of 4000Mhz, alike the lack of increased memory frequency will have hardly any impact on your daily gaming experience.

 

 

 


Stock XMP 2D Results

First test is to run these G.Skill F4-3000C15Q-GRRmodules at their Xtreme Memory Profile (XMP) of 3000C15-15-35 with a Command Rate of 2T. To provide some other data we also included the following speeds in the charts:

  • 2133C15-15-15-35 2T Command Rate
  • 2400C15-15-15-35 2T Command Rate
  • 2666C15-15-15-35 2T Command Rate
  • 2800C16-18-18-35 2T Command Rate (Corsair Vengeance LPX)
  • 3000C15-15-15-35 2T Command Rate (G.Skill Ripjaws 4 3000C15)

 

It will be interesting to see how the similar priced G.Skill kit will perform versus the slower and looser timed Vengeance LPX 2800C16 kit from Corsair; the results differ slightly from those of the Corsair Vengeance LPX article, this because the CPU clock speed is now set at stock and the uncore raised from 3000 to 3500MHz.

First up SuperPI 32M, a single threaded benchmark that loves raw bandwidth: over 10 seconds are gained from the stock Intel rated 2133 frequency versus the XMP settings of the G.Skill F4-3000Q 16GRR kit. The Corsair kit seems to be in for an as whooping, logic as it features 200Mhz less in memory speed and has looser overall timings, yet it retails at the same price level as our G.Skill kit, thus the choice is pretty rapidly made between these two kits....

 

 

 

 

The AIDA64 Bandwidth and latency tests tell the tale. Thanks to the upped Uncore the G.Skill F4-3000 kit soars away from the competition. Breaching over 70K in the Copy test and outperforming with over 3K in the Read test; the AIDA64 latency test is also in favor of the faster G.Skill F4-3000 kit.

 

 

The Maxon Cinebench R15 64bit benchmark also scales mildly with extra memory speed, though benefits more from the Uncore than running your memory over 2666MHz. Interestingly we see a solid increase from 2133 to 2400MHz, though once above the scaling is less impressive.

 

 

 

The X264HD encoding benchmark shows that faster memory means more FPS can be rendered as the entire subsystem benefits from extra bandwidth. Intel's in house developed benchmarking tool eXtreme Tuning Utility also scales with extra memory performance.

 


Stock 3D XMP Profile

High speed memory has a lesser relevance in performance in games than when running 2D tests. One can only recommend these for benchmarking or for bragging rights. Futuremark's Firestrike scales mildly with increased memory frequency. Proper dialed in timings would yield a far higher score, though we test only at the XMP profile.

 

 

 

Catzilla behaves similar than Firestrike, mild scaling from 2133Mhz to 3000Mhz speeds, though nothing earth shattering here. The three tested games, tested at a 1920 x 1080 resolution with every detail setting maxed out hardly give a different output. Everything remains there within the margin of error... time to see if we can tighten up the timings or achieve a higher  frequency.

 

 

 

 

Overclocking

Time to raise the bar and start tweaking the timings. If the memory failed the windows memory test we upped the voltages to see if we can stabilize the memory... Below are the achieved results, pretty impressive stuff if I may say so from a retail kit :)

 

 

CAS 14: 3000C14-14-14-16 Command rate 1T at stock 1.35 volts

 


CAS 13: 3000C13-14-14-16 Command Rate 1T at 1.425 volts

 


CAS12: 3000C12-13-14-15 Command Rate 1T at 1.545 volts

 


CAS11: 3000C11-13-14-15 Command Rate 1T at 1.725 volts

 

 

Next we tried to raise the memory frequency and got a nice 3300MHz bench stable at 1.45Vdimm at the same timings as the stock XMP. Anything higher would fail, now it could be the Integrated Memory Controller of the CPU was reaching its limits and needs cold to remain stable, or our kit is just running at its highest peak.

 

3300C15-15-15-30 2T Command Rate at 1.45 Volts

 

 

Conclusion:

G.Skill is one of the brands that has build a serious reputation amongst the enthusiast crowd. Mainly while as they remain one of the more affordable brands, nevertheless always allow for a lot of tweaking aka overclock potential. This Ripjaws 4 F4-3000C15Q 16GRR kit is no exception to that rule. Checking the DDR4 stats at HWBOT.org we see that G.Skill is the dominating DDR4 memory vendor with over 2000 submissions. Most of them have been achieved with the quad channel kit we reviewed today.

 

 

Though are these a necessity for your build? If you want a solid and fast subsystem than these are great bang for the buck. For those that use memory bandwidth hungry applications these will be a perfect match as they provide a great balance between price, speed and timings. Tweakers and Benchers already found their favorite memory set in this red G.Skill F4-3000C15 Quad Channel kit. Gamers don't need to apply as speeds nor timings really matter for their gaming experience.  A 2133MHz kit will suit their needs perfectly fine.

 

 

Tweaking wise the F4-3000C15 kit has lived up to the high expectations we had from it. Depending on the motherboard you use, you can seriously push these sticks, even beyond higher rated and thus far more expensive retail kits. We have tried these stick on several high end X99 motherboards from ASUS, ASRock and Gigabyte and could run these easily at CAS 14 at stock volts. With added voltage CAS 12 was a no brainer, this at far lower voltages than with the Corsair LPX kit. Pushing even more resulted in an impressive CAS11 bench stable configuration.

G.Skill has a stellar performing kit on offer here, combining great performance at an affordable price. In fact the 280 euros are entirely justified looking at the numbers this kit pumps out. Another great addition to the ever growing G.Skill lineup, therefore recommended from the MadShrimps crew!!

 

 

 

 

PROS:

  • Hynix MFR ICs
  • Huge OC potential
  • Solid programmed SPD
  • Lifetime warranty
  • Great bang for the buck

 

CONS:

  • nothing too mention

 

 

 

 

We wish to thank Joseph from G.Skill for the F4-3000C15Q 16GRR sample

 

 

Sven from ASUS for providing us with the Rampage V Extreme Motherboard

 

 

Kristof and Griet from Intel for the i7-5960X HSW-E processor

  翻译：