G.Skill Ripjaws Z 2133CL9Q-16GBZH Memory Kit Review

Memory by leeghoofd @ 2012-02-02

G.Skill is one of the more prominent market leaders in the RAM department. Offering a zillion of different ram combinations in all sorts of quantities, timings, speeds and even you can pick between different heatspreader colors. With the launch of the SandyBridge-E platform, we received one of the higher end 16Gb quad channel kits, running at 2133Mhz. Timings are the following : CL9-11-10-28. Maybe you are not impressed, but remember this is a quad channel comprising of 4Gb dimms. And best of all the price isn't lethal at all. The X79 ram line is baptized RipjawsZ. The X lineup for the older Sandy Bridges has been renown for being potent overclockers. But most reviewed kits were only 2GB DIMM versions. So being a lot easier to push. Let's have a look if we can get some more juice out of these dimms.

The Kit

With each new ram line, GSkill automatically coincides it's launch with a different heatspreader look. The look has been refined, if we may consider this as an evolution from the X series. A solid blister packaging protects your precious dimms from any outside world abuse.

 

 

  

 

The new heatpsreader is a low profile version. With X79 and if running quad channel, it's required to plug in dimms on the left and right hand side of the CPU socket. Most big CPU air coolers are having clearance issues with older, read bigger sized dimm heatspreaders. A good strategy from the GSkill design team. As other vendors still are using the bigger heatspreader versions, wich would lead to a lot of installation woes.

 

 

We popped the heatspreader and see decent contact between the ICs and the conductive tape. Since we are not allowed to show closeups of the ICs from our contact person, we will strictly adhere to her wishes. Let's stick to the fact that most X79 Quad channel kits are Hynix IC based.

 

 

 

As mentioned in the introduction, GSkill is reknown for varied lineups. Sometimes confusing a lot of buyers, but at least they got a ram kit for everyone. You name it, they got it !

 

System Setup

Since this is a X79 XMP1.3 ready kit, we would have liked to maximise the ram on the already reknown Asus Rampage 4 Extreme. However Asus had sent us the Sabretooth version instead, still giving us an abundance of ram features to tune and tweak. But we are wondering if a more advanced bios/board could have unleashed some more Mhz.

 

  • Intel 3960X ES CPU @4500Mhz
  • Asus X79 Sabretooth motherboard 0709/0907 bios
  • Asus GTX480 GPU
  • Western Digital 1Tb Caviar Green HDD
  • CPU cooled by Corsair Hydro 100
  • Corsair TX850 PSU
  • Windows 7 Pro 64Bit SP1

 

 

 

 

We did the entire [M]Test suite with the CPU running at 4500Mhz at 1.37Vcore

The ram speeds tested are for ->

100 Bclock x 45 CPU Multiplier with rams at :

  • 1333 C9-9-9-24     1.55 Vdimm
  • 1600 C9-9-9-24     1.55 Vdimm
  • 2133 C9-11-10-27  1.65 Vdimm ( stock rated setting )
  • 2400 C10-12-11-28   1.685 Vdimm

 

125 Bclock x 36 CPU Multiplier with rams at :

  • 2333 C10-11-10-27   1.71Vdimm

All the above frequencies are Memtest Dlx stable for 200%, we could run our test suite with 2133C9-11-9-27. But Memtest would error out in a matter of minutes. So we dropped them results as our sample ram kit wasn't stable enough.

 

Test Results

As usual, we start off with SuperPi 32M. Very reknown amongst benchers and a terrible good scaler with increased ram bandwith. An ideal test to show off ram potential. How extra ram speed performs in daily performance that's something else .

 

 

The time gained when going from 1333Mhz to 1600Mhz is a whopping 11secs faster in calculating Superpi32M. Going from 1600 to 2133 we gain another 9 secs. For the die hards, faster is even better. 2400mhz and beyond seems the way to go if SuperPi is your thing.

 

 

SuperPi 32M is single threaded, meaning only one core is loaded. When we run a SuperPi on all the cores (12) the differences are not that pronounced. 2 secs from 1333 to 1600Mhz. And a similar gain when running 2133Mhz. Once above the gains get very tiny, same as we saw with the Superpi32M. Only worthwhile for those that seek the best performance.

 

 

AIDA64 reveals what happens with the bandwith once we up the ram speeds. A solid boost at 1600mhz. At 2133Mhz we seem to hit a sweet spot as only the read results still seem to scale with extra speed. The Copy and Write output are similar to the far lower clocked 2133Mhz run. The below Cinebench Release 10 test scales hardly with extra bandwith, it's all about raw CPU power there.

 

 

The newer Cinebench release 11 version, gives the same outcome as the Pi and AIDA tests. Scaling till 2133Mhz, beyond it all stays sort of put.

 

 

Time for some encoding. We have tested with the 4.0 version of the X264HD encoding test. Surprise surprise, nice increase in rendered FPS till 2133Mhz. Then it all sorta comes to a halt.

 

 

We tested both Futuremarks PCMark05 and 07 edition. While the 05 still goes up with increased ram speed. The newer 07 version isn't too much impressed by the ram clocks we threw at it.

 

 

Close to no gains in 3dmark06. In fact very close to the margin of error.

 

 

Futuremarks 3DMark 11 is a ram lover. The more ram speed you give it, the higher the PhysX score will be. Great for tests, yet the final outcome isn't that spectacularly different between 1333 and 2400mhz.

 

 

Last but not least we tried a few games. Expecially FarCry 2 loves ram speed, hence why we included it in our test suite. Though nothing spectacular to watch. Better OC your GPU iso your Systems rams !

 

Some More Fun

We tried to push with the stock rated timings of 9-11-10-27, but at 1T command rate. We walled at 2252Mhz, Memtest stable. We reqired just 1.67Vdimm, but pumping more volts didn't help at. We pushed up to 1.75Vdimm, no go.

 

 

 

Loosening Command Rate to 2T and some timing tweaking we walled, again, at 2282Mhz. To go higher slack  timings is the way to go. 2400Mhz CL10-12-10-30 1T is bench possible. But to be 24/7 stable with 2400Mhz at 1.67Vdimm, TRP adjustment is a must, so settle then with 10-12-11-30. If you don't want to go over 1.65Vdimm, opt for 11-12-11-30 instead.

 

 

 

 

 

With 1.69Vdimm Memtest stable at 2460Mhz Cl10-12-11-30...

 

 

We did some testing on the X58 platform and got a rock stable 2400Mhz CL10-12-11-30 2T stable. Impressive for a 4gb dimm.

 

 

 

 

Final Conclusion

When I received the Gskill 2133C9 RipjawsZ kit , I wasn't too overwhelmed by it's rated out of the box specifications. Felt more like a déja vu. Secondly testing with some X79 boards and their mediocre performing biosses wasn't a fun ride at all. Luckily ASUS picked up the pace with improved biosses, to unlock some more ram potential. Even though the Sabretooth X79 board is more of a mainstream board, yet with plenty settings to fiddle with, it surprised me how easy it is to get these GSkill rams running rock stable at 2400Mhz.

 

 

 

For this particular kit, we got up to 2250Mhz at the stock rated timings with a slight bump in voltage ( 1.67 ). The quest for more Mhz, had to be achieved via slackening the RAM's timings. If you are a timing purist you can run these rams at 2333Mhz CL10-11-10-27 1T at 1.71Vdimm. But there's no real gain over the much easier to stabilise 2400Mhz CL10-12-11-28. We capped out, in fact our CPU's IMC capped out, at 2460Mhz. But looking at the test results, the out of the box speed of 2133Mhz is a good balance between ram MHz and timings. Plus you can undervolt the rams at their stock speeds to about 1.56Vdimm. Like concluded in the SB-E review, running 1600mhz ram speed is not a necessity, but you are cutting back a bit on performance if you will use 1600 or lower ram speeds. The sweetspot for most X79 users, that look for some extra punch seems to be at 2133Mhz. These RipjawsZ can even go faster, though the gain is hardly worthwhile the extra volts and timing tweaking. But for them that seek a bit more boost then the average joe and still want to obtain it with a high quantity kit, these GSkill RipjawsZ 2133C9's are a perfect combo. Typically for GSkill they spoil us with a very strong lineup with the RipjawsZ.

 

 

Too bad we couldn't stabilise the same timings as the more expensive and rare 9-11-9-27 kit. That would have really been a blast, besting performance of a kit costing trice as much. The price for these black beauties is around 142 euros. If you consider the cost of a 1600mhz 16Gb kit retailing around 85 euro's this does not seem cheap. But believe me the 2133C9's are at a very competitive price level, compared to similar 2133Mhz offerings from the competitors. The latter sometimes easily retailing at 170 and more euros.

 

 

For their excellent balance between the rated speed and timings we award the Gskill RipjawsZ the Madshrimps Performance award. The OC potential is present, with no issues running these rams on the X58 and X79 platform at 2400mhz. In fact it was two times the CPU's IMC that held us back. If you are in for a 16Gb kit and want to have that little extra under the hood, this GSkill RipjawsZ 2133C9 kit is highly recommended !

 

 

PROS :

  • Perfect balanced ram set
  • Compatibility with other platform
  • Strong OC potential
  • Box rated setting at 1.56Vdimm
  • Acceptable price level ( 142 euro )

 

 

 

CONS :

  • No 9-11-9 24/7 possible

 

A big thanks to Mia and Frank from GSkill for the RipjawsZ kit and the patience they had with us. As I always tend to be kindda slow. Big thank you girl !

 

 

And to Rogier from Asus for the Sabretooth X79 sample board.

 

And the boys from Tones webshop for the 3960X CPU

 

  翻译: