Tech Reaction

Introduction

We’ve seen most major memory manufacturers releasing new products optimized for the LGA1366 platform. Reaching speeds above 2000MHz at 1.65v is testament to how far things have come since hitting 1250MHz at 2.3v on our DDR2 sticks was reason for enthusiasts to brag. Now with Bloomfield we all know that we cannot simply push more voltages, provide moving air to the modules and hope we end up with a winning kit. That’s simply not the case. There is now a limit of 1.65v which one shouldn’t cross, to ensure the longevity of one’s processor. Now how does one get top notch performance and timings while not giving up on value as well?

On the topic of value and performance, we’re looking at a set of G.Skill Trident 6GB DDR3 F3-12800CL6T-6GBTD modules that are designed for use in the LGA1366 (Bloomfield/Core i7) platform. We’ve seen in the past that G.Skill modules typically deliver great performance at excellent prices, particularly in the more “mainstream” market, so let’s find out if G.Skill has produced another winner with this Trident kit.

First, a little background:

G. SKILL, established in 1989 by enthusiasts, is a leading memory module Manufacturer based in Taipei, Taiwan. 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. At G. SKILL, we apply a top priority to quality standards. All of its products are put through a series of rigorous tests and strict quality control procedures. 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.

Specification and Packaging

* Speed: DDR3-1600 (PC3 12800)

* CAS Latency: 6-7-6-18

* Capacity: 6GB (2GB x 3)

* Voltage: 1.65 Volts

* PCB: 6 Layers

* Registered/Unbuffered: Unbuffered

* Error Checking: Non-ECC

* Type: 240-pin DIMM

* Warranty: Lifetime

G.Skill packaging is still the same. The memory comes in a clear blister pack lined up above each other and can be inspected without having to open the package. The back lists a few marketing words with the details of the specification and intended settings of the kit. The packaging seems a bit lacking for a high-end kit but that’s hardly a problem, because once the RAM is installed no one really cares about the packing.

Test Setup

CPU Intel Core i7 950

Motherboard MSI x58 Pro-E

Case Open Air

Memory G skill Trident 1600MHz

Video Card EVGA GTX295

Hard Drives 2 x Seagate 3.5″ HDD’s 500GB

Optical Drive Samsung Sata1 DVD Burner

Power Supply Cooler Master Real Power Pro 850W

Type Of Coolers Used Cooler Master V10

CPU Speeds Used 3.2Ghz, 167.5 x 24 = 4021 Mhz

Operating System Used Microsoft Windows Vista x86

Type of Software Used During Testing Super PI, wPrime, Everest, Sisoft.

We tested this kit with the CPU running at 3.2 Ghz and 4 Ghz with varying speed/latency combinations as follows:

1. 1333 MHz CL9

2. 1600 MHz CL9

3. 1600 MHz CL6 (stock speed)

4. 1866 MHz CL8

5. 2000 MHz CL8

Note:

When running the memory at 2000 MHz, the lowest speed that we could run the CPU at was 3.3 GHz instead of the 3.2 GHz used for the other tests. Please keep this in mind when viewing the benchmark results.

Also, 1866 MHz cannot be run with standard settings when the CPU is at 4 GHz, hence that particular frequency has been left out at higher speeds.

Super PI

First set of utility we are testing is Super PI, which is essentially a CPU benchmark. However, the 32M test shows sensitivities to memory subsystem tweaks and total bandwidth hence it has been included in the test suite. Here we calculate pi to 32 million digits and record the time it takes to complete the run. Lower time to completion is better.

*lower is better

Result: As can be clearly seen, this is one test which is effected both by higher frequencies and lower latencies. But as you can see when CPU is running at 4Ghz, 1600Mhz CL6 and 2000Mhz CL8 returns exactly the same time which tells us the importance of tighter timings. So this is one test where a clear frequency winner cannot be pronounced. Take your own pick, lower frequency with tighter timings or higher frequency with loose timings.

WPrime

In hindsight, WPrime should not have been included to benchmark RAM. While it scales fairly linearly with CPU speed, it shows little reaction to increased memory speed or tighter timings. We haven’t removed WPrime from the benchmarks just to show that the memory has been stable enough even at 2000Mhz CL8.

*lower is better

Results : In our tests there was no definitive way this benchmark reacted to memory frequency and timings but still 2000Mhz CL8 is the winner.

Sisoft Sandra

Now we will be testing the modules with SiSoftware Sandra, one of the most popular synthetic memory benchmark utilities. Sisoftware Sandra (System ANalyzer, Diagnostic and Reporting Assistant) is an information & diagnostic utility. It should provide most of the information (including undocumented) that you need to know about your hardware, software and other devices.

First we tested the integer and float bandwidth available to the system at different frequencies with the CPU running at 3.2 GHz. Higher is better.

*higher is better

Then we tested again, this time with the CPU at 4 GHz.

*higher is better

Results : As was expected higher frequencies give higher bandwidth. But note that 2000Mhz tests were omitted here because even after repeated testing at 2000 MHz, the total bandwidth available was dropping drastically which cannot be explained, other than by the assumption that perhaps the motherboard was not up to the task and was bottlenecking somewhere.

Below are the SiSoft Memory Latency Test results:

*lower is better

Result : Again there is no competition when these sticks are running @ 2000Mhz CL8. Though performance @ stock 1600Mhz CL6 is pretty good as well.

EVEREST

Lavalys EVEREST is a system diagnostics and benchmarking utility that provides system and overclock information, advanced hardware monitoring and diagnostics capabilities, including tests for the memory subsystem. It also tests the actual read and write speeds of your memory giving a fairly accurate look of true memory performance. Here we ran the benchmark to record the read and write bandwidths, as well as the latency.

*higher is better

CPU running at 4 Ghz.

*higher is better

Result : Just like the Sandra test at 2000Mhz CL8, memory read/write performance takes a huge performance hit without an explanation, since the system is completely stable at those speeds. However, memory copy does scale with speed as expected.

Below is the Everest Memory Latency Test:

*lower is better

Result: As was expected with higher memory speed and then combined with overclocked CPU, latency dropped by quite a margin. And 2000Mhz CL8 is the clear winner in the latency test.

Overclocking

We have already seen in the past few pages that the memory kit successfully and stably did 2000 MHz CL8-9-8-24. But hold on to your horses, because that is not the highest stable speed that the kit can do! In our tests it was able to do 2100 MHz CL8-9-8-24 1T at 1.7v which is simply an amazing feat for a set of modules that G.Skill doesn’t even claim to be its highest end set. The reason we were unable to add the 2100 MHz results in the tests is because to get to those speeds we had to run our CPU much higher too.

Pricing and Conclusion

We were impressed with the level of performance of these modules. These certainly give a huge improvement over 1333 MHz kits, even at its stock speed of 1600 MHz CL6-7-6-18. But an an absolute overclocking limit of 2100Mhz CL8-9-8-24 is simply an amazing feat. To get the most benefit from this kit though, I would certainly recommend the CPU to be overclocked as well.

This kit is pretty cheap at Rs. 11,500 from SMC International, as far as extreme performance kits go. Though for people looking to save on some money there is another kit which can be recommended and that is the G.Skill Trident 2000Mhz CL9. Though it doesn’t have much overclocking headroom, it’s stock speed should be enough for most users.

Pros :

- Good Looking heatspreaders

- Stock performance

- Good overclocking Headroom

- Lifetime warranty

- Price

Cons :

- None!

Keeping in mind these really excellent results and the fact that these modules lived up to all our expectations. Its only fair to give the kit a 5/5. Congratulations to G.Skill for bringing out such an awesome kit at a fairly affordable price.

Thanks to Mr. Saini from SMC International, India for providing the review sample.