Kingston DDR3 HyperX Memory Review
Choosing the right memory for your desktop PC system is a tricky choice for many. The temptation is to buy the fastest memory possible and then overclock it to go even faster. Common sense suggests to us that memory rated at 2000MHz will be 25% faster than memory rated at 1600MHz but this is not the case in synthetic tests let alone real-world benchmarks.
Today we're going to explore the truth behind the impact of memory choice on system performance and explain in non-technical terms how any system can make the most of memory and how to select the most appropriate memory for any given system. To do this we will use Kingston's high-end HyperX memory in the shape of DDR3-2000 and DDR3-1600 modules. Although we could test the DDR3-2000 in triple channel mode we will use only dual channel so that direct comparisons can be made with previous tests on other manufacturers products.
Founded in 1987 with a single product offering, Kingston now offers more than 2,000 memory products that support nearly every device that uses memory, from computers, servers and printers to MP3 players, digital cameras and mobile phones. Kingston Technology grew out of a severe shortage of surface-mount memory chips in the high-tech marketplace in the 1980s. John Tu and David Sun were determined to find a solution. They put their engineering expertise to work and designed a new Single In-Line Memory Module (SIMM) that used readily available, older technology through-hole components. A new industry standard was born — and, on October 17, 1987, so was Kingston Technology.
Why should all of this matter? We'll discuss the importance of brand value and brand reliability in the conclusion near the end of this review.
The triple channel kit we received is sheathed in heavy heat spreaders and is of the "tall" variety meaning that care must be taken to ensure the DIMMs do not interfere with large cooling solutions. We use Corsair H50s in our test systems and they have never impinged on the memory sockets of any board we have tested - that will not be the case with a Big Boss 2 heatsink and in those situations it may be only possible to use 2 out of the 4 (or 6) available sockets.
The DDR3-1600 chips don't need as much cooling and come with more traditional heat spreaders that will not interfere with any cooling solution.
Here's a summary of the test systems:
Although we used the Intel system for the Sandra benchmarks we tested with our AMD test rig as well for the final test for reasons that will become apparent.
Let us start with using the memory in overclocked systems.
3.5GHz and perfectly stable after an hour of 100% cpu usage.
4GHz and still at stock voltage! Unfortunately not stable during the torture test until we increased core voltage to 1.35v, which is by no means a big increase and well within safety parameters.
of this is to show how well a Lynnfield (and to a lesser extent the Athlon II)
processor overclocks. By contrast, if we overclock the memory, the maximum speed
we can achieve is 2133MHz - a mere 6.7% overclock and one that is not stable
unless we increase the voltage to 1.70v instead of 1.65v (compared to 40% for
the CPU). Worse still, the only timings that are acceptable to allow boot up are
2T instead of 1T and this brings us to one salient point that readers should
note. Overclocking the memory to 2133MHz actually makes it run slower! This is
why we always recommend buying the correct memory for your system and not
relying on overclocking it.
Synthetic tests can all be summed up with one benchmark - SiSoft Sandra's Memory Bandwidth. This test is purely for the memory and controller and will vary by architecture more than by differing memory manufacturers.
This benchmark is the only one where you will see a big difference between different brands. The real-world difference between fastest and slowest memory will be less than 1% for virtually any benchmark. That statement was a generalization for those video editors about to Email us to disagree - the average user will not be able to notice the difference between memory of the same grade). If you don't believe us then just Google (Or Bing) for "DDR3 benchmarks" (we recommend Anandtech or Toms Hardware for comprehensiveness) and you will see many of the latest games benchmarked showing barely 1 frame per second difference from top to bottom due to memory.
Now we have that out of the way let's examine the results. the DDR3-2000 from Kingston is the fastest in that category and the DDR3-1600 Kingston memory comes second in the 1600MHz group. The OCZ and Corsair results have been particularly disappointing given their previous position as brand leaders. Perhaps cost-cutting has impacted product quality? Cell Shock and G.Skill are still relatively unknown in the mass market and we cannot say where they source their chips from to analyse their performance in detail.
Now let's look at something that may surprise you and make comparisons between different brands seem trivial.
The above chart shows the performance of Kingston DDR3-2000 memory on an Intel i7-870 (socket 1156) and an AMD Athlon II X4 630 (socket AM3). We tested by disabling cores in the BIOS to get results for 1, 2, 3 and 4 cores. The results are interesting to say the least. Theoretically, synthetic memory bandwidth tests should be independent of the processor(s) used and rely only on the memory subsystem. The above results show that any Intel system will make the most of its memory but it takes a quad core AMD processor to max out the memory (and that memory is only 1600MHz unlike the 2000MHz RAM used by the Intel system).
The purpose here is to show that the situation is a little more complicated than say, a review of a graphics card or processor where the inherent benefits are what count.
So what can we conclude? Buying the fastest memory doesn't always bring the fastest results. Overclocking memory brings results that are hardly worth the effort. Worse still, if overclocking results in random instability then it can be quite tricky to trace the errors to memory problems. What criteria then are the most important? Reliability is the biggest factor and for this reason it is advisable to purchase from a reputable manufacturer. Kingston have always had dependable memory in the mainstream segment and have now gained considerable traction in the high-end market with their HyperX brand. By contrast Corsair and OCZ have gone the way of Mercedes (who famously lost the top brand desirability slot first to BMW and then to Lexus), they are still reliable brands but no longer seen as "luxury" products. That brings us to the second most important criteria - price. If the performance of memory makes little real-world difference and you have a shortlist based on quality and reliability then why not select the most cost effective?
While it's a good idea to get the fastest supported speed (in this test it was 2000MHz for the Intel platform and 1600MHz for the AMD one) purchasing memory with the expectation of overclocking can lead to disappointment as limits (such as 1.65V DIMMs recommended by Intel as a maximum) are already being pushed by manufacturers. Incidentally, with the introduction of DDR3 for AMD processors and motherboards, the final nail in the coffin of DDR2 seems to have been hammered and it is now no longer advisable to purchase DDR2 memory for reasons of platform longevity and ease of future upgrading.
At the end of the day, memory is an enabler - a PC will not function without memory but spending a fortune on the fastest memory will not make the PC run that much faster.
We compared all of the products in this review on Newegg (www.newegg.com). As of December 2009 the cheapest set of DDR2000 memory was a tie between Kingston and G.Skill at $135. The situation is similar at the $105 mark for the DDR3-1600 memory kits.
We've had problems in the past with unbranded memory and while it is tempting to pick up some no-name DIMMs at a big discount, there have been times when we could not get unbranded memory to even work at the rated timings, let alone have room for overclocking and were forced to underclock to achieve stability.
As a general rule, going from 2GB RAM to 4GB RAM (especially for Vista and Windows 7) will give a much greater performance increase than going from DDR3-1066 to DDR3-2000. If the budget for memory is tight and there's a choice between a minimum and a recommended amount of RAM it's always better to go for the recommended amount to run an application or game or Operating System even at the sacrifice of slower memory..
Kingston are making big inroads into the high-end memory market segment and it's not hard to see why. When everything is taken into consideration we have no hesitation in recommending Kingston's HyperX memory kits in terms of quality, reliability and price/performance characteristics. Surveys show that computer RAM is the most popular PC component given as a present in the holiday season (probably because it is so easy to install) and will be all the more appreciated if selected to perfectly complement the system it is to be fitted into.
All trademarks are the property of their respective owners.