G.Skill ECO DDR3 Memory Kit Review

In the industry of technology, brands offering similar products without innovating will not be easily adopted by the enthusiast market. If we talk about memory, then the mayor factors are frequency, latency and of course... price. Intel specifications for Core i3/i5/i7 products recommend using a maximum DRAM voltage of 1.65v. While most brands try to fit on Intel´s limits on the high side, G.Skill is trying something different. The G.Skill ECO F3 10666CL7D 4GB Dual Channel Kit is intended to run at 1333MHz with barely 1.35 volts benefiting both high-end enthusiasts who want better Overclocks and lower temps, or simply those who want to build an HTPC an every shaved watt can be the difference. Benchmark Reviews will test the G.Skill ECO kit to check if this kind of memory is able to do that difference (considering that frequency and latency are not a major factor on Intel chipsets anymore).

G.Skill designed the ECO series for LGA1156 platforms. While they say they work the best on i5/i7 CPUs with P55 platforms, that´s probably because the H55/H57 chipsets came out just after this kit. The ECOs F3 10666CL7D are meant to run at 1333MHz with 7-7-7-21 2T timings with very low voltage. G.Skill also has the same kit at higher timings or the same kit at higher frequencies (1600MHz to be clear). Benchmark Reviews will test both modules to see how good they perform, how much they can Overclock for enthusiasts, and of course, there will be a time to prove if voltage reduction can really stand up with the ECO name.

Considering the low voltages used on this kit, it could be an excellent choice for Overclockers. With their price being a little bit more expensive than 100 USD the ECO F310666CL7D is definitively an interesting choice for most of the people with a P55/H55/H57 platform. Let´s have a look at the blister:

The package contains two 2GB RAM modules with a grey heatsink and the ECO logo on both sides. G.Skill recommends using 1.35 volts but don´t worry, they added some SPDs to boot at 1.5 volts in case your motherboard doesn´t support low voltages, thus meaning that the memory won´t get damaged with common voltages ranges (1.5v-1.65v).

About G. Skill

Established in 1989 by enthusiast, is a leading memory manufacturer based in Taipei, Taiwan. The company's top priority is Quality. All of our products go through a series of the most rigorous tests and strict quality control processes. In addition to commissioning qualified IC testing houses to test our products, they are hand test 100% twice in factory and office, to ensure the highest product yield and quality.

G. SKILL strives to achieve the highest and most advanced quality from the initial design, through manufacturing solder-paste printing, through surface mounting, to on-line visual inspection, system compatibility testing, packaging and finally to safely and reliably shipping our products to customers.

G.Skill operates a very active support forum for all their products, where their product specialists are quick to respond to issues, real or imagined. They also monitor major e-tail sites for customers having trouble with their products, and try to help them get on the right path towards resolution of their problem.

G.Skill ECO F3 10666CL7D Specifications

Apart from the obvious specifications, G.Skill recommends an Intel chipset P55 based motherboard and 1.35v for stable operation. G.Skill also offers full lifetime warranty giving an extra value to your product. Let´s have a look at the specs:

• Main board: Intel
• System: Desktop
• System Type: DDR3
• M/B Chipset: Intel P55
• CAS Latency: 7-7-7-21-2N
• Capacity: 4GB (2GBx2)
• Speed: DDR3-1333 (PC3 10666)
• Test Voltage: 1.35 Volts
• Registered/Unbuffered: Unbuffered
• Error Checking: Non-ECC
• Type: 240-pin DIMM
• Warranty: Lifetime

*G.Skill guarantees 2 Dimms dual channel operation would reach announced specification.

Closer Look: G.Skill ECO F3 10666CL7D

Leaving aside the simple blisters included in (almost) all memory kits, the G.Skill ECO F3 10666CL7D DIMMs offer great aesthetics. While the PCB is kept in green color, they added a silver heatsink with the brand´s logo and some flamed-shaped figures on it.

The heatsinks are everything but big. G.Skill kept a very sleek and slim design which ensures you won´t have problems with bigger heatsinks for other motherboard components. They barely increase the size of the DIMM and will also help you protect the memory from ESDs. Based in our tests, this kit is one of the coldest Benchmark Reviews has ever tried.

The sticker shows the memory model and the voltage used. Don´t ever remove those stickers if you want to keep your lifetime warranty safe.

G.Skill heatsinks are glued to the Integrated Circuits of the memory. This (sadly) means that we aren´t able to remove them easily to check the brand behind those low voltage chips. If we wanted to take them away, we would have to heat them enough and then, separate that little clip at the top of the heatsink.

Many high-performance memory kits for Core i5 platforms are Elpida based. While it seems G.Skill is shipping different ICs for this model we are not sure how will it OC. They could be Elpida BBSE (low-voltage binned, of course), 1.35v Samsung rated ICs, or who knows, maybe even a Micron based chip. Benchmark Reviews will ignore the fact of the unknown IC and test how well can it scale at stock voltages and up to 1.65v not to pass Intel´s limits.

G.Skill ECO 4GB SPD

CPU-Z shows 3 different JEDECs and the Xtreme Memory Profile for Intel Platforms. The XMP rates the memory at 667MHz (1333MHz DDR) with 7-7-7-23 latencies. Note the command rate is specified at 2T, but most likely support 1T as well. There´s always a JEDEC rated at 1.5v which ensures it will be compatible with any motherboard but you will lose the low-voltage ability. Those JEDECs are rated at 1036MHz, 1184MHz and 1333MHz at different latencies.

Just to verify those weird JEDEC frequencies were real we launched CPU Tweaker. This application gives us a wider view of the timings but the frequencies remain the same, so there´s no mistake. Hopefully, not supported low-voltage motherboards will use the 1333MHz 9-9-9 JEDEC.

Before start testing, we wanted to make sure the kit was able to do 1333MHz 7-7-7 at 1.35v. Memtest was good enough to let us know the kit was working flawlessly.

Memtest 4.0 has been running for 2 hours testing all unused RAM. At 112% coverage it doesn´t show any errors yet.

Testing Methodology: DDR3 RAM

At Benchmark Reviews we need to test memory without adding any other factors to the results. While testing stock speeds is usually easy, it only gets more difficult when we want to add overclocked results. We need to make some changes to the BIOS values and test each configuration to ensure maximum stability. First, I set the values on the BIOS and run Memtest to check if the kit is stable or not. After that, I test with OCCT Perestroika to check if the CPU is stable along with the whole system.

Finally, Benchmark Reviews decides which frequencies are going to be shown in the graphics in order to leave the CPU factor outside of the equation. This way you´ll probably see small differences between memory frequencies, latencies and voltage reflected on our charts without the CPU interfering in the results. Benchmark Reviews uses the following tests suites and applications:

• Lavalys Everest Ultimate Edition v5.3
• Passmark Memory Test v7.0 x64
• SiSoft Sandra Lite 2010
• Crysis Benchmark
• WinRAR 3.9 x64 Benchmark
• Retouch Artists Adobe Photoshop CS4 Script

Test System

• Motherboard: ECS H55H-I Mini ITX
• System Memory: 2x2GB G.Skill ECO F3 10666CL7D
• Processor: Intel Core i3 530 with stock heatsink
• Video: ATI HD 5850 @ stock (only for Crysis Benchmark)
• Disk Drive 1: Seagate Momentus 320 GB 5400 RPM
• Enclosure: Cooler Master Elite 100
• PSU: CM Flex ATX 150 watts PFC
• Operating System: Windows 7 Ultimate x64

As you can see, I tested the G.Skill ECO 4GB kit in a HTPC-like PC. Since the ECOs were done to achieve maximum efficiency and lower temperatures I thought it was a good way to test them. The CM Elite 100 doesn´t have any fans or great airflow, so it was a good way to see if this kit can work in small form factor cases and stay stable while doing some overclock. This configuration will reflect bigger results on power consumption since the system won´t consume as much as a high-end PC.

While we used a H55 based motherboard, as long as we enabled the XMP the motherboard worked at 1.35v without any problems. Unluckily, we could not raise voltage from this point, so either we worked at 1.35v or we disabled the XMP to start at 1.5v. Following this path, I had no other choice but testing different frequencies, latencies, and power consumption at 1.35v, 1.5v at finally 1.6v. A deeper explanation about the Overclocked results will appear on the next pages. The G.Skill ECO setup was the next for the benchmark and applications performance tests:

• 1333MHz 6-6-6-21 1T 1.35v
• 1333MHz 7-7-7-21-1T 1.35v
• 1600MHz 6-7-6-21 1T 1.5v
• 1600MHz 6-8-6-21 1T 1.5v

Benchmark Tests Results

Benchmark Reviews runs each tests at least 3 times to avoid special situations where scores are over/under exposed. Passmark Performance Memory Test does some bench to the memory and reports a test score. I´ve chose Memory Mark along with uncached read and write speed since they´re the most important benchmarks for memory testing on this suite.

Increasing frequency up to 1600MHz shows a little advantage on the final results. In this case, 1600MHz 6-7-6-21 is the best configuration for maximum performance, and that will probably be the best configuration for all the tests done today since it combines decent high frequency with relatively low timings.

EVEREST Ultimate Edition offers three simple memory bandwidth tests that focus on the basics; Read, Write, and Copy. In order to avoid concurrent threads competing over system memory bandwidth, the Memory benchmarks utilize only one processor core and one thread.

Again, Lavalys Everest results increase with higher frequencies and tighter latencies. There is almost 12% performance increase with 1600MHz in each test. Considering how easy is to achieve this configuration (and probably it can be done with lower voltage in the right motherboard) it is definitively the best way to use your G.Skill ECOs.

Sandra is based on STREAM, a popular memory bandwidth benchmark that has been used on personal computers to super computers. It measures sustained memory bandwidth not burst or peak. Therefore, the results may be lower than those of other benchmarks. STREAM 2.0 uses static data (about 12M) - Sandra uses dynamic data (around 40-60% of physical system RAM). This means that on computers with fast memory Sandra may yield lower results than STREAM. It's not feasible to make Sandra use static RAM - since Sandra is much more than a benchmark, thus it would needlessly use memory.

A major difference is that Sandra's algorithm is multi-threaded on SMP/SMT systems. This works by splitting the arrays and letting each thread work on its own bit. Sandra creates a thread for each CPU in the system and assigns each thread to an individual CPU. Another difference is the aggressive use of scheduling/overlapping of instructions in order to maximize memory throughput even on "slower" processors. The loops should always be memory bound rather than CPU bound on all modern processors.

Sandra shows up to 40% performance increase when overclocking up to 1600MHz. The difference gained by tightening memory latencies makes little to no change in this benchmark. From all these Benchmark results, the conclusion is simple: 1600MHz 6-7-6-21 1T will give you the best performance sacrificing 150 mili-volts. While I wasn´t able to test voltages between 1.35v and 1.5v, the G.Skill F3 10666CL7D were very stable at this setup, so they might be able to operate at 1600MHz and tight latencies with lower voltage... who knows?

Application Performance Results

Crysis needs no introduction at Benchmark Reviews. Try playing this 2 years old game with the latest GPUs available without getting 60 FPS as you desire. That's enough to make us choose this game for the tests. Any high-end GPU will scream at high resolutions with this game, which means it is very high GPU dependant.

Crysis Benchmark is a very special case where 99% performance relays on your GPU. This means RAM and CPU overclocking won´t make noticeable differences on final performance. For example, in this case, going from 1333MHz to 1600MHz barely gave more than 1 FPS gain. The reason I choose this game is simple: latest games will become more and more GPU-dependant, so, testing old CPU dependant games won´t show the trend for future games as this will do. Of course, if you want to improve your gaming experience, you´ll need to change your GPU, and leave your CPU and RAM alone as long as they are up to the task and not a bottleneck for the rest of your PC.

Adobe Photoshop CS4 is our new contender for the memory suite. This program can benefit from both CPU and memory, and additionally, it supports Nvidia CUDA and ATI Stream. Retouch Artists Test does a list of basic processes into an image similar to what someone would do to edit their vacation photos. There are some image resizing, duplicated layers, colors levels and image adjustments involved on the test. Notice the more MHz, the faster performance. But again, the gain is so small that I can´t say the memory is worth it enough. Maybe we should try some harder tests so that we can see if the gain is bigger there.

While this is not a formal Benchmark (basically because we need to count the time with an external clock), it reports a little gain when using 1600MHz. From the worst to the best result, there is less than 1 second difference, which means overclocking won´t do a lot here. Now imagine those numbers reflect hours instead of seconds and you will definitively see a change, but still not worth sacrificing power consumption and temperatures.

WinRAR is a very simple, yet useful application. How many times do you compress or decompress random files in a day? Well, WinRAR has its own integrated benchmark which measures an average of KB/s compression. This application supports multi-threading, 64 bits OS, and it´s very CPU dependant. There is a 5% performance improvement using 1600MHz compared to the worst result. Not really a mayor change, but it´s still there.

Summarizing the application performance results, we can see that overclocking RAM isn´t a big deal nowadays, and I have not mentioned Core i7 results, where CPUs with bigger caches definitively ignore RAM frequency and latencies. In fact, for application tests the best configuration is 1333MHz 6-6-6 with barely 1.35v. That will make your whole PC run cooler and shave some watts from your overall power consumption.

Overclocking Results

Some of you readers probably were thinking why I would use such weird settings for performance tests. I need to say that the G.Skill ECOs are one of the most interesting kits I´ve ever tested. This kit simply doesn´t care about the CL or tRP settings. It´s all about tRCD and that´s how Benchmark Reviews chose the settings. Since I did not knew which ICs were beyond the hood all I had to do was to test every setting possible and compile all my results. This quickly lead to me to discover the key for overclocking this kit was simply raising the tRCD while keeping the rest of the timings as low as possible. Let´s have a look to the chart:

Starting with 6-6-6 timings, we could achieve 1260MHz at 1.35v. Raising vDIMM to 1.5v gave us 1333MHz and finally 1.6v didn´t add that much. The G.Skill ECO kit doesn´t really shine with high frequencies but it does a pretty good job on low latencies. The kit was even able to work at 1333MHz 6-6-5-21 Timings!

After that, we moved to 6-7-6-21 1T and this configuration is really where this kit rocks. Reaching 1333MHz was easy with barely 1.35 volts (while the kit is rated at 7-7-7-21 2T with the same voltage). Raising vDIMM up to 1.5v gave us 1600MHz and finally 1.6v didn´t help that much. Please notice that 1600MHz 6-7-6-21 1T was the best configuration for all our performance results, so, even for motherboards which doesn´t support low-vDIMM voltages this kit should be a nice pair for your processor with low timings and just 1.5 volts.

Finally, rising latencies to 6-8-6 let us start at 1600MHz with barely 1.35 volts! This one could be a nice setup for low-vDIMM motherboards since you´re paying for a 1333MHz kit. This time we could only achieve 1730MHz but that´s almost 400 extra Megahertz from default. While Benchmark Reviews tested 6-9-6 settings they didn´t helped anywhere on frequencies, so we just decided to leave that setup out of the chart.

Please keep in mind that while all memory kits are different, some kits could do better or worse than this, but considering the G.Skill F3 10666CL7D kit was done for mid/high-end P55 motherboards paired with i5-i7 processors, there are big chances on achieving higher frequencies than I got, especially considering my setup was based on a low-end H55+Intel Core i3. Now, before you start asking why I chose this setup, remember I wanted to show you how good could this kit be for HTPCs, where you use very "basic" components and the airflow and heat dissipation is very limited because of the small form factor cases. Also, this way I can show you the advantages of power consumption in a better way rather than utilizing a system that consumes 150 watts or more on idle mode. Let´s head to the Power Consumption paragraph to see the results.

G.Skill ECO F3 10666CL7D kit working at 1600MHz 6-7-6-21 1T with 1.5 volts. Nice setup indeed!

Power Consumption Results

As I´ve mentioned before, the reason to try this memory on a low-end H55 motherboard paired with the Intel Core i3 530 CPU was to achieve a very low but still powerful machine. Using a 150 watts PSU would be much more efficient than pairing this with a 700 watts high-end PSU. Also, this way we will analyze power consumption with different voltages and the results will have a bigger gap-percentage compared to a high-end system. All the results were obtained with a P3 International Kill-a-watt and the consumption was measured directly on the plug of the PSU, which means the whole system consumption is being measured (except for the monitor or any external peripherals).

At idle tests, using 1.35v against the 1.5-1.65 volts recommended by Intel gives us close to 3 watts less. In the best case there was a 4 watts difference between 1.35v and 1.65v. Maybe this sounds like a joke for some of you, but considering this is almost 10% from total system consumption I wouldn´t say this is irrelevant. Adding some Load with Prime 95 shows a bigger gap between different voltages. The 300 mili-Volts difference is actually doing something there. In the best case, there is a 9 watts difference between 1.35v and 1.65v. Again, 9 watts could sound like nothing, but in this system it represents 10% of the power consumption, so if you really want to achieve a low power consumption system, low-voltage memory kits like this are something you must definitively consider, and hey, they are cheap and clock very well with low latencies. What else could you ask for? Maybe they would even work with lower voltages if your motherboard allows you to do so (some motherboards offer 1.2v and higher for vDIMM operation).

If you have a high-end PC and you don´t care about 5-10 watts difference, don´t forget this kit is also good when overclocked and it´s cheaper than many kits on the market. Also, there are high probabilities that yours will achieve higher frequencies on a high-end P55 motherboard with a Core i5/Core i7 CPU.

G.Skill ECO DDR3 Final Thoughts

I really don´t have a lot to say here. This kit has impressed me in many aspects and I think it´s worth getting it. The reasons are simple: the memory is made by G.Skill and it has a great reputation behind. You also receive a nice, sleek looking heatsink which works great specially since the memory works at low voltages, and finally, the ECO F3 10666CL7D overclocks very well and will outperform many kits rated at the same price point. The G.Skill ECOs fit on low-end machines like the HTPC tested today, but also represent a great choice for enthusiast and overclockers. As if that wasn´t all, you will save some watts with a less stressed PC (using low voltages) and you have lifetime warranty with G.Skill, which also has a great support team on their forums to answer every question you have. I really can´t ask for anything else. This kit just rocks! I ´m very tempted to try it on a high-end P55 motherboard and pair it with a Lynnfield processor to check if the MHz barrier was related to the motherboard and not RAM, so expect a little update on the forum soon!

Let´s summarize all the results and head to the conclusions of this article.

G.Skill 10666CL7D Conclusion

G.Skill F3 10666CL7D performance is impressive. I was very happy with the results obtained at different configurations and the low temperatures at the time of testing them. Perhaps, the only problem I would see here is that your motherboard needs to work with low voltages if you want to take advantage of your new acquisition, but as all DDR3 motherboards will at least work with 1.5v, you can always use 1600MHz 6-7-6 or similar configurations without heating the memory or consuming that much power.

The appearance is simple, but elegant. You won´t have tall heatsinks and LEDs on your RAM but they won´t pass unnoticed either. Use them on a HTPC and no one will even notice them, but you will know you have some good RAM there, and that should be enough for you considering you didn´t pay that much for them. On the other side, heatsinks will help avoid damage to the memory since they just reinforce the construction.

While all tests show that high frequencies or low latencies won´t give you a big boost on your daily tasks, functionality is still there. The G.Skill ECOs offer very low latencies with decent frequencies, and despite the fact that many brands offer similar specs, the vast majority won´t offer low voltages as G.Skill does. You will shave some watts and keep better temps on your setup with this memory.

I´ve been mentioning the great value this memory has but I haven´t mentioned the real price of this kit. As for April 2010, this kit retails at $119.99 at NewEgg, which is a very good price for its value. You won´t find many similar memory on this price range, I promise.

My recommendation is: either you have a high-end PC or a low-end HTPC, take this kit which will offer low power consumption and temps. It overclocks really well with relatively low voltages and it definitively won´t disappoint you in any way. You´ll be safe with G.Skill´s lifetime warranty too. Let´s finish this article with the Pros and Cons of the product.

Pros:

Cons:

Ratings:

• Performance: 9.50
• Appearance: 9.00
• Construction: 9.50
• Functionality: 9.00
• Value: 9.50

Final Score: 9.3 out of 10.

Excellence Achievement: Benchmark Reviews Golden Tachometer Award.

Related Articles:

• Hiper HPU-4M880 & HPU-4M630 Type M PSU's
• Zotac GeForce GTX-480 Fermi Video Card
• NVIDIA GeForce GTX 460 768MB Video Card
• Mad Catz S.T.R.I.K.E. 7 Gaming Keyboard
• Radeon HD 5770 CrossFireX Performance Scaling
• NZXT Hades Computer Case HADE-001BK
• ASUS P5K3 Deluxe Wi-Fi-AP Intel P35 Motherboard
• MAD CATZ R.A.T. M Wireless Mobile Gaming Mouse
• Zaward VIVO PCJ004 Exposed Copper Heatpipe CPU Cooler
• Palit GeForce 9600 GT 1GB Sonic NE/960TSX0202