by Ilya Gavrichenkov
12/22/2009 | 04:17 PM
What factors affect the general performance of contemporary desktop systems? Of course, it is first of all the CPU and graphics sub-system. The CPU determines how fact the computer can process data in computational and system tasks, while the graphics card determines the result visualization speed. Obviously, if we are talking about applications processing and creating content, the role of the graphics accelerator becomes secondary. If we turn to entertainment applications, such as 3D games, for instance, then the primary role will be exactly on the graphics card, as it will be responsible for the spectacular experience the user will get from interacting with the computational algorithms. Most of the articles on our web-site are dedicated to optimization and choosing the best combination of contemporary CPUs and graphics cards.
However, processor and graphics card are not the only two components in a contemporary computer system. There are a lot of other parts, which importance shouldn’t be underestimated. However, while we understand perfectly well the practical value of mainboard, hard drive or power supply unit specifications without any additional explanations, then the system memory choice may sometimes become a tricky quest even for experienced users. No, memory capacity is not our concern at this point, things are pretty clear with this parameter. We have long known that the majority of contemporary computer systems can do perfectly fine with as much as 4 GB of RAM: this capacity is quite sufficient for proper work of operating systems, general-purpose applications and last-generation games. Only in certain specific cases you may need more RAM and those users who deal with cases like that are definitely aware of it.
It is much more interesting to decide on the operational memory frequency and timings. And it is mostly interesting since a lot of lances are being broken over this matter: numerous memory makers go out of their way trying to convince us that memory performance is one of the key parameters of a contemporary computer system that has just as much effect on the general performance as a CPU or a graphics card. As a result, memory makers vary the prices of their kits in a very broad interval. For example, 4 GB of inexpensive DDR3-1067 memory can be purchased for as low as $80. At the same time the same amount of “elite” DDR3-1600 SDRAM can easily cost you over $200. But is this tremendous price differentiation justified?
Our previous experience shows that it isn’t. The earlier memory tests for LGA775, Socket AM3 and LGA1366 platforms proved: computer performance in real applications depends on the memory speed very insignificantly. And systems with high-speed memory can only please you with high scores in synthetic benchmarks measuring peak performance and latency of the memory sub-system, such as Lavalys Everest or SiSoft Sandra.
However, so far we haven’t discussed the influence of memory speed on the performance of LGA1156 platforms built around Core i7, Core i5 and Core i3 processors. In the meanwhile, these CPUs have a number of peculiarities connected closely with the memory: they feature a built-in high-performance DDR3 SDRAM controller, but unlike the LGA1366 modification of Core i7, work with only two memory channels instead of three. Therefore, it is possible that memory specifications may have a somewhat different effect on the performance of LGA1156 systems. We decided to dedicate this article to checking this assumption out, namely, to studying the connection between the bandwidth and latency of DDR3 SDRAM and performance of systems with LGA1156 processors.
Before we get to the actual benchmark results, we decided to remind you what the major operational principles of the memory controller in LGA1156 systems are. Core i7 and Core i5 processors (as well as the upcoming Core i3 and Pentium) manufactured in LGA1156 form-factor feature a built-in dual-channel DDR3 SDRAM controller located on the same die as the CPU cores. Like the on-die L3 cache, this controller works at its own frequency, which is different from that of the CPU. This frequency varies depending on the processor model and together with other parameters has certain influence on the overall memory sub-system performance.
Another feature of the LGA1156 processor memory controller is the support of different memory speeds depending on the processor. More expensive models with faster memory controller support DDR3-1600 SDRAM, mainstream solutions can only work with DDR3-1333 SDRAM being the fastest, while low-cost models can only clock the memory as DDR3-1067.
Here is the summary of DDR3 SDRAM frequencies supported by LGA1156 processors:
In other words, systems equipped with less expensive CPUs can’t accommodate fast DDR3-1600 SDRAM, at least without base clock frequency (BCLK) increase beyond the nominal value. However, the owners of Core i5 and Core i3 based systems can get access to faster memory modes during overclocking, although they will have to overclock not only the memory but also the CPU.
The thing is that LGA1156 processors use the same 133 MHz clock generator (BCLK) and a number of independent multipliers:
As we see, the owners of the most expensive Core i7 processors have the most flexibility in configuring their memory sub-system. Therefore, we used this particular processor in our today’s tests. However, all conclusions we are going to make today will also be true for other LGA1156 CPUs. The only difference for the owners of less expensive processors is that they have fewer opportunities to have their memory working at higher speeds. However, all junior processor models still have all the same tools for manipulating the memory timings.
We checked out the effects of the memory frequency and timings on the overall system performance in a Core i7-860 based platform. We decided to go with this particular CPU, because on the one hand it supports the widest range of DDR3 memory but at the same time features a relatively democratic price tag. This combination of parameters makes Core i7-860 processor a pretty popular solution these days.
The platform built around this CPU will be tested with DDR3 SDRAM working at 1067, 133 and 1600 MHz. DDR3-800 modules were not included into our today’s test session, because they turned out pretty hard to find in retail: it was a transitional standard that is why this memory has almost completely vanished from the market by now. As for the timings, we tried to cover almost all most popular settings combinations.
During our tests we used the following hardware and software components:
We would like to thank AMD Russia for providing us with ATI Radeon HD 5870 graphics card for this testbed.
First of all we ran some tests in the system nominal mode, when none of the components were overclocked. We only changed the memory frequency multiplier and timings. I have to say that during this test we tried to emulate the most typical operational conditions that is why we decided not to disable any processor technologies. Hyper-Threading, Turbo Mode and Enhanced Intel SpeedStep worked the way they were supposed to: the operating system saw our CPU as an eight-core one, its frequency dropped to 1.2 GHz in idle mode and under operational load of different intensity it increased to 3.46, 3.33 and 2.93 GHz.
First of all, we checked out the results in synthetic benchmarks, which measure the performance of the memory sub-system separately from the rest of the platform. Lavalys Everest 5.30 is a perfect example of a test application like that.
As you can see from the obtained results, memory sub-system settings do have some effect on the measures system parameters. As the memory frequency increases and the latencies drop, we see that the practical bandwidth starts growing and the practical latency – decreasing. As a result, the performance difference between DDR3-1067 and DDR3-1600 in Everest test reaches 18%.
We also used another synthetic test called MaxMem2. The peculiarity of this test program is that it not only measures the memory bandwidth and latency in single-threaded mode, but also can estimate the memory performance in case of multi-threaded operation.
Overall, we see almost the same situation as in Everest. However, the results in multi-threaded mode are in fact pretty interesting. According to the numbers shown on the corresponding diagram, the memory frequency appears way more important than timings, when multi-threaded memory access takes place. DDR3-1333 shows more than 20% advantage over DDR3-1067, while DDR3-1600 appears almost 40% faster than a similar system equipped with slow DDR3 SDRAM.
However, the above mentioned numbers are the results of synthetic benchmarks, which are not directly connected to the real-life memory sub-system performance. Therefore, let’s go over to the real applications. To estimate the average performance in popular applications we used CPC Benchmark 2007, which measures the performance in GIMP graphics editor, during video transcoding using Handbrake utility and under multi-threaded load created by simultaneous HD video playback and data compression in 7-zip archiving tool.
And this is where we see very clearly that in real life things are completely different from what we have just seen in synthetic benchmarks. In many cases fast DDR3-1600 memory provides only an imaginary advantage over DDR3-1333 and DDR3-1067 platforms: very often the results differ by less than 1%. Very often, but not always. For example, we see in the multi-threaded test that faster memory may still have a noticeable effect on performance. By raising the DDR3 SDRAM frequency only one step higher we can increase the system performance by the good 4-6%. And as you can notice, this frequency increase causes a more substantial performance improvement than lowering of the timings.
There are also individual applications that appear very sensitive to memory sub-system parameters:
For example, during data compression using WinRAR we can achieve 8% acceleration only by adjusting the memory sub-system settings.
Image processing in Adobe Photoshop, on the contrary, barely depends on the memory frequency and timings settings.
x264 codec is also pretty indifferent to DDR3 SDRAM frequency and timings.
However, during video transcoding using Cyberlink MediaShow the performance may vary within 5% interval depending on the memory settings. And by the way, just like with the archiving application, this performance is more sensitive to frequency rather than timings.
We can also see certain influence from the memory parameters on system performance during final rendering in Autodesk 3ds max. The dependence here is not that serious, but still pretty noticeable.
Contemporary 3D games are a separate part of our today’s test session.
As we see, gamers shouldn’t underestimate the importance of adjusting the memory sub-system settings. Of course, we are not implying that faster memory will provide you tens of percents of improvement, but the correlation between frames per second and memory frequency and timings can be seen clearly with a naked eye. One step higher memory frequency will provide about 2% change in gaming performance. You can achieve the same by adjusting the timings by a few clocks.
As a result, summing up everything we have just discussed, we can conclude that it hardly makes sense to invest additional finances into getting faster memory for your LGA1156 system. Dual-channel DDR3-1333 or even DDR3-1067 SDRAM offer sufficient bandwidth for normal operation of a contemporary LGA1156 system. However, if you are looking to squeeze the maximum out of our platform, then fast DDR3-1600 may come in quite handy.
Although we didn’t discover any obvious advantages in using high-speed memory during the first part of our today’s test session, it doesn’t mean anything yet. High-speed memory like DDR3-1600 SDRAM is targeted primarily for computer enthusiasts and will most likely find its way into overclocked systems. In fact, during our previous high-speed memory tests we often noticed that its frequency potential can best be revealed during CPU overclocking. That is why the second part of our test session will be dedicated to performance analysis in an overclocker system.
In order to get as close to real-life conditions as possible, we overclocked our Core i7-860 processor to 4.0 GHz. Since the maximum multiplier for this processor is 22x under any load, we had to raise the base clock from 133 to 182 MHz in order to hit 4.0 GHz mark.
Turbo Mode technology, which changes the processor clock frequency multiplier dynamically, was disabled in this case, because this way we could achieve highest possible overclocking.
The increase in clock generator frequency caused certain changes in the list of supported memory frequency modes that can be set using available multipliers. Instead of DDR3-800, DDR3-1067, DDR3-1333 and DDR3-1600 SDRAM, the CPU automatically got support for DDR3-1092, DDR3-1456, DDR3-1820 and DDR3-2184 memory. The latter is obviously hardly possible at this point, because there are no memory modules in the market with these specifications. The remaining three modes are quite possible, which means that the currently available overclocker as well as regular memory modules can easily be used in LGA1156 systems with overclocked processors. We checked out all three of these modes in our second test session using various timings configurations that could be applied in each of these modes.
As usual, synthetic benchmarks come first.
As we have expected, during CPU overclocking the scattering of practical bandwidths and latencies of the memory sub-systems based on modules with different frequencies and timings grew bigger. Which is, in fact, not surprising at all, since the increment between different DDR3 SDRAM modes increases when you raise BCLK frequency.
We see a similar picture in another synthetic test called MaxMem2.
I would like to point out that processor overclocking speeds up the memory sub-system, even if the memory itself keeps working at the same frequency. For example, if we compare our results for DDR3-1092 SDRAM against the performance of DDR3-1067 from the previous part of our test session working at a close frequency, we will notice a significant increase in real practical bandwidth and lowering of the memory latency. There is nothing surprising here: it all comes from L3 processor cache and memory controller overclocking following the BCLK frequency increase.
However, the results of synthetic benchmarks are not so interesting for us, because they only give us an idea of theoretical advantages of high-speed memory in the most favorable testing conditions. As for the real-life performance, we can only explore it in real applications. Just like in the previous part of our tests, we will start with CPC Benchmark 2007.
Those applications where memory speed didn’t have that big of an influence on the system performance in nominal mode, do not seem to react that much to the memory sub-system settings during overclocking. However, in those tests where we did see certain influence, high-speed memory can work a small wonder. For example, during multi-tasking test the performance difference between the fastest and the slowest DDR3 SDRAM mode is over 15%, which is roughly speaking twice as much as the performance difference between the two closest processor models within a family. And even if we look at the total score in CPC Benchmark 2007, we definitely shouldn’t disregard memory sub-system settings. However, hitting the highest memory frequency at any price is hardly the best strategy here. In particular, you can significantly improve your system performance by simply lowering the timings of the memory working at lower frequency.
Archiving uses a lot of system memory. That is why the results obtained in WinRAR illustrate everything we have just said in a great way.
However, there are applications, which do not depend on the memory sub-system performance in overclocked systems. One example like that is Adobe Photoshop. Although, to be completely fair I have to say that in this case we still do see a slight difference in the performance of our test script applying typical actions to a 12-megapixel image.
We can say the same thing about x264 decoding speed: even in an overclocked system its performance barely depends on the memory frequency and timings settings.
However, during video transcoding in Cyberlink MediaShow things turn totally different. We obtained about 4% performance improvement from raising the memory frequency only one step up, and another 1.5-2% - from adjusting the timings by one cycle.
As for the effect from the memory sub-system settings on the performance in Autodesk 3ds max, we could say that it was average. In practice it means that if you are putting together a system mostly for final rendering work, then it makes more sense to focus on the CPU speed first and only then on high-speed memory.
And now comes the most interesting part. Memory frequency and timings in an overclocked gaming system have the same influence on the overall performance as in a system working in its nominal mode. And it means that even though we can clearly see the connection between the memory sub-system settings and fps rate on the diagrams, it will be barely noticeable during actual gameplay.
So, it turns out that overclocking doesn’t really change the dependence between the memory settings and system performance. As a result, during overclocking the potential of almost any DDR3 memory should be sufficient for building a well-balanced system. In other words, with a few specific exceptions in individual circumstances, overclockers shouldn’t really waste their time searching for high-speed memory. Especially, since LGA1156 processors have low minimal multiplier for the memory frequency, which allows using widely spread DDR3-1333 SDRAM even at a seriously increased base clock without any problems.
Although we didn’t discover any obvious and strong connection between memory speed and system performance, it doesn’t mean that you should totally underestimate the importance of choosing the right memory modules for your needs. I doubt that you would agree to lose a few percents of performance if you can get them practically for free. Yes, although some overclocker memory makers set pretty high prices on their well-promoted solutions with exclusive heat-spreaders, there are quite a few affordable DDR3-1600 kits out there, which will cost you only a few dollars more than the regular DDR3-1333 or DDR3-1067 SDRAM. We strongly recommend paying special attention to these particular modules: they will not only allow you to use the potential of your LGA1156 platform fully, but will also please you with their good overclocking functionality. There is only one thing you should keep in mind in this case: if you are not going to overclock your system, then DDR3-1600 makes sense only for Core i7-800 processors that support this default speed mode.
Today we are going to introduce to you two inexpensive and popular DDR3-1600 SDRAM kits like that for LGA1156 platform, which we can definitely recommend as a good buy.
Corsair offers a lot of different solutions for LGA1156 platform, but we picked a 4 GB DDR3-1600 kit from the Dominator series. These memory modules are equipped with special “tall” aluminum heat-spreaders with a comb-like top edge that promise improved overclocking potential. I have to say that modules like that cost only a little more than their analogues with traditional heat-spreaders: the difference usually doesn’t exceed $15. Unfortunately, while engineers did their best to ensure proper cooling of the memory chips, you may have problems installing these modules into systems with large CPU coolers. Luckily, the top part of the modules heat-spreaders can be unscrewed and removed.
CMD4GX3M2A1600C8 memory modules we picked support “intermediate” timings of 8-8-8-24. Of course, Corsair also offers similar modules with more aggressive timings, but they also cost more, and honestly, the performance improvement in this case is too small to justify the price difference. CMD4GX3M2A1600C8 modules should work at 1.65 V, because LGA1156 processors set the same limitations to memory voltage as their LGA1366 fellows.
The parameters in modules SPD make them compatible with 1333 MHz frequency. The modules nominal settings are saved in the XMP profile, which allows setting memory timings automatically on those mainboards that support XMP technology.
During our tests of Corsair CMD4GX3M2A1600C8 modules we tried to find their maximum stable frequency and timings. We used Memtest 2.11 and Prime95 25.7 x64 benchmarks for that.
First of all we determined that at 1.65 V voltage and 1600 MHz frequency these modules remain stable not only with their nominal timings of 8-8-8-24, but also in a slightly more aggressive mode. We managed to get this DDR3-1600 memory to work fully stably at 8-7-7-20 timings.
We also checked out what these modules are capable of with different timings configurations. The table below shows the maximum frequencies, at which the memory passed our stability tests with different timings. We ran these tests on Corsair Dominator kit also at 1.65 V voltage and 1T Command Rate.
As we see from the obtained results, overclockers should really be pleased with excellent overclockability of Corsair CMD4GX3M2A1600C8 kit with less aggressive timings. Namely, it can work at over 1900 MHz frequency with the timings set as 9-9-9-27, which may be very helpful during CPU overclocking by raising the BCLK frequency.
This particular peculiarity allows us to recommend this memory kit for consideration, because as we have just seen, high frequency is often more important than lower timings.
Kingston is one of the most well-known names in the enthusiast memory market. That is why we were not surprised to see that they offer just as many high-speed products for the LGA1156 platform, as they do for LGA1366. Their frequencies reach as high as 2133 MHz, which is impressive. However, the company doesn’t forget about the mainstream DDR3 SDRAM as well, and 1600 MHz kits are definitely among them. For our today’s tests we chose a 4 GB kit of two HyperX KHX1600C8D3K2/4GX modules that are selling at a very democratic price, which is very close to that of DDR3-1333 kits. Just like the Corsair modules we have just discussed, HyperX KHX1600C8D3K2/4GX DDR3-1600 memory supports 8-8-8-24 timings and 1.65 V voltage.
These pretty popular specifications come with a traditional cooling system. These modules are equipped with heat-spreaders of Kingston’s brand name design but pretty standard size. They cover both sides of the modules. Since Kingston decided not to experiment with large heat-spreaders, these modules will fit into any system featuring a large CPU cooler.
SPD of Kingston’s memory modules is pretty similar to that of Corsair’s. The standard settings profile is supposed to provide maximum compatibility at low frequencies. The XMP profile that makes it a lot easier to configure the memory sub-system in the mainboard BIOS Setup contains all the default settings.
Although HyperX KHX1600C8D3K2/4GX memory is intended to work with CAS Latency 8, we managed to achieve stability at much lower settings. This DDR3-1600 SDRAM kit successfully passed stability tests in Memtest 2.11 and Prime95 25.7 x64 with 7-7-6-20 timings.
I have to say that it is a pretty remarkable result. It turns out that HyperX KHX1600C8D3K2/4GX with default CAS Latency 8 can compete successfully against memory kits with CAS Latency 7, which are somewhat more expensive.
As for the overclocking potential of Kingston HyperX KHX1600C8D3K2/4GX with different timings, the table below lists all confirmed frequencies, when the tested memory remained stable:
This is where Kingston memory really surprised us: with 9-9-9-27 timings it went way beyond the psychological DDR3-2000 barrier and worked at an impressive frequency of 2172 MHz, which is truly unbelievable for DDR3-1600 SDRAM.
With other timings configurations HyperX KHX1600C8D3K2/4GX managed to go far beyond its nominal settings. These impressive overclocking results give us every right to recommend this memory kit to all overclockers: priced fairly low, it offers unimaginable potential for frequency increase as well as lowering the timings. We are proud to give Kingston HyperX KHX1600C8D3K2/4GX kit our Recommended Buy title:
Overall, I can conclude that the quality of memory chips used for overclocker memory modules has improved significantly lately. A few months ago we tested triple-channel DDR3-1600 SDRAM kits and at that time we didn’t see anything as impressive as today.
System memory speed doesn’t have a serious effect on the platform performance – this rule also works for LGA1156 systems. In fact, we are not surprised with this outcome: every time we test memory modules for contemporary platforms we arrive at this conclusion in the end. The results of our practical performance tests suggest that DDR3 SDRAM frequency and timings can only make as little as 5% difference on average in LGA1156 systems. Of course, there are a few exceptions to this rule, such as synthetic benchmarks, for example, but they do not affect the overall picture too much.
All this allows us to conclude that it is hard to underestimate the importance of optimal memory sub-system parameters, but at the same time we shouldn’t completely ignore this aspect either, because in most cases getting the best DDR3 SDRAM for your LGA1156 system doesn’t require any serious additional financial investments. As a result of the recent price drop on DDR3 SDRAM, DDR3-1067, DDR3-1333 and DDR3-1600 kits sell at very close prices. Moreover, the best memory models boast excellent overclocking potential, which makes it possible to build a highs-speed memory sub-system at a very moderate cost.
At the same time, it is important to keep in mind that most LGA1156 platforms working in nominal mode can only clock system memory at 1333 MHz. And only systems with CPU from Core i7-800 series support multipliers that allow settings the memory frequency at 1600 MHz. It means that if you are not planning to overclock, then in most cases you should be fine with DDR3-1333 SDRAM with low timings. However, it you consider yourself an enthusiast and think of doing some overclocking, then it makes more sense to go for a higher-speed DDR3-1600 like Kingston HyperX KHX1600C8D3K2/4GX or Corsair Dominator CMD4GX3M2A1600C8 kits we have discussed today. This memory will provide you with additional flexibility in configuring the system parameters and will deliver a few percent of extra performance. However, we do not recommend considering those overclocker kits that cost much more than the regular DDR3-1333 SDRAM. Instead you can use the available finances for a faster CPU or graphics card.