The current series' flagship CPU, known as the Core i7-3960X, boasts a massive 15MB L3 cache and six cores clocked at 3.3GHz. As we found in our review last November, the i7-3960X is 20-30% faster than the Core i7-2600K, though at $999 it's over 3 times more expensive. As crazy as that price might seem, there are those willing to fork out the dough to purchase the world’s most powerful desktop processor.

There is also a slightly slower and considerably more affordable option. The Core i7-3930K is already selling for $599, and a third alternative, the Core i7-3820 is set to be released shortly. But if you are after a hexa-core processor with L3 cache to spare, then the Core i7-3930K is going to be the cheapest option for a while.

So if you're already spending ~$600 on a processor alone, you'll want to make sure the memory that you equip it with is able to unleash its true potential. Unlike the original Sandy Bridge processors with using dual-channel memory, the new LGA2011 variants feature a quad-channel memory controller.

This means Sandy Bridge-E processors are designed be equipped with four individual DDR3 modules and as such memory module makers have begun releasing LGA2011 kits. In fact there are already over one hundred different quad-channel memory kits to choose from, with manufacturers such as G.Skill and Corsair flooding the market with options.

Although we are focusing on how the Sandy Bridge-E platform performs when using various memory timings, frequencies and channel configurations, we will also be taking a look at three new memory kits. Those kits include the Division 4 Viper Xtreme from Patriot, Ballistix Elite from Crucial and the HyperX Genesis from Kingston.

These features are aimed at overclockers and enthusiasts alike, as it allows them to monitor in real time the health of their memory modules. This can help users to avoid the dangers of excessive heat when pushing the modules well beyond the designed operating specifications.

In conjunction with the real-time temperature monitoring, Crucial also radically redesigned the Ballistix heat spreader for improved thermal performance. Across the top of the module are 32 "fins" designed to increase surface temperature. Crucial claims that the new design cools the memory modules by up to 30% more than the previous Ballistix heat spreader design.

The finned version of the Crucial Ballistix Elite memory is now available in several kits ranging from 2GB to 16GB in capacity, with single modules coming as large as 8GB. There are actually no official quad-channel kits yet but as Crucial claim they are on the way we have accepted a pair of dual-channel kits.

Therefore today we will be checking out a pair of the 8GB dual-channel 1866MHz kits for a total capacity of 16GB across four channels. The Crucial Ballistix Elite PC3-14900 kits are designed to use CL9-9-9-27 timings at 1866MHz with just 1.5v. At just $88 per 8GB kit ($176 for quad-channel configuration) this Crucial memory seems like a steal, it will be interesting to see how they perform.

Unlike the HyperX T1 series which feature tall heat spreaders, the Genesis series relies on more traditional low-profile heat spreaders. Kingston claims that the Genesis heat spreaders, which are ventilated, provide maximum cooling efficiency.

Honestly we are not that concerned with the cooling performance, as most memory heat spreaders are purely included for aesthetics (marketing purposes). What we like about this design is its simplicity, the low-profile design means users won’t run into compatibility issues when using the Genesis memory.

Today the HyperX Genesis series offers a number of quad-channel kits supporting a wide range of capacities, timings and configurations. The supported frequencies range from 1333MHz through to 2400MHz, while capacities are offered from 8GB up to 32GB.

The kit that we have chosen to look at features four 4GB modules designed to operate at 2133MHz with CL11-12-11-28 timings. In order to operate at 2133MHz these modules require 1.65v of power, which is the maximum voltage the Intel memory controller can safely handle. Overall we really like the look of these Kingston HyperX Genesis memory modules and their bright blue appearance goes well with a number of high-end X79 motherboards.

Today we have the 16GB (PXQ316G1600LLQK) memory kit which is designed to operate at 1600MHz in a quad-channel configuration using CL 8-9-8-24 timings. Those are relatively aggressive timings for 1600MHz memory, though there are a number of CL7 kits from the likes of Corsair, G.Skill and Mushkin.

The Viper Xtreme modules utilize a custom heat shield design which features an extruded aluminum outer shield built around a copper core to provide improved cooling. Each module has 6 grams of copper at its core paired with an extruded aircraft-grade aluminum outer shield, to deliver a 5% improvement in cooling over traditional aluminum heat shields.

Patriot claims that the Division 4 Viper Xtreme series is built using only the highest quality pre-sorted IC's available and is subjected to rigorous testing and validation on the X79 platform to achieve optimized performance and maximum quality.

Hardware - Intel Core i7-3960X Extreme Edition (3.30GHz) - x4 4GB Kingston HyperX Genesis DDR3-2133 (CAS 11-12-11-28) - x4 4GB Patriot Division 4 Viper Xtreme DDR3-1600 (CAS 8-9-8-24) - x4 4GB Crucial Ballistix Elite DDR3-1866 (CAS 9-9-9-27) - Kingston HyperX 240GB (SATA 6Gb/s) - HIS Radeon HD 6970 (2GB) - Gigabyte G1.Assassin2 (Intel X79) Software - Microsoft Windows 7 Ultimate SP1 64-bit - Nvidia Forceware 285.38

As we mentioned earlier we will be basing our tests on three quad-channel DDR3 kits from Crucial, Kingston and Patriot. That said, we will not be comparing the kits directly as the focus is on Sandy Bridge-E memory performance. Therefore we will be testing to see how the Core i7-3960X responds to using different memory frequencies, timings and channel configurations.

Something we found interesting here is that the memory frequency had no impact on the write bandwidth, at least when testing between 1333MHz and 2133MHz. The write performance remained at 17.4GB/s despite the memory frequency which was surprising. The read performance on the other hand did change quite considerably and we consistently saw improvements as the frequency was increased.

At 1333MHz the read bandwidth was actually lower than the write bandwidth at just 16GB/s. Increasing the memory frequency to 1600MHz boosted the read bandwidth by 11%, while at 1866MHz we saw another 9% increase in bandwidth. Finally at 2133MHz the read bandwidth reached 20.7GB/s, which was 8% more bandwidth than what the 1866MHz memory delivered.

The memory frequency has a huge impact on latency and as expected the DDR3-2133 configuration was the fastest here providing roughly 11% more performance than the 1866MHz memory. The latency performance continued to decrease as the frequency dropped, and whereas the 2133MHz memory took 45.5ns the 1333MHz memory took 66.1ns.

As you can see the memory timings do not impact the performance nearly as much as the memory frequency. However like the frequency the memory timings have no impact on the write bandwidth performance. They do influence the read performance though not drastically. Here we found that the loose timings of CL12 were just 7% slower than the CL9 timings when measuring read performance, this equates to roughly 1.25GB/s.

The memory latency is also only slightly impacted by memory timings as we only saw an 8% improvement in performance when going from the CL12 timings to the CL9 timings.

It is very interesting to note that the quad, triple and even dual-channel configurations all delivered roughly the same bandwidth. In fact we found that the triple-channel configuration delivered slightly better read and write performance. It was not until we removed all the modules with the exception of one for single-channel mode that we saw a dive in performance. Still that said, for a quad-channel platform the Sandy Bridge-E is not a great deal slower when using only a single channel.

As the memory frequency and timings remained the same the latency was unaffected and all four configurations delivered roughly 64ns performance.

When testing with WinRAR we find that the higher the memory frequency the greater the performance. The gains were quite consistent, when going from 1333MHz to 1600MHz we saw a 7% performance increase, while going from 1600MHz to 1866MHz provided another 6% performance. Finally when testing the DDR3-2133 configuration we found that it was 6% faster than the 1866MHz memory.

The gains were not as evident when testing in Photoshop CS5 as the high-speed 2133MHz memory provided just 7% more performance than the much slower 1333MHz memory. This equated to just 1.1 seconds separating the slowest configuration to the fastest.

The x264 HD Benchmark was used to represent encoding performance and again the margins were not huge, although we did see consistent performance gains each time the memory frequency was increased. There was just an 8% performance gain seen when comparing the DDR3-1333 memory to the 2133MHz memory.

This time we find that the memory timings have even less of an impact on performance as we only saw a 7% increase in performance when going from the CL12 timings to the more aggressive CL9 timings. While there were clear performance advantages to using memory with tighter timings, the gains are not huge.

Here we found that the memory timings had almost no impact on Photoshop performance and just half a second improvement was seen between the CL12 and CL9 memory.

When encoding we only saw a little over a 1% performance gain separating the DDR3-1333 and 2133MHz memory.

These are by far our most surprising results yet. Here we find that the Core i7-3960X processor performs almost as well with a single stick of DDR3 memory running in a single channel configuration as it does with 16GB of quad-channel memory. Of course so far this is only true when using WinRAR, but they are surprising results nonetheless.

Here we see that it’s not just WinRAR that finds single-channel memory only slightly slower than the quad-channel configuration. Here just 0.1 seconds separate the two, making single-channel memory just as fast as quad-channel memory in this high-end graphic design application.

When testing the encoding performance we find the performance of the quad, triple and even dual-channel configurations to be virtually identical. Even the single-channel configuration is not a great deal slower which is again very surprising.

Here we see when testing with Dirt 3 that the memory frequency has virtually no impact on performance at either tested resolution.

It is the same case with Just Cause 2 as we again find no difference between the various memory frequencies.

Finally The Witcher 2 confirms what we have seen in the previous two games, the memory frequency has no impact on gaming performance.

Here we see that the memory timings have a slight impact on performance at 1680x1050, though at 2560x1600 the GPU is by far the greatest bottleneck.

Just Cause 2 shows no difference in performance between the various timings as CL12 and CL9 deliver the same performance.

Finally The Witcher 2 again shows virtually no difference in performance when adjusting the memory timings.

Although we were not entirely surprised to learn that memory frequency and timings really have no bearing on gaming performance, we must admit to being quite shocked about the channel performance. Here we see when testing with Dirt 3 that there is absolutely no change in frame rate when using a single DDR3 module compared to four, even at 1680x1050. Not only are we comparing single-channel to quad-channel performance, but it is also a 4GB vs. 16GB comparison.

Amazingly Just Cause 2 also shows no difference in performance between the single and quad-channel memory configurations.

Again we find the same when testing with The Witcher 2, which shows single and quad-channel memory providing the same performance.

While we were not shocked by this finding, we were surprised at how well the Core i7-3960X performed with just a single 4GB module. In many cases there was little to no difference between using quad-channel memory with a 16GB capacity, to a single-channel configuration with a capacity of just 4GB. That said, there will be certain applications that will max out 4GB of memory and such scenarios would cause a decline in performance.

Moving on we also found that memory timings had little impact on performance and it is certainly not worth paying a price premium for low-latency memory. If you look at quad-channel DDR3-1600 kits for example, CL9 memory such as that from the Patriot Extreme Performance 16GB series can be had for just $85, whereas the equivalent CL7 memory costs upwards of $160.

The next major consideration is frequency and this had by far the greatest impact on performance, particularly application performance. The best bang for your buck is going to be DDR3-1600, which starts at around $80 for a 16GB kit, and there are dozens of options priced under $100. That said, we did find DDR3-1866 memory faster though it does start at $130 and range to about $160.

Keeping those prices in mind, we do not imagine many are going to invest in a Sandy Bridge-E processor and then look to save a few bucks on the memory. That being the case, we found for maximum performance DDR3-2133 memory is the way to go and kits start at $140 and can range up to around $300 in price.

Of course we suggest investing in one of the cheaper 2133MHz kits as the more expensive versions are simply lower-latency kits and we already know they don’t do much to improve performance.

At this stage we are not sure where the Patriot Division 4 Viper Xtreme and Kingston HyperX Genesis quad-channel kits stand in terms of pricing, so it is impossible to say where they stand in terms of value. The Ballistix Elite is also only available through Crucial at this point and both their DDR3-1600 and 1866 kits are much more expensive than other alternatives you will find online.

For the majority of users investing in the Sandy Bridge-E platform the fact the quad-channel memory doesn’t really improve performance is irrelevant. The key advantage of this feature is the extra DIMM slots, which take the maximum memory capacity to 64GB. That said, those looking at purchasing the upcoming Core i7-3820 which is said to cost $285, perhaps an 8GB dual-channel memory kit is the way to go.