The whole truth about multi-core processors. Computer resource U SM What is a processor core and multi-core

Intel, in order to remain the leader in the processor market, steadily continues to follow its “Tick-Tock” concept, approximately every two years transferring production to a new, thinner technical process (“Tick”), and a year later introducing a new architecture, which is produced using already mastered technical process (“So”). So, just over a year ago, the Nehalem architecture for desktop processors was introduced to the world, the most powerful and expensive of which use a 45 nm Bloomfield core. And now the time has come to transfer the production of “top” processors to a new technical process, which, by the way, has already been successfully tested on mass-produced processors with the Clarkdale core that were introduced for the New Year. However, in these models with a built-in graphics core, only the computing part was produced according to 32 nm standards, and you need to master the technical process in order to produce full-fledged processors.

And so, by transferring the production of processors with the Nehalem architecture to the 32 nm process technology, Intel decided not only to repeat the same thing, but with a smaller element size and increase the operating frequency, as was usually the case before. This time the updated processor also received noticeable architectural changes - it became six-core. Of course, the Nehalem architecture itself has undergone virtually no changes, but simply the new processors, codenamed Gulftown, include two more of the same computing cores as in Bloomfield.

In parallel with the increase in the number of cores, the volume of third-level cache was increased by one and a half times, which is now 12 MB. Moreover, the L3 cache memory still operates using Smart Cache technology, i.e. is integral and can be distributed dynamically between cores depending on their needs, up to the point that it will be captured by one of the busiest computing cores.

But one small expansion of capabilities was also made - finally, support for acceleration instructions for the AES encryption algorithm, which have been implemented in mass-produced dual-core processors with the Clarkdale core, has been implemented for the “top” processors. Otherwise, the Gulftown core is exactly the same as the Bloomfield, the features of which are described in more detail in the review of the Intel Core i7-920 processor, even the built-in three-channel memory controller officially supports working only with DDR3-1066 modules. Naturally, the new processors based on the Gulftown core use exactly the same Intel LGA 1366 processor socket, communicate with the system using the QPI bus, support the same set of proprietary technologies and can be installed in motherboards based on the Intel X58 Express chipset (the main thing is to remember to update first BIOS).

True, for now, when we talk about new processors based on the Gulftown core in the plural, we mean only one model, which has a very high cost and is intended for enthusiasts. More affordable mass models will appear later. Well, while waiting for less expensive six-core processors to appear, we will study the capabilities of the Nehalem architecture transferred to 32 nm, expanded and slightly updated.

Our testing laboratory received an engineering sample of the Intel Core i7-980X Extreme Edition processor in a box without printing, although the packaging dimensions themselves fully correspond to the retail version. Moreover, in terms of dimensions, this box has become almost twice as large as the packaging of previous models of Core i7-900 series processors. The thing is that now the “top” processor also comes with a corresponding cooler.

Finally, Intel has met the needs of buyers of very expensive Extreme Edition processors, offering them a good proprietary cooling system - Intel DBX-B Thermal Solution. We will definitely take a closer look at this cooling system and study its capabilities. In addition to the processor and cooler, inside the box the buyer will have to find a user manual, warranty statements and a company sticker.

Let's move on to consider the technical characteristics of the Intel Core i7-980X Extreme Edition processor.

Specification:


Marking
CPU socket
Clock frequency, MHz
Factor
Bus frequency, MHz
L1 cache size (Data\Instructions), KB
L2 cache size, KB
L3 cache size, MB

Number of Cores
Instructions support	MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AES, EM64T
Throughput QPI, GT/s
Supply voltage, V
Power dissipation, W
Critical temperature, °C
Technical process
Technology support	Enhanced Halt State (C1E) Enhanced Intel Speedstep Technology Hyper-Threading Technology Execute Disable Bit Intel Virtualization Technology Intel Turbo Boost Technology
Memory controller specification
Maximum memory capacity, GB
Memory types
Number of memory channels
Maximum throughput, GB/s
ECC support

Studying the specification of the Intel Core i7-980X Extreme Edition, it is interesting to note that when switching to a new technological process, an increase in operating frequencies was not provided, because the previous top-end processor Intel Core i7-975 Extreme Edition operates at exactly the same nominal frequency of 3.33 GHz. Apparently that's why the Intel Core i7-980X Extreme Edition has only a slightly larger model number.

Please also note that, unlike regular (not extreme) processors of the Intel Core i7-900 series, the Intel Core i7-980X Extreme Edition processor, like all Intel Core i7 Extreme Editions, uses a faster QPI bus operating mode - 6.4 GT/s instead of 4.8 GT/s, which should speed up data exchange with the system a little.

The heat dissipation cover of a retail processor, unlike an unremarkable engineering sample, will have to indicate the model, sSpec number, country of origin, as well as technical information:

frequency – 3.33 GHz;
L3 cache size – 12 MB;
QPI bus clock frequency – 6.4 GT/s;
Compatibility requirements – PCG (Platform Compatibility Guide) 08.

As you would expect, the number and location of matching elements on the back of the processor is radically different from other models of the Intel Core i7-900 family.

Having finished with the external examination of the Intel Core i7-980X Extreme Edition processor, let's take a look at it, so to speak, from the inside, using the CPU-Z information utility.

As you can see, the utility quite correctly visualizes the declared technical characteristics and shows some other interesting details. In addition to the increased number of computing cores to 6, and thanks to support for Hyper-Threading technology with the ability to simultaneously execute up to 12 program threads, the Intel Core i7-980X Extreme Edition processor has a one and a half times increased third-level cache memory - up to 12 MB. It's very interesting to look at the organization of this extended cache memory.

Unfortunately, the architecture of the L3 cache has not changed - the same 16 association lines of 64 bytes each, as in models with 8 MB. In this case, theoretically, an increase in cache memory size of 50% led to a slowdown of 33% with other parameters unchanged. In addition, in order to reduce the power consumption of the processor, and it remained within the thermal package of up to 130 W, the operating frequency and supply voltage for the Uncore logic, including the built-in memory controller, were slightly reduced. Let's say right away that low-level synthetic tests perfectly record an increase in the latency of the third level cache and RAM, but it is much more interesting to see in more practical and universal tests how critical such a slight slowdown of memory and cache memory is with a noticeable increase in the volume of the latter, as well as the addition the processor has two more computing cores. We will try to answer this question during the testing process.

Separately, it is worth mentioning the operation of the processor's memory controller: it officially supports only three-channel DDR3 memory modules at frequencies up to 1066 MHz. Even a kernel update did not change the situation. However, it is not documented that it is possible to use the Intel Core i7-980X Extreme Edition processor in conjunction with high-frequency DDR3 memory modules ranging from DDR3-1333 and, thanks to the free divider, to probably the fastest DDR3-2533 to date. We were unable to check the latter, but the modules available in the test laboratory started without problems at an effective frequency of 1866 MHz.

Concluding the story about the declared capabilities of the Intel Core i7-980X Extreme Edition processor, we should remind you of the support for the following proprietary technologies from Intel:

Enhanced Halt State (C1E) turns off some processor blocks during inactivity, thereby reducing power consumption and heat dissipation;

Enhanced Intel Speedstep Technology allows you to reduce the supply voltage and clock frequency during low load on the processor;

Execute Disable Bit – support for hardware and software protection against buffer overflow, a mechanism used by many malware to cause damage or penetrate the system;

Intel Virtualization Technology enables virtual machines to access hardware resources;

Hyper-Threading Technology – each core of the Intel Core i7 processor supports the simultaneous execution of two program threads;

Intel Turbo Boost Technology - allows you to increase the processor multiplier depending on the load, in fact it is a dynamic overclocking function, but without a noticeable increase in power consumption, which is limited by the declared thermal package, and heat dissipation.

During testing we used Processor Test Stand No. 1

Motherboards (AMD)	ASUS M3A32-MVP DELUXE (AMD 790FX, sAM2+, DDR2, ATX)GIGABYTE GA-MA790XT-UD4P (AMD 790X, sAM3, DDR3, ATX)
Motherboards (AMD)	ASUS F1A75-V PRO (AMD A75, sFM1, DDR3, ATX)ASUS SABERTOOTH 990FX (AMD 990FX, sAM3+, DDR3, ATX)
Motherboards (Intel)	GIGABYTE GA-EP45-UD3P (Intel P45, LGA 775, DDR2, ATX)GIGABYTE GA-EX58-DS4 (Intel X58, LGA 1366, DDR3, ATX)
Motherboards (Intel)	ASUS Maximus III Formula (Intel P55, LGA 1156, DDR3, ATX)MSI H57M-ED65 (Intel H57, LGA 1156, DDR3, mATX)
Motherboards (Intel)	ASUS P8Z68-V PRO (Intel Z68, sLGA1155, DDR3, ATX)ASUS P9X79 PRO (Intel X79, sLGA2011, DDR3, ATX)
Coolers	Noctua NH-U12P + LGA1366 KitScythe Kama Angle rev.B (LGA 1156/1366)ZALMAN CNPS12X (LGA 2011)
RAM	2x DDR2-1200 1024 MB Kingston HyperX KHX9600D2K2/2G2/3x DDR3-2000 1024 MB Kingston HyperX KHX16000D3T1K3/3GX
Video cards	EVGA e-GeForce 8600 GTS 256 MB GDDR3 PCI-EASUS EN9800GX2/G/2DI/1G GeForce 9800 GX2 1GB GDDR3 PCI-E 2.0
HDD	Seagate Barracuda 7200.12 ST3500418AS, 500 GB, SATA-300, NCQ
power unit	Seasonic SS-650JT, 650 W, Active PFC, 80 PLUS, 120 mm fan

Select what you want to compare Intel Core i7-980X EE with

As you can see, the six-core processor operating at 3.33 GHz confidently outperforms all models we have previously tested. But whether you can experience this increase in performance will greatly depend on the tasks you perform. Thus, in mathematical, some multimedia packages and applications for three-dimensional modeling, it will be possible to obtain a noticeable acceleration. But in the vast majority of computer games, using a six-core processor will be of little use, although it will be possible to quite painlessly run some demanding application in parallel with the game, for example, video transcoding or a full scan with an antivirus.

Real benefits of six cores: Bloomfield vs. Gulftown

When testing the Intel Core i7-980X Extreme Edition processor at the nominal frequency, we, unfortunately, were not able to unambiguously and completely answer how much a six-core processor with an increased amount of third-level cache memory is superior to a four-core processor with almost the same architecture, because The compared models operated at different clock frequencies. But given that older models with four and six cores operate at the same frequency, it is likely that more affordable models on the Gulftown core expected in the near future will compete with solutions of equal frequency on the Bloomfield core. To test this, we slowed down the Intel Core i7-980X Extreme Edition processor to the frequency of the Intel Core i7-950 that was in our test lab.

After running a series of standard tests, we got this result:

Test package		Result		Change in productivity, %
Test package		Intel Core i7-950	Intel Core i7-980X @3.06 GHz	Change in productivity, %









	Rendering, CB-CPU





	DirectX 9, High, fps
	DirectX 10, Very High, fps

Performance in various applications depends on many parameters, including the characteristics of the applied algorithms, as well as optimization for multi-threaded execution. This is probably why we recorded a serious spread of values - from a small negative result, most likely due to poor optimization for execution on multi-core processors and a large dependence on the speed of cache memory and RAM, to a fairly impressive increase in performance almost reaching a theoretical +50% due to a well-implemented algorithm with support for parallel computing. But on average, the Gulftown core was only ≈12% faster than Bloomfield. This is exactly the kind of system acceleration that average users who switch from a quad-core to a six-core processor will be able to get in the near future, although in the professional sphere the effect of replacing the processor will be much greater.

Using faster RAM

We have already established that a six-core processor will not always provide a noticeable acceleration in task execution, and some slowdown in the third level cache and the built-in memory controller is partly to blame for this. On the other hand, at least with the Intel Core i7-980X Extreme Edition, you can install fairly fast memory modules into the system, exceeding the speed of “standard” DDR3-1333.

We have already shown above that in practice the system worked stably with DDR3-1866, although such and faster modules have a noticeably higher cost than DDR3-1333. That is why we did not conduct experiments using clearly overclocking frequencies for memory modules, but limited ourselves to the frequency of 1600 MHz, at which more affordable and common modules operate, sometimes even without radiators. After all, it is DDR3-1600, as it seems to us, that will be most relevant in the near future, when affordable six-core processors become available for sale. But will this lead to a noticeable acceleration of the system?

Test package		Result	Productivity gain, %
Test package			Productivity gain, %









	Rendering, CB-CPU

Fritz Chess Benchmark v.4.2, knodes/s



	DirectX 9, High, fps
	DirectX 10, Very High, fps

Judging by the results obtained, from using faster DDR3-1600 modules, at best, you can expect a performance increase of 5-7%, although on average this is 1-2%. Even if you use more expensive sets with aggressive timings, this will not change the situation much. Perhaps this is why officially only DDR3-1066 support is still declared for Intel Core i7 processors under LGA 1366. But, nevertheless, if mass-produced six-core processors are able to work with memory modules faster than DDR3-1333 without overclocking and the latter also have an affordable price, they will provide some slight increase in performance.

How Intel Turbo Boost Technology works

While the ability to use fast memory modules is optional, and not yet guaranteed for mass models, then all Intel Core i7 processors will be equipped with support for Intel Turbo Boost technology. Let us recall that Intel Turbo Boost technology provides intelligent adjustment of processor performance to the user's needs by slowing down unloaded cores and slightly accelerating the rest, without a noticeable increase in power consumption (within the thermal envelope). Thus, poorly parallelized tasks are executed a little faster. In addition, Intel Turbo Boost has an acceleration mode by increasing the multiplier by one step, i.e. at 133 MHz of all computing cores, which in any case guarantees some increase in performance, the main thing is not to forget to activate Intel Turbo Boost in the BIOS.

For six-core processors, the acceleration formula now looks like 1/1/1/1/2/2. That is, when one or two cores are loaded, their frequency increases by 2x to 3.6 GHz, naturally with the rest slowing down, and in all other cases the processor will become faster by 133 MHz. However, do not forget that the processor will begin to consume a little more electricity.

Let's try to estimate how much acceleration the system will receive after enabling Intel Turbo Boost technology.

Test package		Result		Productivity gain, %
Test package		Intel Turbo Boost OFF	Intel Turbo Boost ON	Productivity gain, %









	Rendering, CB-CPU

Fritz Chess Benchmark v.4.2, knodes/s



	DirectX 9, High, fps
	DirectX 10, Very High, fps

The effectiveness of enabling Intel Turbo Boost in most tasks exceeds the benefit of installing faster memory modules, and this does not require any additional costs, and the technology itself will be guaranteed for all processors.

In general, it is recommended to leave Intel Turbo Boost technology always on, because in idle mode the core frequency and supply voltage will still decrease, and a slight increase in power consumption under load will not be a problem even if you are using a “boxed” cooler. And in this case, thanks to the “boxed” Intel DBX-B Thermal Solution, you can try to get good overclocking results.

Overclocking Intel Core i7-980X Extreme Edition

Holding a processor with a free multiplier, such as the Intel Core i7-980X Extreme Edition, the easiest and most accessible way to overclock seems to be to increase the multiplier, although this is not the most optimal mode. We decided to try out various options, but first we found out what results can be obtained if we simply increase the processor multiplier, naturally ensuring stability at an increased frequency by slightly increasing the supply voltage.

In this simple and convenient way, we managed to achieve stability from the Intel Core i7-980X Extreme Edition with a x31 multiplier, i.e. at a frequency of 4125 MHz, which is almost 24% more than the nominal frequency. Unfortunately, it was not possible to make the processor work with a x32 multiplier even at a higher core voltage. But +24% should provide a noticeable acceleration of the system.

Test package		Result		Productivity gain, %
Test package		Rated frequency	Overclocked processor	Productivity gain, %









	Rendering, CB-CPU

Fritz Chess Benchmark v.4.2, knodes/s



	DirectX 9, High, fps
	DirectX 10, Very High, fps

As we can see, in a number of tasks the increase in system performance is almost directly proportional to the frequency of the processor, but in complex tasks the acceleration is not so great and on average was only ≈13.5%. In general, this result is quite expected, because many resource-intensive applications are also dependent on other computer subsystems.

Therefore, we tried to achieve the same frequency of 4.12 GHz by increasing the reference frequency, which leads to acceleration of all buses and the memory controller built into the processor, as well as the memory modules themselves. Since in this situation not only the frequency of the computing cores has increased, but also of all other nodes, we can expect a noticeably greater increase in performance.

Test package		Result		Productivity gain, %
Test package		Rated frequency	Overclocked processor	Productivity gain, %









	Rendering, CB-CPU

Fritz Chess Benchmark v.4.2, knodes/s



	DirectX 9, High, fps
	DirectX 10, Very High, fps

Now you can notice an increase in productivity in almost all tasks: the average increase in performance was 18.6%. Thus, it is quite obvious that the presence of a free multiplier in the processor only adds flexibility when overclocking.

The result of comparing different overclocking methods will be the conclusion that overclocking using a multiplier is the simplest and most affordable, but will be more acceptable when using less expensive processors with a free multiplier, for example, Intel Core i5-655K or Intel Core i7-875K. For a professional who wants to get the most out of overclocking a very expensive model, there is practically no benefit from a free multiplier, because overclocking by increasing the frequency of the system bus and all associated nodes and components provides the greatest performance increase.

But during overclocking, the power consumption of the processor also changes, which must be taken into account:

System power consumption	Nominal mode with energy saving technologies enabled	Nominal mode with energy saving technologies turned off	Overclocking the processor to 4.2 GHz at a supply voltage of 1.4 V
System downtime, W

Load using stress test in EVEREST, W

Overclocking the processor by 26% significantly increased the processor's power consumption, and therefore its heat dissipation. It's nice to note that we carried out all these experiments using the Intel DBX-B Thermal Solution cooler included with the processor.

Intel DBX-B Thermal Solution Bundled Cooling System

As has already been mentioned more than once throughout the review, a feature of the “top” six-core processor is the powerful Intel DBX-B Thermal Solution cooler with copper heat pipes. It is this cooling system that should allow experiments with overclocking this processor. This step is very important because... Previously, “extreme” processors were equipped with ordinary, simple coolers, which the buyer of a fairly expensive processor often simply threw away when purchasing a cooler worthy of the processor. Let's take a closer look at the design features of the Intel DBX-B Thermal Solution and evaluate its effectiveness.

The Intel DBX-B Thermal Solution cooler is based on four 6mm heat pipes that accelerate heat transfer from the copper base to a dense block of aluminum fins.

The heat pipes themselves are laid in deep grooves in the base, and the contact is improved using solder. In most cases, this heat sink design is the most optimal.

Moreover, to improve efficiency, the ribs were fixed using hot-melt adhesive. This makes the cooler design quite high quality and reliable.

However, the heatsink of the Intel DBX-B Thermal Solution cooling system seems too dense, because in it, fairly wide plates with a thickness of 0.5 mm are mounted with an indentation of 1.0 mm. This design will require the fan used to be able to generate a high enough static pressure for the system to be truly effective. In addition, a small gap between the plates will contribute to the accumulation of dust there, which will reduce the efficiency of the cooler over time.

To ensure high performance, the radiator is equipped with a 100 mm F10T12MS2Z9 fan manufactured by NIDEC, whose nine translucent blades with a high angle of attack are capable of rotating at speeds of up to 2600 rpm. Moreover, part of the air flow at the very bottom passes under the radiator, providing ventilation of the “near-socket” space.

The fan has a 4-pin power connector, i.e. Supports dynamic PWM rotation speed control. But to accurately set operating modes, the cooler has a switch between quiet and performance modes. In quiet mode, the fan rotates at speeds of up to 1800 rpm and creates an average noise level, not particularly highlighting the Intel DBX-B Thermal Solution inside the system unit. In performance mode, the rotation speed can increase to 2600 rpm and the cooler becomes very noisy.

The base of this “boxed” cooler is also very well processed - polished to a mirror finish. But the shape of the base is not quite optimally chosen - it is rectangular 31x37 mm. In our test system, the most complete contact of the cooler with the processor was only in the case when the air was vented towards the power supply, which was not entirely optimal.

To install the Intel DBX-B Thermal Solution cooler, a plastic stop plate is used, i.e. It will not be possible to secure the cooling system without removing the motherboard from the system unit. To facilitate the installation process, there are two adhesive strips on the frame, with which it is simply glued to the motherboard, and there is no need to hold the frame while screwing the cooler. The cooling system itself is fixed using “stationary” screws with a large head. Thus, the Intel DBX-B Thermal Solution cooler is installed quite simply and quickly, even by hand, although to ensure good pressure to the processor it is advisable to finally fix it with a screwdriver.

To evaluate the effectiveness of the Intel DBX-B Thermal Solution, we suggest comparing it under the same conditions (overclocking the Intel Core i7-980X Extreme Edition processor to 4.1 GHz at a core voltage of 1.36 V) with several high-performance coolers: Scythe Kama Angle, Noctua NH -U12P, Noctua NH-U12P SE2, Noctua NH-U9B and Noctua NH-U9B SE2.

In high performance mode, the Intel DBX-B Thermal Solution delivers performance even higher than some established cooling leaders. However, not everything is so rosy - the noise is noticeably higher than a comfortable level. But if you are conducting overclocking experiments, then the Intel DBX-B Thermal Solution will help you with this and, most likely, you will not want to replace it. And for continuous operation, you can reduce the overclocking level and switch the cooler to quiet mode. Of course, it will not become silent, but it will no longer be annoying.

Bottom line

Assessing the capabilities of the most productive desktop processor today, Intel Core i7-980X Extreme Edition, you begin to forget about its various features and nuances, because... Its level of performance, especially in well-optimized applications for multi-threaded execution, is impressive. And this is truly a confident step into the future, since the Intel Core i7-980X Extreme Edition is also one of the most complex processors today, which means Intel has perfectly mastered the 32 nm process technology, and soon we can expect the transfer of other processors to it, which will be noticeable more affordable and will have excellent overclocking potential. However, in order to increase the number of processing cores and the amount of third-level cache, while maintaining a thermal package of up to 130 W, some sacrifices had to be made - the latency of the cache memory increased and the speed of the built-in memory controller decreased, which may affect some unoptimized applications . This negative effect can only be smoothed out by enabling Intel Turbo Boost technology and using high-speed memory modules, and, of course, overclocking. After all, the Intel Core i7-980X Extreme Edition processor, traditionally for the Extreme Edition series, has a very high cost and is aimed at wealthy enthusiasts. Moreover, in this case, the effective “boxed” cooler Intel DBX-B Thermal Solution on heat pipes, which is an important addition to the Intel Core i7-980X Extreme Edition processor, will help with the experiments.

When you're buying a new laptop or building a computer, the processor is the most important decision. But there is a lot of jargon, especially regarding kernels. Which processor to choose: dual-core, quad-core, six-core or eight-core. Read the article to understand what this really means.

Dual core or quad core, as simple as possible

Let's keep it simple. Here's everything you need to know:

There is only one processor chip. This chip may have one, two, four, six or eight cores.
Currently, an 18-core processor is the best you can get on consumer PCs.
Each "core" is the part of the chip that does the processing. Essentially, each core is a central processing unit (CPU).

Speed

Now simple logic dictates that more cores will make your processor faster overall. But it is not always the case. It's a little more complicated.

More cores only give more speed if a program can divide its tasks among the cores. Not all programs are designed to split tasks between cores. More on this later.

The clock speed of each core is also a decisive factor in speed, as is the architecture. A newer dual-core processor with a higher clock speed will often outperform an older quad-core processor with a lower clock speed.

Power consumption

More cores also result in higher CPU power consumption. When the processor is turned on, it supplies power to all cores, not just the ones involved.

Chip makers are trying to reduce power consumption and make processors more energy efficient. But, the general rule is that a quad-core processor will drain more power from your laptop than a dual-core processor (and therefore drain the battery faster).

Heat release

Each core affects the heat generated by the processor. Again, as a general rule, more cores lead to higher temperatures.

Because of this extra heat, manufacturers must add better radiators or other cooling solutions.

Price

More cores are not always a higher price. As we said earlier, clock speed, architectural versions, and other considerations come into play.

But if all other factors are equal, then more cores will fetch a higher price.

All about the software

Here's a little secret that processor manufacturers don't want you to know. It's not about how many cores you use, but what software you run on them.

Programs must be specifically designed to take advantage of multiple processors. This kind of "multithreading software" is not as common as you might think.

It's important to note that even if it's a multi-threaded program, what it's used for is also important. For example, the Google Chrome web browser supports multiple processes, as well as the Adobe Premier Pro video editing software.

Adobe Premier Pro offers different engines to work on different aspects of your editing. Given the many layers involved in video editing, this makes sense as each core can work on a different task.

Likewise, Google Chrome offers different kernels to run on different tabs. But therein lies the problem. Once you open a web page in a tab, it is usually static after that. No further processing is necessary; the rest of the work is to store the page in RAM. This means that even though the kernel can be used to lay out the background, it is not necessary.

This Google Chrome example provides an illustration of how even multi-threaded software may not give you much of a real performance boost.

Two cores don't double the speed

So let's say you have the right software and all your other hardware is the same. Will a quad core processor be twice as fast as a dual core processor? No.

Increasing cores does not address the software scaling issue. Scaling to cores is the theoretical ability of any software to assign the right tasks to the right cores, so each core computes at its optimal speed. This is not what is really happening.

In reality, tasks are split sequentially (as most multithreaded programs do) or randomly. For example, let's say you need to complete three tasks to complete an activity, and you have five such activities. The software tells core 1 to solve problem 1, while core 2 solves the second, core 3 solves the third; Meanwhile, core 4 is idle.

If the third task is the hardest and longest, then it would make sense for the software to split the third task between cores 3 and 4. But that's not what it does. Instead, although cores 1 and 2 will complete the task faster, the action will have to wait for core 3 to complete and then compute the results of cores 1, 2, and 3 together.

This is all a roundabout way of saying that the software, as it is today, is not optimized to take full advantage of multiple cores. And doubling the cores does not equal doubling the speed.

Where will more cores really help?

Now that you know what cores do and their performance limitations, you should ask yourself, "Do I need more cores?" Well, it depends on what you plan to do with them.

If you often play computer games, then more cores on your PC will undoubtedly come in handy. The vast majority of new popular games from major studios support multi-threaded architecture. Video gaming is still largely dependent on what kind of graphics card you have, but a multi-core processor helps too.

For any professional who works with video or audio programs, more cores will be useful. Most popular audio and video editing tools use multi-threaded processing.

Photoshop and design

If you are a designer, then higher clock speeds and more CPU cache will increase speed better than more cores. Even the most popular design software, Adobe Photoshop, largely supports single-threaded or slightly threaded processes. Lots of cores won't be a significant incentive for this.

Faster web browsing

As we've already said, having more cores doesn't mean faster web browsing. While all modern browsers support multi-process architecture, kernels will only help if your background tabs are sites that require a lot of processing power.

Office tasks

All core Office applications are single-threaded, so a quad-core processor won't add speed.

Do you need more cores?

In general, a quad-core processor will perform faster than a dual-core processor for general computing. Each program you open will run on its own kernel, so if the tasks are separated the speeds will be better. If you use many programs at the same time, often switch between them and assign their own tasks to them, choose a processor with a large number of cores.

Just know this: Overall system performance is one area where there are too many factors. Don't expect a magical performance boost by replacing just one component, even the processor.

IntroductionIntel has long secured the title of manufacturer of the fastest processors for desktop computers. And if you can argue about which processes for computers in the middle and lower price categories should be considered the most optimal choice today, in the upper price category there is not even a hint of choice. Intel Core i7 is a family of processors to which AMD cannot offer worthy alternatives. At least at the moment, when the release of the six-core Phenom II, also known under the code name Thuban, is still several weeks away. At the same time, we can say that the existing quad-core Phenom II processors are more profitable: they are inferior in performance to the Core i7 by only a couple of tens of percent, and at the same time they are several times cheaper, but this does not change the situation. The most demanding computer enthusiasts are willing to pay a premium for high performance, which is why Core i7 processors are quite popular.

Even in the absence of direct competition, this consumer interest in high-performance and expensive processors is pushing Intel to continue improving its expensive products, which increase clock speeds, acquire microarchitectural improvements and even get an increased number of cores. The main character of this article is the recently announced representative of the Core i7 family, which became the first processor for desktop computers to receive six computing cores.

However, it should be understood that the appearance of a six-core model in the Core i7 line is far from the beginning of the six-core revolution. Today, Intel is ready to offer the only such processor, the Core i7-980X, belonging to the Extreme Edition series. This means that for now the six-core CPU is a kind of demonstration product that will be of interest from a practical point of view only to the wealthiest enthusiasts who are willing to pay about a thousand dollars for the processor alone. Moreover, this state of affairs will last at least until the fall, when, in addition to the Core i7-980X, another, not so expensive model of a similar processor may be released. However, this will not change the general situation - we will have to wait a very, very long time for the mass arrival on the market of products with the number of cores exceeding four. At least if we talk about processors made by Intel. Of course, AMD can make certain adjustments to the situation with “publicly available six-cores”, which is going to start selling processors with six cores in the mid-price category in the near future, but for now we do not have the opportunity to get acquainted with these products in practice, and therefore we will postpone conclusions until a more convenient occasion.

For us, getting to know the Core i7-980X is more interesting for another reason. This processor is based on a new Gulftown semiconductor chip, combining six computing cores and a 12-MB L3 cache. The implementation of all these nodes in a monolithic silicon crystal was made possible thanks to the use of a technological process with 32 nm production standards. The same process is partially used in the manufacture of Clarkdale family processors, but the Core i7-980X is the first product to produce which the most modern manufacturing process is used from start to finish. Thus, it is on the Core i7-980X that the evolution of the Nehalem microarchitecture should be fully visible. The recently announced Core i5 and Core i3 processors turned out to be a very bad example in this regard. The distribution of processor units across two semiconductor chips, one of which is produced using a 45 nm process technology, led to the emergence of additional bottlenecks that made a negative contribution to the consumer quality of the final products.

In other words, the Core i7-980X is what Intel engineers are currently capable of when combining an advanced process technology with the most modern microarchitecture. And it is from this rather theoretical point of view that Gulftown is interesting. In practice, such processors in the foreseeable future will be available only in the most expensive computers, and they will definitely not make it into the mass market segment this year. And no cheaper versions of Gulftown are planned for 2011, since Intel is immediately going to move on to introducing the next generation of microarchitecture, Sandy Bridge.

Core i7-980X Extreme Edition in detail

While we've described the Core i7-980X as a revolutionary product, we can't share any shocking details about its microarchitecture. Intel engineers simply assembled a six-core processor from their standard Nehalem design, combining the usual elements - computing cores, L3 cache, memory controller and QPI bus controller. It’s just that in one case there were more of these elements - the number of cores increased to six, and in the other - the size of the element increased - the capacity of the L3 cache increased to 12 MB. Yet these components fit on a single chip thanks to a new technological process with 32 nm production standards. As a result, although the Gulftown die contains 1,170 million transistors, which is approximately 1.6 times the number of transistors in the Bloomfield die, its area is 248 square meters. mm versus 263 sq. mm at Bloomfield.

If you look at the photograph of the Gulftown crystal and the placement of various blocks on it, the conclusion suggests itself that this is the result of a simple transfer of parts of the old core to production using a new technological process with minimal adjustments.

If we do not take into account the appearance of two additional cores, this is how it is. The processor cores themselves and the memory controller of the Core i7-980X are completely similar to the cores and memory controller of the Core i7-900 processors, which have been produced for more than a year. In fact, the only difference is in the production technology. The only innovation is the appearance of seven new AES-NI instructions aimed at speeding up the operation of cryptographic algorithms. However, these instructions are already familiar to us from Clarkdale processors.

So all we can do is report the main technical characteristics of the new product, comparing them with the characteristics of the Core i7-975 - the older processor of the Bloomfield generation, which is being replaced by a new six-core flagship.

The fact that the memory controller and QPI bus controller used in Gulftown are the same as the corresponding Bloomfield processor units means that they can be used on the same platforms. There is no PCI Express bus controller in Gulftown, and a set of logic, played by the well-known Intel X58 Express, is responsible for supporting the graphics subsystem.

Based on this, it is quite logical that the Core i7-980X has an LGA1366 design and works without problems in motherboards equipped with this connector. All that is required to support a new CPU with old boards is a BIOS update.

By the way, despite the one and a half times increase in the number of processor cores, the Core i7-980X fits into the same thermal package as its quad-core predecessors. Moreover, the transition to a more advanced technological process did not entail a decrease in the processor supply voltage - this can be clearly seen from the CPU-Z screenshot.

However, Intel has equipped its six-core processor with a new tower cooler that uses four 6mm heat pipes and a two-speed fan with a 100mm impeller.

But this was done not in connection with increased heat dissipation, but as another step towards enthusiasts, who now, after purchasing an Extreme Edition processor, can easily use a standard cooling system with good efficiency.

L3 cache and memory subsystem

In presenting Gulftown as the most powerful processor to date, Intel relies on two of its key features - an increased number of processing cores and an increased amount of cache memory. At the same time, it is quite obvious that at the moment there are not so many applications capable of simultaneously loading six processor cores, and most of them relate to the field of either three-dimensional modeling or the creation and processing of digital content. Therefore, from the point of view of common applications, another Gulftown property is much more important - L3 cache memory, the volume of which has been increased to 12 MB. It is thanks to this that in systems based on the new processor, performance gains can be noticeable in older tasks that are not optimized for multi-threaded environments. Moreover, the third level cache is common to all cores, which means that, depending on the nature of the load, it can be monopolized by one or several cores.

However, we remember well that even a simple increase in the amount of processor cache memory always entails some more negative consequences. That's what happened this time too. Since Intel engineers did not touch the logical organization of the L3 cache, leaving it with 16-channel associativity, the increase in volume and the need for arbitration between the increased number of cores led to a 33 percent increase in its latency.

The second factor that can negatively affect performance is that Gulftown processors have a reduced frequency of the Uncore part, which includes, in addition to the L3 cache, the memory controller. Intel engineers have already practiced slowing down Uncore in Lynnfield processors, in which, by reducing the frequency and voltage of the L3 cache and memory controller, they managed to significantly reduce power consumption. Similar motives motivated the developers this time too. The speed of the memory subsystem in platforms based on Gulftown was sacrificed for two additional computing cores. Otherwise, the six-core Core i7-980X simply would not fit into the 130-watt thermal package installed for LGA1366 processors.

As a result, when comparing the cache memory characteristics of older Gulftown, Bloomfield and Lynnfield processors, a rather contradictory picture emerges.

It is quite natural that Gulftown is inferior to its predecessor in the speed of working with cache and memory. The magnitude of this loss can be assessed, for example, by the results of the Everest Cache & Memory Benchmark. When testing, we used DDR3-1600 SDRAM with timings of 9-9-9-24.

Core i7-980X (Gulftown)

Core i7-975 (Bloomfield)

The difference in practical cache memory performance is immediately noticeable. Bloomfiled beats Gulftown by approximately 33% in L3 cache read speed and 25% in latency. The new product is also inferior in terms of memory processing speed. The practical memory bandwidth and latency of a six-core processor turns out to be approximately 15-20% worse than that of its four-core predecessor, which has a seemingly similar three-channel DDR3 SDRAM controller.

Thus, despite a larger number of processing cores and a more capacious cache, in real applications the Core i7-980X may be inferior in performance to the Core i7-975 - there are quite objective reasons for this. Actually, it now becomes clear why Intel gave the new product such a small processor number. After all, the new Gulftown turns out to be better than the old Bloomfield in not everything, and its weaknesses cannot be called insignificant.

Turbo Boost and Hyper-Threading technologies

Turbo Boost and Hyper-Threading technologies were introduced in the very first Bloomfield processors, and now we can say with complete confidence that they have stood the test of time and confirmed their effectiveness. And if Hyper-Threading allows you to increase the speed of the system under multi-threaded loads, then Turbo Boost technology plays the opposite role - it helps to increase performance when loading only part of the cores. It is not surprising that both of these technologies were transferred to the new six-core Gulftown processor.

With six processing cores in the Core i7-980X, Hyper-Threading technology adds six more virtual cores to this processor, resulting in as many as twelve cores visible in the operating system at once.

When looking at this funny screenshot, a completely reasonable question arises: are there any applications that are capable of using all these resources to the fullest? In addition, a single memory bus is shared between all cores, so it is possible that computing resources will spend too much time waiting for data, since the memory bus bandwidth may not be enough for simultaneously working cores. To dispel all these doubts, we conducted a simple experiment - we checked the level of system performance in a popular 3D shooter while a number of processes using computing power and the memory bus were running in the system in the background. More specifically, we tested the speed in Far Cry 2 by running in parallel several copies of the performance test built into the WinRAR archiver (which itself also supports multithreading). During these tests, the memory operated in DDR3-1600 mode, and for comparison with Gulftown, a similar test was performed on platforms with older processors from the Bloomfield and Linnfield families.

Overall, Gulftown handles multi-threaded workloads much better than its quad-core counterparts. The drop in performance as the background load increases is much slower for this processor, which means that the bandwidth provided by the three-channel memory subsystem is generally sufficient when working in multi-threaded environments.

As for Turbo Boost technology, its implementation in the Core i7-980X is somewhat disappointing. After Lynnfield processors for the LGA1156 platform were able to increase their frequency by 667 MHz above the nominal within the framework of this technology, we expected to see a similar frequency increase in Gulftown. However, Intel engineers thought differently, and in the new six-core processor Turbo Boost technology turned out to be just as conservative as in Bloomfield. As a result, the Core i7-980X with a nominal frequency of 3.33 GHz can increase by only 266 MHz - up to 3.6 GHz. Details about the frequencies of older processors in the Gulftown, Bloomfield and Linnfield families when turbo mode is enabled are shown in the table.

As a result, the maximum frequency of all older processors with the Nehalem microarchitecture is the same - it is 3.6 GHz. At the same time, according to official data, the Core i7-980X is capable of maintaining this frequency even when two computing cores are loaded. But in practice, we were able to observe the Core i7-980X operating at a frequency of 3.6 GHz exclusively with a single-threaded load, while loading the second processor core with work led to a decrease in frequency to 3.46 GHz.

However, it must be remembered that the possibility of overclocking a processor using Turbo Boost technology is determined not only by the activity of the cores, but also by the power consumption of the processor at each moment in time. So the inability of the Core i7-980X to operate at 3.6 GHz with a dual-threaded load is probably due to the fact that the power consumption of this processor in this mode goes beyond the limits established by the specification.

How we tested

There is no doubt that the Core i7-980X is one of the fastest processors. Therefore, in performance tests, for comparison with it, we took a pair of the fastest quad-core Intel Core i7 series processors and the senior processor of the Phenom II X4 family. As a result, the test systems included the following set of components:

Processors:

AMD Phenom II X4 965 (Deneb, 3.4 GHz, 4 x 512 KB L2, 6 MB L3);
Intel Core i7-980X (Gulftown, 3.33 GHz, 6 x 256 KB L2, 12 MB L3);
Intel Core i7-975 (Bloomfield, 3.33 GHz, 4 x 256 KB L2, 8 MB L3);
Intel Core i7-870 (Lynnfield, 2.93 GHz, 4 x 256 KB L2, 8 MB L3).

Motherboards:

ASUS P7P55D Premium (LGA1156, Intel P55 Express);
Gigabyte MA790FXT-UD5P (Socket AM3, AMD 790FX + SB750, DDR3 SDRAM);
Gigabyte X58A-UD5 (LGA1366, Intel X58 Express).

Memory:

2 x 2 GB, DDR3-1600 SDRAM, 9-9-9-24 (Kingston KHX1600C8D3K2/4GX);
3 x 2 GB, DDR3-1600 SDRAM, 9-9-9-24 (Crucial BL3KIT25664TG1608);

Graphics card: ATI Radeon HD 5870.
Hard drive: Western Digital VelociRaptor WD3000HLFS.
Power supply: Tagan TG880-U33II (880 W).
Operating system: Microsoft Windows 7 Ultimate x64.
Drivers:

Intel Chipset Driver 9.1.1.1025;
ATI Catalyst 10.3 Display Driver.

Performance

Overall Performance

The SYSmark 2007 test, which shows system performance when executing typical scenarios in real applications, immediately highlights the shortcomings of Gulftown that we mentioned above. In the case where the applications used are not well optimized for multi-core processor architectures, the Core i7-980X can easily lag behind its predecessor, the quad-core Core i7-975. This is exactly the picture that is observed in the E-Learning and Productivity scenarios - in them, the higher result is shown not by the processor with more cores, but by the one with a faster L3 cache and memory controller. Scenarios that simulate the creation and processing of digital content put Gulftown in first place, which is not surprising, since applications used for this type of activity are usually good at distributing the load across multiple computing cores. But as a result, the overall SYSmark 2007 result of the new Core i7-980X is practically no different from the result of the Core i7-975.

Gaming Performance

Many modern games can already effectively use the resources of dual-core processors. Some of them are capable of loading quad-core CPUs. The six-core Gulftown, which also has support for Hyper-Threading technology, is clearly beyond the capabilities of modern games to fully load with work. Therefore, the differences in the results of the Core i7-980X and Core i7-975 are not so striking. Another factor is much more important for gaming applications - the L3 cache increased to 12 MB. It is thanks to him that Intel's new CPU can become a useful acquisition for gamers.

3DMark Vantage

The popular 3DMark Vantage performance test can effectively load any number of processor cores. That is why the result of the Core i7-980X looks very impressive. So new world records in this test will now be set primarily by systems based on this processor.

Application Performance

Adobe Photoshop is an application optimized for multi-core architectures. But not all operations and filters performed in it use the maximum number of cores. Therefore, the advantage of the six-core processor turned out to be not so significant, and partly it is explained not so much by the number of Gulftown cores, but by its increased L3 cache.

Video transcoding is a highly parallelizable task. Therefore, here the new Core i7-980X with six cores naturally demonstrates more than 40 percent superiority over the Core i7-975, which has only four processing cores.

A similar picture is observed with non-linear editing of high-definition video in Premiere Pro.

WinRAR can also use several processor cores, but when their number increases beyond three, the performance gain becomes almost unnoticeable. Therefore, the Core i7-980X and Core i7-975 demonstrate similar speeds. And by the way, the 12 MB L3 cache of a six-core processor does not provide any visible effect: its large size, unfortunately, is neutralized by high latency.

Arithmetic calculations in Excel 2007 can be efficiently parallelized. As a result, our test task is calculated significantly faster on a new processor with a larger number of cores.

The software audio studio Sonar 8 Producer also runs slightly faster in the final mixing of tracks on a system with a six-core processor. The advantage of the Core i7-980X over the Core i7-975 is about 5%.

The final rendering is one of those types of workloads that always respond positively to an increase in the number of processing cores in the system. So, at least a 20% superiority of the Core i7-980X over its competitors is a completely logical result.

Single-threaded performance

In order to see how processors cope with a single-threaded load, we included two additional tests in the study: the MaxxPi computing test and the Fritz chess program, in which the number of processor cores used was manually set to one. This test is of interest because the older processors of the Core i7 family have Turbo Boost technology, thanks to which their clock frequency when loading a single processor core is leveled at 3.6 GHz.

As you can see, in these tests the Core i7-980X and Core i7-975 show relatively close results with a slight advantage for the older processor, which has a more efficient cache memory in terms of operating speed. Moreover, the Core i7-870 is also catching up to them, the slight lag of which in this case is mainly due to the lower bandwidth of the memory subsystem.

Energy consumption

Formally, the increase in the number of cores in the new Core i7-980X processor did not entail a change in the calculated heat dissipation. Its TDP compatibility with the LGA1366 platform is ensured by both the more modern technological process used in the production of Gulftown semiconductor crystals and the reduction in frequency and supply voltage of Uncore. As a result, the estimated typical heat dissipation of the Core i7-980X, like that of the Core i7-975, is 130 Watts.

However, to get a more detailed picture, we also conducted practical testing of energy consumption. The following graphs show the total system consumption (without monitor), measured “after” the power supply and representing the sum of the power consumption of all components involved in the system. The efficiency of the power supply itself is not taken into account in this case. During measurements, the load on the processors was created by the 64-bit version of the LinX 0.6.3 utility. In addition, to correctly estimate idle power consumption, we activated all available energy-saving technologies: C1E, AMD Cool"n"Quiet and Enhanced Intel SpeedStep.

Without load, the consumption of the LGA1366 platform exceeds that of other platforms, regardless of which processor it uses. This is explained by the fact that the Intel X58 Express logic set has a very “gluttonous” temperament. The consumption of the processors themselves during idle time accounts for no more than a few watts.

Under load the situation looks much more interesting. The new six-core processor turns out to be even more economical than its quad-core brother called Core i7-975. However, the 32 nm process technology does not work any special miracles, and the Core i7-980X remains a very energy-intensive device: its consumption significantly exceeds the consumption of older CPUs for the LGA1156 and Socket AM3 platforms. On the other hand, given that Gulftown has one and a half times more computing potential, energy efficiency (the ratio of performance to energy consumption) also reaches a new level.

Overclocking

The transfer of processor production to a new technological process usually entails an increase in frequency potential. The Core i7-980X is the first processor produced exclusively using the 32nm process technology. That is why the results of its overclocking are of particular interest.

The only Gulftown currently available is the Extreme Edition. This means that Intel does not lock its multiplier, giving the user an easy path to overclocking. It was this opportunity that we used when conducting our experiments. To remove heat from the processor during tests, a Thermalright Ultra-120 eXtreme air cooler was used.

First of all, we tried to establish the overclocking limit of the Core i7-980X, which can be achieved without increasing its supply voltage above the standard 1.2 V for our CPU instance. As we showed in our recent material, it is this kind of overclocking that is the most energy efficient and does not lead to a catastrophic increase in power consumption and heat generation.

Practical tests have shown that stability of operation without increasing the processor voltage is not lost at a maximum frequency of only 3.6 GHz.

Unfortunately, this frequency is very close to the standard one and is unlikely to satisfy enthusiasts. Therefore, the second series of experiments was carried out with an increase in the voltage on the CPU to 1.35 V. Moreover, as we know from the example of Clarkdale, processors produced using 32 nm technology should respond to voltage increases very well.

By increasing the voltage, we were able to achieve stable processor performance at a much higher frequency - 4.13 GHz.

But frankly speaking, this is not the result that we hoped to see when overclocking the new Core i7-980X. It turns out that, despite the fact that this processor was produced using the most modern technological process, it overclocks no better than a year-old CPU built on 45 nm semiconductor crystals. In other words, when overclocked without the use of special cooling means, the frequency potential of Gulftown approximately corresponds to the potential of Bloomfield processors, the overclocking limit of which is in the region of 4.0-4.2 GHz.

By the way, I would like to note two features that we noticed when overclocking the Core i7-980X. Firstly, Gulftown maintains a relatively low temperature even when its frequency increases with increasing supply voltage. 60 degrees at maximum load is very low compared to the temperatures at which Core i7 processors from the Bloomfield families overclocked with increasing supply voltage usually operate. Secondly, successful overclocking of Gulftown requires a fairly careful selection of voltage, and increasing it too much leads to deterioration in overclocking results. For example, our processor instance worked at a frequency of 4.13 GHz when its voltage was raised above the nominal by 0.15 V, but when the voltage was increased by 0.2 V, it could not pass stability tests even at a frequency of 4.0 GHz.

conclusions

Despite the fact that Gulftown is not only the first six-core processor for desktop computers, but also the first CPU, which is manufactured exclusively using a 32-nm process technology, we would not classify it as a next-generation product. In fact, Intel offered us everything that we have already seen in Bloomfield processors, only this time, to introduce the next model in the Core i7 family, they chose not to increase the clock frequency, but to add processing cores. Which, given the block structure of processors with the Nehalem microarchitecture, is not such an innovation.

As a result, the new Core i7-980X theoretically has one and a half times higher performance, which formally allows it to be considered the fastest processor for desktop computers. In practice, everything depends on application optimization. As tests have shown, there are not so many tasks that receive a commensurate performance gain when running on a six-core processor, and they relate exclusively to the creation and processing of digital content. It turns out that the Core i7-980X is an excellent option for use as a workstation base, rather than in a home computer.

It is not at all surprising that when releasing the six-core Gulftown to the market, Intel limited itself to offering a single model costing $999. Under normal conditions, using a processor with six computing cores does not make much sense, and Gulftown, moreover, under certain circumstances, may be slower than its four-core predecessors due to the increased latency of the L3 cache and a sluggish memory controller. So the Core i7-980X is clearly designed for those high-income enthusiasts who are drawn to everything new primarily out of curiosity, and not based on sound calculation. Pragmatists, even after the appearance of the Core i7-980X, will probably not lose interest in existing quad-core processors, the performance of which is quite sufficient for everyday work and for modern 3D games. Moreover, the 32-nm process technology does not provide any significant dividends: as tests have shown, the Core i7-980X has become only slightly more economical than its quad-core LGA1366 predecessors, and its overclocking potential does not at all exceed the capabilities of 45-nm processors.

In general, we will have to wait for truly innovative Intel processors, which may be of interest to the general public, at least until the beginning of 2011, when the microprocessor giant should bring to the market dual-core and quad-core products with an updated Sandy Bridge microarchitecture, for the manufacture of which it will use 32- nm technical process. Regarding the new product discussed in this article, I just want to say: “Nothing special.”

Main competitor

Perhaps the world would not have seen an AMD processor with six cores on store shelves, but this became inevitable for the manufacturer when its main competitor, Intel, introduced the new Core i3-2125 to the market in the mid-range budget class. The power of budget-class processors (on the Russian market there was only one crystal with the old Phenom technology, which all buyers bypassed) AMD at that time was not enough to fight the competitor, and modifications with eight cores competed for primacy with the Core i5. There was an urgent need to occupy an open price niche.

The manufacturer presented to the world community several products for the budget price niche. All of them differed slightly in cost and performance. Such a decision was supposed to drive the competitor out of the market. The new FX-6100, its characteristics and price immediately attracted the attention of the buyer. The new product was definitely interesting because there were six cores on board the crystal, which worked independently of each other. It was from this moment that the battle of the titans began:

Specifications

The consumer clearly liked the manufacturer's approach to creating the processor, because to create it, the company's technologists developed a completely new Zambezi core, refusing to use old technologies. As a result, the new AMD FX-6100 BOX received the following technical characteristics:

Interaction with motherboards on
The production uses a new 32-nanometer process technology, which makes it possible to place 1.2 billion transistors on one chip.
The number of computing processes corresponds to the number of cores - 6 pieces.
The nominal core frequency is 3300 MHz (3900 in Max Turbo mode).
Full amounts of cache memory of all three levels are used.
The processor supports two DDR3 memory channels operating at frequencies of 1333/1600/1866 MHz.
All instruction sets are supported for running 32-bit and 64-bit applications, including MMX.
Heat dissipation at peak loads does not exceed 95 watts. We are talking about a base frequency of 3300 MHz. With increased performance, heat dissipation can reach 150 watts.

Appearance and packaging

The AMD FX-6100 processor may be the smallest device in the system, but it’s a given that the packaging for the “heart” of a computer will always be beautiful and huge in size. It is worth noting that the manufacturer has completely changed the design of the box for all Black Edition series processors. Instead of the utopian black color, it is made in a red and white style, to match all AMD products. The main attribute - packaging with a window for contemplating the processor - has not changed. The contents of the box for all brands are identical and have long been considered the norm:

the processor itself is in a miniature plastic package that can protect the device from physical impacts during transportation;
AMD branded sticker on the owner’s system unit;
cooling system assembly (fan and radiator);
colorful instructions with pictures for installing the processor and mounting the cooler;
a lot of “waste paper” (leaflets, certificates, guarantees and recommendations).

Questions about the cooling system

Coolers that came in the same package with devices always received low ratings from those who like to test processor power through overclocking. Therefore, the AMD FX-6100 Six-Core Processor product was no exception. The aluminum radiator with a copper core at the base and a 70mm cooler looks a bit weak. However, according to the manufacturer, such a system can cope with processor cooling at peak loads of up to 100 W.

As users note in their reviews, for standard frequencies (3300-3900 MHz) this cooling system is quite sufficient, but overclocking enthusiasts should think about more advanced coolers from well-known world brands. Accordingly, in such cases it is recommended to purchase the processor not in the BOX version, but to give preference to OEM delivery.

Sports interest

Naturally, all potential buyers are interested in comparing two processors in the same niche from different manufacturers. For the purity of the experiment, the AMD FX(TM)6100 Six-Core must be compared in tests with the Intel Core i3-2125. In fact, these are two identical processors, judging by their technical characteristics and price, although the latter has only two cores.

As test results show, in resource-intensive applications that require processor power (archivers, password crackers, video and audio encoders, mathematical calculations), the new product from AMD is the leader. Undoubtedly, 6 cores are more productive than two. However, the situation changes dramatically in benchmarks, when only one core is involved in the test - Intel Core i3-2125 wins by a huge gap over its competitor (Cinebench R11.529, 3DMark).

But with games the issue is controversial. Applications that are designed to run on one or two cores undoubtedly show better results with an Intel processor. And all the rest, which require overall system performance, show decent results with the AMD FX-6100 crystal. It is worth noting that recently many game manufacturers have been writing codes without reference to processor threads, and accordingly, AMD’s new product has a greater chance of winning in performance over its competitor.

Higher, faster, stronger

In the media you can find many arguments from “experts” who assure others that the AMD FX-6100 Six processor is a lightweight version of its older brother FX-6300. Logically, these two processors have many of the same parameters: number of cores, cache, memory bus, instructions, heat dissipation and technical process. But the chipsets used for their production are different, and the technologies have slight differences. Testing will put everything in its place.

The GeekBench CPU benchmark shows the 6300's CPU performance to be 7,677 (the 6100's 6,945).
The AMD FX-6100 processor does not support FMA3, which is used to speed up tasks.
The 6300 chip runs 10% faster with all applications running the updated version of the Turbo Core AMD instructions (video editing and 3D modeling).

The right approach

For many potential buyers who want to overclock the AMD FX-6100 processor, reviews from owners lead them to believe that it is worth purchasing a decent cooling system to protect the crystal from overheating. The choice falls on expensive high-end devices, the cost of which is not comparable to the price of the processor itself. Naturally, the buyer immediately gives up his desires. There is no need to rush here, the main thing is to know one truth: any cooler on the market positioned for a certain heat dissipation is definitely more efficient than the BOX version.

The choice of a decent cooling system within 3,000 rubles is quite large, and for most buyers it is not carried out at the level of characteristics, but is tied to the brand. Devices from Zalman, Scythe, Deepcool, and Cooler Master have proven themselves to be excellent. Any cooler you like from the proposed options is guaranteed to cope with the task. For an AMD FX-6100 processor with a heat dissipation of 95 W, it is worth choosing a cooling system with a coefficient of one and a half. That is, the cooler must cope with the dissipated power of the processor of 142.5 W.

Overclocking potential

Many beginners, after installing the proprietary AMD Catalyst software, discover that on one of the application tabs there is information on the processor, which indicates the nominal processor frequency and overclocking potential. Very often the user contemplates a figure equal to 4.3 GHz; naturally, he overclocks the crystal to the maximum.

You should not do this at the initial stages; IT experts recommend overclocking the AMD FX-6100 3.3 GHz processor to the maximum permissible level specified in the device specification - 3.9 GHz in Max Turbo mode. It is necessary to work in this mode, observe the temperature characteristics of the cooling system, including programmatically, using special utilities. If there are problems, reduce the frequency by 100 units. If heating is under control and the processor is operating stably, you can begin to increase the frequency in 100 MHz increments.

Overclocking instructions

How to overclock AMD FX-6100? Judging by reviews in the media, many users are interested in complete step-by-step instructions with recommendations for overclocking the processor. No problem:

Go into the computer's BIOS.
Go to the "Advanced" tab.
Select the item “JumperFree Configuration”.
Find the “CPU Ratio” menu.
To the right of the found menu there is an “Auto” option. You need to click on it and select the correct multiplier in the list that appears (19.5x corresponds to a frequency of 3900 MHz).
Save and restart your computer.

But the overclocking saga doesn’t end there, because many users pay minimal attention to the cooling system, so the manufacturer took full responsibility. The AMD FX-6100 Six processor is equipped with an overheating protection system (58 degrees Celsius). The protection works great - it simply reduces the core frequency by half by setting the required parameter in the BIOS. There are two ways to solve the problem: either install a powerful cooling system, or deceive the lock.

Walking on the blade

As a result, through trial and error, the user will arrive at the correct result. It is worth noting that for each computer these indicators are different (3600 MHz and 1.24 V, 3900 MHz and 1.36 V). After overclocking, many owners in their reviews recommend not focusing on the maximum, because at peak loads the processors do not have time to cool in a timely manner, and accordingly, blocking will occur.

CPU stress test

Many users, judging by their reviews, are not particularly clear about the question of testing an overclocked processor, because there is so much similar software on the Internet that their eyes wander when choosing. Professionals recommend testing the stability of the AMD FX-6100 processor, the characteristics of which have been changed by the user, using the OCCT program. The fact is that only this application can perform testing according to specified parameters and provides a lot of useful information.

To set parameters in the OCCT program, the user must set the test time (10-20 minutes is considered normal). Be sure to specify the test version (32 or 64 bits). Select the maximum testing mode - a large set, and it is better to set the number of tests to “Auto”.

In addition to the test result, at the end of the program, the user is given the opportunity to monitor the temperature and voltage of the processor while the cores are loaded. The system will naturally freeze. This is normal, because OCCT takes over all the resources.

Until recently, Intel processors were developed according to the time-tested Tick-Tock system, that is, according to the principle of a pendulum: at each “tick” a new, significantly redesigned architecture is born, and at each “tick” the existing architecture is transferred to a new one , a more advanced technical process. Intel plans to continue to adhere to this approach, but the pendulum does not swing quite evenly, and therefore some “intermediate” solutions appear periodically. One of these products is the Intel Core i7 980X processor we are considering, which represents the Nehalem architecture, which is being transferred as part of the next “so” to a 32-nm process technology. But in this case, the swing of the pendulum is slightly different from usual - the transition to a new technological process most often makes it possible to increase the operating frequency of the processor, but Intel chose a different path and increased the number of cores to six. So, the Intel Core i7 980X is the first six-core processor for desktop computers in our testing laboratory. Let's take a closer look at its architecture.

⇡ Architecture

The Intel Core i7 980X processor belongs to the Gulftown family and is its first and so far the only representative of processors of this family. There are no fundamental differences from the architecture of the Bloomfield family, on which all other processors for the LGA1366 platform are based, in the Intel Gulftown architecture. We can assume that the Core i7 980X is the same Bloomfield, operating at a frequency of 3.33 GHz, with a third-level cache increased by 4 MB and manufactured using a 32 nm process technology. However, there are some significant differences.

First, thanks to Intel HyperThreading technology, this six-core processor can handle up to twelve data threads, which is four more than all other Core i7 processors.

Secondly, the Core i7 980X received a new AES-NI (Advanced Encryption Standard New Instructions) instruction set, consisting of twelve different instructions designed to speed up all applications that actively use the AES algorithm. The AES-NI instruction set is already used in Clarkdale processors, but this is the first solution for the LGA1366 platform with this instruction set. Adding them will significantly increase processor performance in tasks such as encryption, VoIP, Internet firewalls and other applications that rely heavily on encryption. For other applications, the presence of AES-NI will have virtually no effect.

Thirdly, the L3 cache increased to 12 MB can have a positive effect on performance in games and other applications that use large amounts of cache memory. At the same time, other applications may lose some performance, since the increase in cache memory also led to an increase in latencies - the Uncore bus frequency in the new processor was reduced from 3.2 GHz to 2.6 GHz.

Finally, fourthly, the transfer of the processor to a 32-nm process technology using transistors with a metal gate had a positive effect on its physical dimensions: the Gulftown die has an area of 248 mm², while the quad-core Bloomfield die has an area of 263 mm², and the Lynnfield die has an area of 263 mm², and altogether 296 mm². Reducing the technological process standards should have a positive effect on the heat dissipation of the processor and its overclocking potential. The Core i7 980X has 1.17 billion transistors, making it the first home processor to surpass one billion transistors.

Otherwise, the Core i7 980X is similar to the Core i7 975: the same QPI bus frequency of 6.4 GT/s, that is, 25.6 GB/s, a similar built-in memory controller that allows you to work with DDR3 1333 memory in three-channel mode. Both processors operate at the same frequency and have an unlocked multiplier, the value of which can vary from 12 to 60 (in nominal mode - 25, in Turbo Boost mode - 27).

⇡ Cooling system

Many buyers of top-end Intel processors were very surprised when they took out of the box with a processor for several tens of thousands of rubles a simple aluminum radiator with radially diverging fins and a small noisy fan. Standard Intel cooling systems practically did not change from processor to processor, except that the height of the fins increased. With the release of the Core i7 980X, for the first time in many years, Intel changed its approach to standard processor cooling and equipped the new product with a much more serious cooler, called the Intel DBX-B Thermal Solution.

The new cooler is a tower heatsink with four heat pipes running through a copper base. On one side there is a fan with a diameter of 100 mm with a transparent impeller and blue backlight. Let's look at the cooler in a little more detail.

The radiator itself consists of aluminum fins of medium thickness, and the distance between them is very small - it will be difficult for fans with low speeds to blow through such a structure. Four heat pipes with a diameter of 6 mm are neatly sealed in the hollows of the base - there is, of course, no technology for direct contact of heat pipes with the processor itself, but this is not necessary. The top of the radiator is covered with a cover with protrusions for heat pipes, on which the Intel logo is placed.

The fan impeller is the strangest part of the cooler: its blades have a slightly curved shape, and it is not enclosed in a frame. As a result, only a small part of the air flow is sent directly to the radiator, but the airflow around the motherboard around the processor is at a high level.

The processing of the base of the cooler is at an average level: it is not mirror-like, but without any distinct irregularities. At the same time, the base is slightly convex, which ensures good contact with the processor cover in the middle, where the crystal itself is located. This solution is ineffective if the processor cover is perfectly flat, but in our case it turned out to be slightly concave, and here the convexity of the cooler base came in very handy.

The Intel DBX-B thermal Solution attaches to the motherboard with four finger-tight screws. A soft plastic plate is installed on the back of the motherboard, into which screws are screwed. Despite the inconvenient location of the screws (you have to reach up to the heads of two of them) and the flimsy design of the plate, this mount is a huge step forward compared to all previous versions of mounts.

There is a two-position switch at the top of the radiator. The letter "S" stands for Silence, while the letter "P" stands for Performance. In the first mode, the fan rotates at a speed of approximately 800-900 rpm, and in the second - about 1800 rpm. And if in the Silence mode the fan can be called moderately noisy, then in the Performance mode it is very loud: its noise drowns out both the power supply fan, the video card fan, and the sound from the hard drive heads. The blue illumination of the impeller cannot be turned off, but it is not too bright and does not hurt the eyes.

In general, despite the huge number of shortcomings, the Intel DBX-B cooler is far superior to all previous cooling systems that were equipped with Intel processors. Unfortunately, it is intended only for Gulftown processors - other processors will be equipped with old coolers. Let's see what the new cooling system is capable of in action - let's try to overclock the processor.

The maximum frequency we were able to load the system at using air cooling was almost 4.5 GHz. At this frequency it was even possible to pass some tests, but stability was not observed. Therefore, the frequency had to be reduced to 4.2 GHz - at this frequency, all tests passed properly, and the processor with the Intel DBX-B Thermal Solution cooler installed on it did not warm up above 65 degrees Celsius. However, when trying to check the stability of the processor in the OCCT utility, the Core i7 980X processor with a standard cooler still warmed up to 85 degrees, and the system eventually produced a blue screen. Despite this, we will consider the operation of the processor at this frequency to be conditionally stable, since the loads created by the OCCT LinPack utility are not encountered in real applications.

⇡ Temperature and power consumption

Let's move on to the processor performance tests and compare its results with the results of other latest generation Intel processors, but first let's evaluate the system's power consumption.

Test bench configuration:
Processors	Intel Core i7 980X 3.33 GHz Intel Core i7 920 2.66 GHz Intel Core i7 870 2.93 GHz
Cooling systems	Intel DBX-B Thermal Solution for Core i7 980X Titan Fenrir for Core i7 920 and Core i7 870
motherboards	Asus Rampage II Extreme MSI P55-GD65, Socket LGA1156 ASUS P6T Deluxe Palm OS Edition, Socket LGA 1366
RAM	3x 1GB Apacer DDR-3 2000 MHz (9-9-9-24-2T) @ 1333 MHz (7-7-7-24-1T) 2x 2 GB Corsair XMS 2 @ 1066 MHz (5-5-5-15-2T)
Hard disks	Seagate Barracuda 7200.10 750 Gb Samsung SpinPoint SP750
Video card	NVIDIA GeForce GTX 295, WHQL 186.18 drivers
power unit	Hiper M730

At standard frequencies, our test bench, together with the Core i7 980X processor, consumed only 185 W, which is not bad for a computer with the most powerful desktop processor and a dual-chip video card. Under load using the OCCT utility, the system's power consumption increased significantly and amounted to 297 W - this is only due to the processor, because the OCCT LinPack test does not load the video card.

Overclocking with an increase in processor voltage to 1.35 V does not greatly affect the system's power consumption when idle - it is 192 W, but under load, power consumption increases to 344 W - almost 50 W more than without overclocking.