by Ilya Gavrichenkov
12/02/2010 | 01:23 PM
Recently we have begun to publish a series of reviews with the results of a comprehensive test of modern CPUs we did in our test labs. Right now you can read two articles on that topic:
Thus, we have already covered all desktop processors from both AMD and Intel that are officially priced at below $200. It’s time to move on further now, to more expensive and faster products. As we noted in our earlier articles, $200 is the threshold after which the performance of CPUs does not grow up at the same rate as their price. Therefore we don’t think that choosing expensive CPUs for general-purpose desktop computers is a rational solution. If you seek higher overall performance, your investments into the graphics subsystem or, for example, into SSDs are going to be far more effective.
Well, of course we don’t mean that purchasing top-end CPUs is a sheer waste of money. There are situations and scenarios when high processing power is called for and quickly returns the investment into it. It is when the computer is used for computations-heavy applications. Home users and amateurs often dabble in video processing such as simple transcoding, nonlinear video editing or application of visual effects. Various 3D modeling suites and audio editing/mastering systems may also need high processing power. Gamers may also be interested in top-end processors, especially if they have premium-class graphics solutions and strive for the highest image quality possible. Thus, expensive CPUs are not a luxury but a demanded class of products, even though not as popular and demanded as entry-level and mainstream CPUs.
It must be noted that the top-end CPU market is structured differently than the low-end and midrange segments. The main difference is that there are almost no offerings from AMD here. AMD had to leave the top-end market sector not through the company’s own volition. While Intel’s engineers have been constantly improving their microarchitecture for the last few years, AMD has largely limited itself to polishing off its manufacturing process. As a result, AMD is now unable to offer a serious alternative to Intel’s top-of-the-line CPUs.
Thus, this article will largely be concerned with Intel’s solutions, especially as Intel has spent a lot of effort to develop its high-performance products and even offers a special platform for them.
But let’s be systematic and start out with those AMD models that manage to be objectively priced above the $200 mark.
Talking about AMD’s top-end processors we have to limit ourselves to the Socket AM3 platform only. Today, AMD supports but one platform and offers CPUs of very different specs and prices for it. This is the reason why all modern processors from AMD have to be equipped with a dual-channel DDR2/DDR3 controller, but that’s okay for the top-end segment we are discussing today.
The problem with AMD’s products is that they are based on the Stars microarchitecture which has a very low IPC parameter by today’s standards (IPC stands for Instructions per Clock). Therefore even those of AMD processors that work at high frequencies prove to be not as fast as their opponents from Intel. Besides, AMD has not yet been able to implement 32nm tech process, which has a negative effect on the power consumption and heat dissipation of its products.
All of this makes AMD’s position in the top-end segment rather weak. There are actually only two CPU models from AMD in this review and both belong to the same series.
Phenom II X6. The six-core CPU series is a rather recent addition to AMD’s product line-up. A special monolithic semiconductor die with improved energy efficiency was developed for it, but the manufacturing process is still 45 nanometers. The clock rates of the senior Phenom II X6 models that fall into the top price segment are 3.0 to 3.2 GHz. They are additionally equipped with Turbo Core technology which can raise the clock rate to 3.5-3.6 GHz if at least three out of the six cores are idle. The rest of the specifications are ordinary enough. Each core has 512 kilobytes of L2 cache. Moreover, each Phenom II X6 has 6 megabytes of shared L3 cache. The integrated memory controller supports DDR2 and DDR3 memory at frequencies up to 1600 MHz. The top-of-the-line model Phenom II X6 1090T is a Black Edition modification and can be overclocked by increasing its frequency multiplier.
It may seem quite natural that the Phenom II X6 series is positioned among expensive CPUs. After all, Intel offers its six-core CPUs in the top price segment, too, and offers them at a much higher price. However, when you take a look at the test results, you realize that the Socket AM3 platform doesn’t have bright perspectives in the top market segment. Therefore it is no wonder that AMD’s six-core processors are quickly getting cheaper under the pressure from Intel. In other words, the Socket AM3 platform may soon leave this market segment altogether. We are only expecting one new model in the Phenom II X6 series and then these CPUs will steadily go down to the mainstream, leaving their place for new products with the Bulldozer microarchitecture which are promised by AMD in the second quarter of the next year.
Anyway, Socket AM3 can be viewed as a top-end platform as yet. There are expensive and feature-rich mainboards for Socket AM3 processors which provide broad opportunities for assembling high-performance computers. For example, there are hardcore gamers oriented Socket AM3 mainboards which allow building graphics subsystems out of three or four graphics cards!
The next table shows the specifications of the few AMD processors that fall into the top-end category and took part in our tests.
So, it is Intel that will play the first fiddle in the top-end market segment. And, surprisingly enough, the company offers not only different CPU models but also different platforms here. The LGA1336 platform is positioned as the main top-end solution. On the other hand, you can buy a $200 and more expensive CPU even for LGA775 or LGA1156. The internal competition and the availability of all the three platforms in the top price segment is an indication of certain flaws in the company’s marketing policy, but here two out of the three platforms can coexist thanks to their unique features.
The benefits of LGA1366 solutions are obvious. It is exclusively for this platform that Intel offers six-core CPUs whereas quad-core LGA1366 processors differ from their LGA1156 counterparts in having a triple-channel DDR3 SDRAM controller. On the other hand, this platform is not without downsides for all its elite status. Coming to desktop PCs from the world of servers and workstations, it doesn’t have a suitable system architecture: the controller of the graphics PCI Express bus resides in the chipset which calls for a high-speed QPI bus to connect the CPU to the chipset. The addition of two QPI controllers into the main chips makes LGA1366 systems consume more power in comparison with LGA1156 computers.
Core i7-900. This series combines rather variegated processors which are designed for the LGA1366 platform. In fact, there are two varieties of products with the Nehalem microarchitecture within this series. The first variety is quad-core CPUs based on 45nm dies, and the second variety is six-core CPUs with 32nm dies. The quad-core models have clock rates of 3.06 to 3.2 GHz, 8 megabytes of L3 cache, and support triple-channel DDR3 SDRAM. They feature all the performance-boosting technologies available today: Hyper-Threading for virtual CPU cores and Turbo Boost for automatic overclocking. The six-core CPUs have higher frequencies (3.2-3.33 GHz) and 12 megabytes of cache. Otherwise, they are similar to their quad-core LGA1366 counterparts in their specs. Particularly, they also have a triple-channel memory controller and support Hyper-Threading and Turbo Boost.
It must be noted that there is a very slim separating line between the quad- and six-core CPUs for LGA1366 systems in terms of model names. The model names up to i7-960 are quad-core CPUs whereas the i7-970 and higher models are six-core ones. The senior Core i7-980X is additionally referred to as Extreme Edition which means that you can overclock it by changing its frequency multiplier.
Interestingly, the LGA1366 platform is likely to have the longest lifecycle among all the platforms available today. It is going to be replaced no sooner than in the end of the next year. Therefore this platform can be recommended for users who like to upgrade their computers regularly.
The LGA1156 platform looks like a more mature solution than LGA1366 because it came out later and its designed was optimized by integrating the PCI Express controller into the CPU. However, the senior CPU models for this platform have lower frequencies and are equipped with a dual-channel memory controller. Most importantly, they only offer one PCI Express x16 but which, even though splittable in two, limits the performance of multi-GPU graphics subsystems. As a result, many LGA1156 processors, even though top-end solutions, are positioned lower in the product hierarchy than their LGA1366 counterparts.
Core i5-600. These are dual-core LGA1156 processors based on a 32nm semiconductor die that additionally incorporates a graphics core die. Such processors are identified by the OS as quad-core ones because they support Hyper-Threading which enables each physical core to execute two instructions threads simultaneously. Top-end models from the Core i5-600 series have clock rates of 3.46 to 3.6 GHz and 4 megabytes of L3 cache. They also support Turbo Boost and to automatically increase the clock rate when one core is idle. The integrated memory controller is compatible with dual-channel DDR3 SDRAM at 1067 or 1333 MHz.
Core i5-700. The junior series of quad-core LGA1156 processors also has its representation in the above-$200 category. This is the Core i5-760 model which is based on a 45nm die and has a clock rate of 2.8 GHz. It supports Turbo Boost but doesn’t have Hyper-Threading. Therefore, like the Core i5-600, it is identified as a quad-core CPU by the operating system. The Core i5-670 has 8 MB of L3 cache and its integrated memory controller supports dual-channel DDR3-1067 and DDR3-1333. As opposed to the 600 series, the 700 series have no integrated graphics core.
Core i7-800. These quad-core LGA1156 processors are the masters of the top-end market segment. Although based on the same semiconductor die as the Core i5-700, they have higher clock rates (2.8 to 3.06 GHz) and support both Turbo Boost and Hyper-Threading. As a result, a Core i7-800 is identified by the OS as an eight-core processor. Besides, the Core i7-800 features a faster memory controller which supports not only DDR3-1067 and DDR-1333 but also DDR3-1600. The Core i7-800 has the same amount of cache as the Core I5-700, which is 8 megabytes.
Right now, the LGA1156 platform looks highly appealing but it is getting closer to its demise. In a couple of months it will be replaced with the new LGA1155 platform which will support more advanced CPUs working at higher clock rates. The two platforms will be incompatible and there will be no LGA1156 processors faster than what is already available.
As for the LGA775 platform, there is no reason for it to be in the top-end market sector in late 2010. We can only view it as a reminder of the highly successful Core 2 Duo and Core 2 Quad series which helped Intel beat its opponent in terms of CPU performance back in 2006. In other words, the Core microarchitecture of the LGA775 processors is just as outdated as the Stars one, but Intel, unlike AMD, does not even try to increase the clock rates of its old CPUs and offers 2- or 3-year-old products at high prices.
Core 2 Duo. The dual-core LGA775 CPU series based on a 45nm core is available in the top price segment, too. There is but one model, Core 2 Duo E8600. This CPU has a clock rate of 3.33 GHz and a 333MHz system bus. It also has a shared 6MB L2 cache.
Core 2 Quad. There are several quad-core LGA775 processors in this market segment. These are models with clock rates of 2.83 to 3.0 GHz and a 333MHz system bus. By the way, a quad-core LGA775 processor is in fact two dual-core semiconductor dies within a single package. Therefore the L2 cache, which is 12 MB large, is divided by half between the two pairs of CPU cores.
The next table shows the specifications of all top-end CPUs from Intel available today.
Simultaneous testing of a large number of processors is a pretty complicated task that not only takes a lot of time and effort, but also requires access to dozens of different processor models at the same time. Therefore, most results databases available online are either updated over a large period of time, which causes the testbed configuration, drivers and applications versions to change accordingly, or have the results obtained on the same old platform that hasn’t been updated for a while. We did our best to avoid any of these problems – all our tests are current and have been performed at the same time for all CPUs. We used only the latest hardware components with the most recent BIOS and driver versions as well as the operating system with all available updates.
Here is the list of hardware and software components we used for our today’s test session:
To estimate the processors performance in general-purpose apps we use SYSmark 2007 test that emulates usage models in popular office and digital content creation and processing applications. The idea of this benchmark is very simple: it produces a single score that characterizes average system performance.
We can see Intel’s new microarchitectures being indeed superior to the older solutions: the LGA1156 and LGA1366 processors are faster than their LGA775 and Socket AM3 opponents. Even the dual-core models like the Core i5-670 are easily ahead of the six-core products from AMD in SYSmark 2007. So, it is no wonder the Phenom II X6 is so inexpensive. This series is only competitive to the Core 2 Duo and Core 2 Quad series in practical applications.
The dual-core Core i5-670 and Core i5-680 CPUs are indeed very fast in SYSmark 2007 because most of popular real-life applications use only two CPU cores. They also have high clock rates and feature Hyper-Threading for multithreaded applications.
But it is the Core i7 series for both LGA1366 and LGA1156 that wins in SYSmark 2007. The LGA1366 platform justifies its higher positioning, and nearly all of the LGA1366 processors prove to be faster than their LGA1156 counterparts. For example, the senior model of the Core i7 800 series is but slightly ahead of the Core i7-950, but that’s hardly a surprise considering their prices.
The table below contains the detailed scores from the SYSmark 2007 suite sorted according to the application type:
Easy to see, the high overall scores of the dual-core processors Core i5-670 and Core i5-680 are explained by their high performance in office and image-editing applications. The CPUs with more cores are ahead at processing video content and 3D modeling.
As you know, it is the graphics subsystem that most often determines the overall performance of a computer in modern games. Therefore we run gaming tests in our CPU reviews without full-screen antialiasing and at low resolutions. So, the results are not indicative of how fast the particular game can run on modern computers but rather how well the tested CPUs can cope with gaming load. This can give us some insight into the future when graphics cards will get faster, making the CPU the limiting factor.
We’ve got expected results here. Top-performance CPUs are not really necessary in most games because the frame rate is going to be limited by the graphics card’s speed. Even at low resolutions and without FSAA the Core i7 series processors deliver close results. In other words, top-of-the-line CPUs, like Intel’s six-core ones, may only be interesting in some particular situations. They only show their full potential in a few games and only if you have a high-performance multi-GPU graphics subsystem. Generally speaking, each Core i7 series processor is more than enough for any gaming configuration.
To test the processors performance during data archiving we resort to WinRAR archiving utility. Using maximum compression rate we archive a folder with multiple files 560 MB in total size.
There are a lot of factors affecting the performance of WinRAR: the CPU’s clock rate, the number of CPU cores, the mount of cache memory, etc. However, this time the diagram looks quite simple. We’ve got LGA1366 models at the top, the quad-core LGA1156 ones in the middle, and the quad-core LGA775 at the bottom. The only disappointment is that the Phenom II X6 is only capable of beating its dual-core opponents.
The processor performance during encryption is measured with an integrated benchmark from a popular cryptographic utility called TrueCrypt. I have to say that it can not only effectively utilize any number of processor cores, but also supports special AES instructions.
Encryption can easily be paralleled, therefore we’ve got six-core processors from both AMD and Intel in the lead. Intel’s ones are faster due to their more advanced architecture and support for the AES instruction set.
We measured the performance in Adobe Photoshop using our own benchmark made from Retouch Artists Photoshop Speed Test that has been creatively modified. It includes typical editing of four 10-megapixel images from a digital photo camera.
There are no surprises here, either. The LGA1366 and LGA1156 processors are ranked up according to their model numbers. The LGA775 oldies can’t compete with the newer products whereas the six-core processors from AMD are slow.
We use Apple iTunes utility to test audio transcoding speed. It translates the contents of a CD disk into AAC format. Note that the typical peculiarity of this utility is its ability to utilize only a pair of processor cores.
Apple iTunes uses two CPU cores only, which explains the results. The dual-core Core i5 series processors with higher clock rates are in the lead.
In order to measure how fast our testing participants can transcode a video into H.264 format we used x264 HD benchmark. It works with an original MPEG-2 video recorded in 720p resolution with 4 Mbps bitrate. I have to say that the results of this test are of great practical value, because the x264 codec is also part of numerous popular transcoding utilities, such as HandBrake, MeGUI, VirtualDub, etc.
In our previous reviews AMD processors used to be good enough at transcoding video with the x264 codec. But when it comes to the top-end products, the Phenom II X6 falls behind Intel’s six-core CPUs as well as many of the quad-core Core i7 series models that support Hyper-Threading.
The performance in Adobe Premiere Pro is determined by the time it takes to render a Blu-ray project with HDV 1080p25 video into H.264 format and apply different special effects to it.
We’ve got the same standings in the Adobe Media Encoder test as when using the x264 codec. Summing up these results, we can say that the six-core Core i7 series is most appropriate for video editing. Installing a Core i7-970 or a Core i7-980X instead of an expensive quad-core model may boost the computer’s performance in such applications by 50% and more! This is quite normal as video processing algorithms can effectively run in multiple parallel threads.
We launch standard MathematicaMark7 test from Wolfram Mathematica suite to measure the systems’ performance here.
We’ve got expected results again. The Core i7 series has no competition. Within this series, the CPU models perform according to the official price list.
We use special Cinebench test to measure the final rendering speed in Maxon Cinema 4D.
The more cores, the higher the performance is in this test. This explains the results if you take into account that Hyper-Threading allows executing two instruction threads on a single physical core.
Rendering speed in Autodesk 3ds max 2011 with both, Scanline as well as Mental Ray, was measured using SPECapc test.
The speed of rendering in 3ds max 2011 is similar to the Cinebench results. We can again see that AMD’s six-core CPUs can only compete with Intel’s quad-core solutions and that the LGA775 products are outdated.
Performance is not the only practical spec that may be of interest to potential buyers of mainstream CPUs. In many cases their power consumption matters a lot, as it has direct connection not only to the amount of your next power bill. The same parameter sets certain restrictions and criteria when it comes to picking out a system case. Therefore, we decided to add power consumption tests to our performance research.
The graphs below show the full power draw of the computer (without the monitor) measured after the power supply. It is the total of the power consumption of all the system components. The PSU's efficiency is not taken into account. The CPUs are loaded by running the 64-bit LinX 0.6.4 utility. We enabled all the power-saving technologies for a correct measurement of the computer's power draw in idle mode: C1E, AMD Cool'n'Quiet and Enhanced Intel SpeedStep.
There are three groups of CPUs in terms of power consumption in idle mode. The most energy-efficient group includes the whole LGA1156 platform. The second group combines somewhat less efficient CPUs from the LGA775 and Socket AM3 series. The LGA1366 platform has the highest power consumption when idle, which might be expected considering its architectural specifics.
The picture is somewhat different at load even though the LGA1366 processors still consume the most. The dual-core Core 2 Duo and Core i5 processors expectedly need the smallest amount of power. As for the Socket AM3 platform and Phenom II X6 processors, they are similar to the senior quad-core LGA1156 processors in terms of power consumption.
Although there were a lot of processor models participating in our today’s test session, we hardly witnessed any cut-throat battles. Unlike mainstream and value segments, the high-end price range is extremely broad, so the CPUs that fall into it may have dramatically different prices and rarely compete directly against one another. In other words, the price is often the main criterion determining the CPU choice in this segment. The general rule, however, is the following: more expensive processors run faster.
However, like any other rule, this rule also has exceptions. Firstly, if you are looking to buy a processor priced beyond $200, then LGA775 shouldn’t be even considered. They are obviously outdated and cannot compete against newer models with dramatically higher performance.
Secondly, you should also be careful about expensive dual-core solutions in the Core i5 lineup. Most contemporary resource-hungry tasks, which may require a more expensive CPU, can be easily split into parallel threads. Therefore, quad-core processors are generally a better choice, so that even high clock rates don’t help the top Core i5-670 and Core i5-680 to win you over. So, these processors may only be appealing for two reasons: they may suit for some specific applications that do not utilize all the advantages of multi-core CPUs for some reason, or they may be appealing to those users who value energy-efficiency over performance.
In order to illustrate the correlation between the consumer qualities of the tested processors in the high-end price segment, we put together the following diagram showing the average performance vs. price.
As you can see, Core i7-950 appears to be one of the best processors in the upper price segment in terms of price-to-performance ratio. While its price is quite reasonable (for a high-end product), this quad-core processor is not only faster than many of its competitors. Its another indisputable advantage is that it is designed for LGA1366 platform. This platform not only has great future ahead, but also offers everything necessary for building high-speed graphics sub-systems of any potential.
However, LGA1366 mainboards are pretty expensive and a system with an LGA1366 processor will be far from energy-efficient, that’s for sure. If these aspects matter a lot to you, then you should check out Core i7-870. Its performance is quite competitive with that of Core i7-950, but at the same time it is much more energy-efficient and doesn’t require an expensive mainboard. Although you will have to put up with certain limitations when it comes to configuring your multi-card graphics sub-system, but I am sure not everyone needs them anyway.
Summing up everything we have just said, we would like to award Intel Core i7-950 and Core i7-870 processors with our Recommended Buy title:
Six-core processors are also worth your immediate attention. They are the low-cost Phenom II X6 and expensive Core i7-970 and Core i7-980X. Of course, we wouldn’t recommend them for regular high-performance home systems, their average price-to-performance ratio doesn’t look very appealing to the end-users. However, there are a lot of specific applications where six fully-fledged cores may become very handy. Among such tasks we could name high-definition video processing, final rendering and some other similar applications that require a lot of calculations. So, if you are going to use your computer for tasks like that, then you should definitely bet on one of the six-core products.