You may have noticed a common theme amongst our latest reviews within the last 6 weeks. We have assessed three Socket AM3 motherboards as well as the new Phenom II X6 1075T. Indeed, AMD seems to be the flavour of the month here at Overclock3D and whether you like it or not we have yet another AMD orientated product in store for you; the G.Skill FLARE PC3-16000 DDR3 set.
So lets just say that you're in the market for a performance machine that revolves around AMD's Phenom II series of processors and 870/890GX/FX motherboards. It stands to reason that you would probably want a quality set of RAM as well right? To start off with, we'll quickly walk you through today's memory kit.
|Model||2 x 2GB FLARE PC3-1600|
|Intel XMP Profile||No|
|AMD BEMP Profile||Yes|
Active RAM Cooler
First and foremost, yes this is a DDR3-2000 set of memory. Whats more is that it is rated with a relatively agressive timing set of 7-9-7-24 @ 1.65V. The kit definitely appears to have potential, that's for sure! Lets move on.
Packaging & Initial Impressions
G.Skill waste no time or resources on fancy box styling; this is exactly how we like it. Solid, protective and simple. After all the only important aspect of the item is the internal product itself!
Upon opening the box, the first thing you'll find is the G.Skill flame's cooling system. This is a dual 60mm fan set thats mounted on a metal frame. The mount itself is self explanatory as it uses a hook shape to fix itself to the memory area.
Finally on either side of the box you'll find the star of the show :- the memory modules.
Like a lot of performance RAM today, the G.Skill Flame series feature particularly large heatsink modules. As always you will need to take appropriate caution with memory compatibility with certain processor coolers, which may clash with the RAM.
For those who are interested in aesthetics, we would say these modules are very well styled, although it would have been great to to have followed the theme with a black PCB; sadly they are green.
Right, time for a bit of background information about the testbed and then some testing!
AMD Phenom II X6 1090T Black Edition
BIOSTAR TA890FXE Motherboard
4GB G.Skill Flare PC3-16000 RAM
4GB Kingston HyperX DDR3-1866 RAM
Samsung Spinpoint F1 320GB SATA II Hard Disk Drive
nVidia GeForce GTX 460 768MB GDDR5
Samsung 22x DVD+/-RW SATA
Windows 7 Home Premium x64
Forceware 258.96 Drivers
The Road to 2000MHz DDR...
Normally this would be a section dedicated to memory overclocking, but we felt that the circumstances commanded a slight name change.
The truth of it is that not every AMD Phenom II processor will be able to assist you productively in your endeavours to reach high memory frequencies. While processor yields continue to improve, there are still samples in circulation which are very flakey beyond 1600MHz.
The particular sample in my possession could well be one of the worst around but regardless I set about pushing our G.Skill Flame samples as close to their rated frequencies as possible. Sadly as great as my intentions may have been, my attempts weren't particularly fruitful.
With voltages notched up across the board, we reached a maximum frequency of just 1720MHz. No amount of refinement (looser timings, greater voltages, different memory ratio combinations) could offer a greater result. As it was clear that the memory itself was not the limitation we then decided to set its timings as tight as possible. After a short play, we found ourselves with a timing set of 8-8-7-24 1T.
Another point of view to consider is that sheer memory frequency isn't entirely the answer on AMD's Socket AM3 platform. With this in mind we hoped to determine the lowest timing set possible at the highest default frequency available on most boards; DDR3-1600.
Again, it seemed that our processor was a great limiting factor, as it was only possible to achieve this with a higher base HTT and lower memory ratio. Regardless, with a memory frequency of 1600MHz we attained a timing set of 7-7-6-20 1T. This to us was an impressive result.
SiSoftware Sandra (the System ANalyser, Diagnostic and Reporting Assistant) is an information & diagnostic utility capable of benchmarking the performance of individual components inside a PC.
The CPU arithmetic test ascertains the processor's capabilities in terms of numerical operations. Two subtests named Dhrystone and Whetstone are carried out respectively. This is not a measure of latency and thus higher is better.
The Arithmetic test shows frequency of memory taking precedence; albeit the difference remains marginal.
The CPU Multimedia Test focuses on CPU based operations that may occur during multimedia based tasks. The magnitude of the score depends on the processor's ability to handle Integer, Float and Double data types
Meanwhile the Multimedia test takes a preference towards memory latency.
In terms of SiSoft's memory bandwidth test, the results are near enough level.
Without surprise, the G.Skill configuration at 7-7-6-20 timings offers the lowest memory latencies.
CPU Queen is based on branch prediction and the misprediction penalties that are involved.
PhotoWorxx as the name may suggest tests processors by means of invoking functions that are common to Photo Manipulation including Fill, Flip, Crop, Rotate, Difference and Colour to B&W conversion.
This is an integer based benchmark that will test the CPU and Memory by means of the CPU ZLib compression library.
The Everest Ultimate Edition series of processor benchmarks show memory making a negligible difference. Pehaps the tests pertaining to core memory operations will offer more interesting results.
Both the tighter timing set and higher memory frequencies offer a larger memory bandwidth result over the Kingston set.
Interestingly the memory latency results show similar results regardless of frequency or timing configuration. Lets move on to more general system benchmarks.
WPrime is an excellent multicore compliant alternative to SuperPi.
The 32M test shows the lower latency memory taking a marginal lead over the higher frequency configuration.
Strangely the tables turn with the 1024M test where raw memory makes a 3 second difference.
PCMark Vantage is Futuremark's flagship "System Wide" benchmark. With a large focus on day to day operations, it's an excellent means of judging the capability of a computer as a whole.
The two G.Skill configurations perform quite similarly here, while the Kingston set at 1600MHz C7 scores strangely low.
3DMark Vantage is Futuremarks flagship gaming oriented benchmark at present and is considered to be a demanding one at that. Our tests were carried out under the "Performance" prefix.
Finishing off with the 3DMark Vantage test, we find all configurations performing quite similarly.
All in all, the G.Skill Flare memory kit performed well on our Socket AM3 testbed. Our tests showed positive results at both tighter timings and higher frequency. Neither configuration was noticeably superior.
As we had previously highlighted, it was a shame that our AMD Phenom II X6 1090T was unable to push the RAM to its rated 2000MHz frequency. It should be stressed that at no point did it appear as though the memory was the limiting factor in our overclocking endeavours. As much of a shame that it is however, our findings should be food for thought for those who intend on purchasing high frequency memory for their Socket AM3 systems.
It would be unfair of us to assert that no Phenom II X6 processor is capable of pushing RAM to high frequencies. Infact, with advances in the Thuban core's production, it should be quite possible to take memory upwards of 1850MHz if you so wished with a good processor. Needless to say however, this is risky business on Socket AM3 and that there is every chance that memory such as this is wasted on the platform.
Now lets discuss price. Unfortunately this particular set isn't widely available but we were able to find it on sale with a small supplier via Amazon at £159. If this is indeed the expected retail price for this 4GB kit in the UK, then it is very difficult to justify it. When these days it is possible to purchase C8 DDR3-1600 for well under £80, one really has to think twice when purchasing this kit. Just to put things into perspective, an £80 saving in memory is enough to build a machine around the Intel Core i7, rather than the Phenom II X6.
To summarise, the G.Skill Flare PC3-16000 memory series is a great performer. For those who really care about squeezing every point out of their benchmarks, then this is what you're after. Despite the RAM's recommended Socket AM3 usage, we believe it will prove its capabilities to a greater degree on a Core i5/i7 configuration instead. As for Phenom II users, do you really need this memory? We think not.
- Relatively quiet active cooler
- Good Styling
- Intended usage on Socket AM3 is arguable futile.
Thanks to Gskill for the Flare on test today, you can discuss our results in the forums.