The long-awaited AMD Bulldozer processor. Processors

AMD FX-8350 | Meet the AMD FX line of processors based on Piledriver

Us as columnists computer hardware, are not very interested in the problems faced by its manufacturers. Many would agree that last year was a terrible year for AMD's processor design division, starting with the power-hungry Bulldozer CPUs, which slowly fell in price over the course of twelve months in response to the introduction of processors Intel Core third generation. New samples recently arrived at our laboratory in Southern California. Talk about shortcomings in management, layoffs and the difficult past of AMD should not particularly worry end consumers. So let's get down to business.

Sometimes you can predict the ending of an article. If AMD had sent us the FX-8170 running 200MHz faster than its predecessor, we'd expect the processor to show the same weaknesses in light-threaded applications, likely outperforming the Core i5-2500K in more intensive tasks, but compared to 77W chips its energy consumption will be simply terrible.

But instead we got a processor AMD FX-8350, which, by analogy with APU Trinity, introduced less than a month ago, is based on the Piledriver architecture. Experience tells us that in terms of performance per core and per clock, Piledriver can outperform the Bulldozer design by 15%. It is also worth considering that AMD FX-8350 will run at least 400 MHz faster than FX-8150. By the way, the architecture provided a noticeably smaller increase in speed for the Intel Core processor line. There's a good chance that today's comparison will be much more interesting than a complete rip-off FX-8150 in last year's review.

Meet the FX family on Piledriver architecture

True to its tradition since Bulldozer, AMD sent out to the press the fastest models of the new lineup, which will include eight-, six-, and quad-core models. Although they are all built on the Piledriver architecture, the chips themselves are called Vishera, and will continue to be sold under the FX brand.

AMD Vishera crystal

The Vishera processor occupies 315 square millimeters and consists of 1.2 billion transistors. Exactly the same indicators characterize the previous generation Zambezi based on the AMD Bulldozer architecture.

Two of the four processors use eight processing cores, or four Piledrivers, as AMD calls them. Base frequency of the flagship model AMD FX-8350 is 4 GHz. In low-threaded applications Turbo technology Core can increase this figure to 4.2 GHz, although most of the increase in chip speed will be associated with the base frequency. How much does Turbo Core speed up? AMD FX-8350? Not by much. In the single-threaded iTunes benchmark, the result improved by only three seconds.

The FX-8320 lowers the base frequency to 3.5 GHz, but within the allocated thermal package, the Turbo Core increases it to 4 GHz (the 500 MHz boost makes a greater difference on the FX-8320). Both eight-core models are equipped with 8 MB of L2 cache (divided into 2 MB for each module) and 8 MB of L3 cache (common for all four chip modules). Recommended price for AMD FX-8350 is $195, and the FX-8320 is $169.

The FX-6300 already has three active modules (six cores), and the price has dropped to $132. The 3.5 GHz frequency gives the architecture an advantage in multi-threaded tasks, and Turbo Core tries to compensate for the disadvantages in single-threaded applications by increasing the frequency to 4.1 GHz. Like four-module chips, the FX-6300 uses 2 MB of L2 cache per module (6 MB total) and a shared 8 MB L3 cache. Fewer active resources, as well as a lower northbridge frequency (2 GHz), allow the FX-6300 to be within the 95 W thermal package, which is noticeably different from the 125 W of the older FX-83x0 processors.

The TDP of the dual-module FX-4300 processor also does not exceed 95 W. The base frequency of 3.8 GHz rises to 4 GHz in low-threaded applications, and the northbridge operates at 2 GHz like the FX-6300. However, the L3 cache capacity is reduced to 4 MB, and the price is only $10 lower than the three-module chip, which will encourage many buyers to spend the extra $10.

The AMD architecture does not experience any particular shortage of memory bandwidth, however, the dual-channel DDR3 controller officially supports a data transfer rate of 1866 MT/s. To reduce the overall cost of the platform, we will use DDR3-1600 modules with low latency, especially since the test results (except Sandra 2013 Beta) show that spending more money on more fast memory You won't get any speed increase.

The entire FX line has an unlocked multiplier, which makes overclocking much easier. Does Vishera have enough overclocking potential? What about 5.125 GHz with closed loop liquid cooling support?

AMD FX-8350 | Overclocking and platform compatibility

Overclocking

Although AMD processors have not been recognized as the fastest processors for several years, the company is trying to keep the attention of enthusiasts by providing features that are important to them. Software that allows you to change configuration settings in Windows in real time, unlocked multipliers and platforms with more PCI Express slots are just a few of the key features, which AMD offers to people who know how to use them to increase the performance of their systems.

Those same overclockers who were disappointed in the scalability of FX Zambezi processors with a conventional cooling system will be greatly surprised this time, even though we are considering a very similar architecture on the same 32nm core.

With CPU and northbridge voltages of 1.375 V and 1.175 V, respectively, we were able to achieve stable operation AMD FX-8350 at 4.8 GHz at full load. In the screenshot above, a single-threaded test is running to “spin up” the chip, but the highlighted maximum temperature corresponds to the peak load of the entire test package.

AMD FX-8350 could work even faster, but the key factor here is setting the voltage so that the temperature does not exceed 70 degrees Celsius. At this point, the temperature sensor forces the cores to lower their frequency (the image above proves this), preventing the chip from overheating and negatively impacting performance. Without multi-threaded loads, we even managed to run tests at 5.125 GHz (for this we need to set the voltage to 1.4375 V for the CPU and 1.2 V for the northbridge).

Obviously, in most user systems, cooling will be the bottleneck. The AMD reference heatsink and fan are, to put it mildly, not enough, and a powerful third-party solution will increase the cost of the platform with the FX processor. But for testing purposes only, we used the closed-loop liquid cooling system that AMD offered with its FX processors last year. It will cost you about $70. In that case, you can consider $300 as an alternative. Fortunately, we have test results.

Overclocking to 4.8 GHz is enough to AMD FX-8350 overtook in a multi-threaded environment like 3ds Max 2012, but it did not help the AMD Piledriver architecture outperform in a single-threaded environment iTunes app. Of course, if you are willing to spend an extra $30 on and even more on a third-party cooler, then its base frequency can be relatively easily increased from 3.5 GHz to 4.5 GHz and matched with AMD FX overclocking.

Compatibility

All four FX Piledriver chips are compatible with the existing Socket AM3+ processor interface. To the old motherboards recognized new FX series processors, you need to update the BIOS. However, boards that previously experienced problems with FX processors are unlikely to get rid of this shortcoming.

As an experiment, Asus added support for the FX line in 2011. However, the company has not yet released the updates needed to fix the issue. blue screen death in some situations. Therefore, we do not believe that older AM3 platforms will provide adequate support for FX processors, and the problems will not go away with the arrival of Piledriver. AMD assured that the problem is not global, and suppliers can solve it with updates. But some board makers don't seem to care much about updates to older products.

AMD FX-8350 | Piledriver architecture: what has changed compared to Bulldozer

Current AMD architecture x86 was discussed in great detail in the review FX-8150 (AMD FX-8150 Review: From Bulldozer to Zambezi and FX). All these principles can be transferred to the Piledriver architecture. However, AMD engineers gained experience when they put the Bulldozer concept into practice. We know that the process technology has evolved over the year, even though the company still uses a 32nm core in the production of Vishera processors. Therefore, you should not be surprised if the new architecture turns out to be simply a modified old one, where the changes introduced only turned out to be a long-planned “work on mistakes.”

Input improvements

After the premiere of AMD Bulldozer, it was argued that branch prediction is one of the weak points of the architecture. The Piledriver module concept includes some shared resources across two execution pipelines, and the architects tried to minimize bottlenecks in the preprocessor by implementing one branch prediction queue per thread. The company claims that Piledriver has improved prediction accuracy.

Piledriver has added support for several ISA extensions, which we first looked at in our review APU Trinity. The unified multiplication block was introduced a year ago in Bulldozer. That version was called FMA4 and allowed instructions to have up to four operands. But in the upcoming Haswell architecture, Intel only plans to use the simpler FMA3 instruction set with three operands, so AMD retains this advantage in Piledriver. Another extension is called F16C. It includes support for converting up to four half-precision floating point values ​​at a time. The Intel architecture already has such an extension, so Piledriver is simply catching up with its competitor. This is not to say that Bulldozer felt an urgent need for FMA3/F16C, since compiler-level support was only added in Visual Studio 2012.

Integer block

Each of the two integer cores uses a separate load/store unit, capable of performing two 128-bit loads per clock or one 128-bit store per clock. AMD found that in some cases Bulldozer was unable to detect stored data in the register file that was already there. After correction, instructions get to the integer block faster.

The integer core still contains two execution units and two address generation units (simply called AGen). This time, AGen's capabilities have expanded and they can process MOV instructions. When AGen block activity is low, the architecture will route MOV instructions through these channels.

One of the important changes is the increase in the buffer fast conversion addresses (TLB) for L1 cache from 32 to 64 entries. Because the L2 TLB has a fairly high 20-cycle latency, increasing the L1 cache hit rate can significantly improve performance in data-intensive applications. This is especially true for server environments, however, according to AMD engineers, some games are also sensitive to this, although they did not expect this.

L2 cache optimizations

Hardware sampling in L2 has also been improved. The minimum latency has not changed, which is why the cache latency in the Sandra 2013 test did not improve. However, the prefetch unit and L2 cache are used more efficiently, and AMD says average latency (which is very hard to measure) should be reduced. The same Sandra 2013 module shows minimal changes to L3 latency, and Vishera architects confirm that there have been no changes to the L3 cache shared across all FX processor modules.

Putting it all together: five architectures at 4 GHz

What effect do all these changes have on Piledriver performance? To compare relative results, we'll test five different architectures at 4 GHz.

However, this is a very generalized result. The subtests show how each platform affects the overall gaming performance score.

It seems that the only processor that stands out from the general pattern is the AMD Phenom II X6 1100T, and only by a couple of percent. All others show the same results, since the Graphics subtest isolates GeForce GTX 680 .

In the Physics subtest, CPU performance plays a major role as Futuremark divides the world into several distinct regions, creating multiple threads.

AMD FX-8350 faster than AMD Phenom II X6 1100T. But the six-core chip has lower power consumption, and if its speed is not much lower, efficiency may be at a higher level. This will be a disaster for AMD.

Unfortunately, the platform based on the Vishera chip is unlikely to be able to catch up Intel processors in terms of efficiency, since the results of average energy consumption differ dramatically.

This is where AMD can be proud of itself. AMD FX-8350 finishes second among the processors we selected (CPUs were selected in such a way that there was no large scatter in the results). It was very interesting how AMD FX-8350 will resist Core i5-3570K, and as a result, the new processor beat the more expensive model by 12 seconds.

The fact that AMD is asking for a new flagship that completed the tests almost 10 minutes faster than FX-8150, less than $200, only adds to its value. But what about efficiency?

AMD FX-8350 almost 13% more efficient than its predecessor. But more importantly, the new processor is more efficient than the Phenom II X6 1100T.

A year ago, Phenom surpassed Bulldozer in this regard. And there was no denying that AMD introduced a processor that used more power and ran slower. Today we recognize higher performance, improved efficiency and a more attractive price. Is this enough to get a recommendation?

AMD FX-8350 | There's still something missing

Having checked the new processor in all respects, we can confidently say that AMD FX-8350 Core i5-3570K, even despite the serious differences in efficiency and energy consumption. This is our opinion in 2012.

As already stated, AMD FX-8350 turned out to be much stronger FX-8150 and allows AMD to regain ground lost by the Bulldozer architecture. Piledriver doesn't fix all of Bulldozer's shortcomings, but fine tuning design and power consumption allow the company to increase the clock frequency without affecting the thermal package, which is 125 W. The changes are not so significant, but they are enough to create a good alternative the best models Intel Core i5.

Of course, if AMD decided to ask $245 for the new chip as for FX-8150, processor AMD FX-8350 I would expect the same “success” as last year’s model. However, the recommended price does not exceed $200. Thanks to this AMD FX-8350 stands on par with the Intel Core i5-3470, with a locked multiplier, which ended up behind in many benchmarks. In the same tests, the new FX chip outperforms Core i5-3570K for $230. It is only in single-threaded applications that Intel processors remain out of reach.

But there is still the issue of energy consumption. In Russia, thank God, electricity is relatively cheap. And it's unlikely that anyone will worry about the extra 50 W unless a loud cooler is needed to dissipate it. But, for example, residents of Denmark pay $0.40/kWh, and the difference is only 10 W between Core i5 and AMD FX-8350 idle time will cost a few dollars per month. A system running around the clock under load will already ask for an extra $15 per month. Advantage for Intel.

Still, let's try to decide on a recommendation. Considering that professional users prefer AMD over Intel, we think many enthusiasts will be seriously interested in AMD FX-8350, unlike Bulldozer chips, and rightly so. Everyone wanted to see more speed, improved efficiency and more low price, and AMD gives it all. Are there compromises? Yes, I have. Performance in single-threaded applications didn't impress us at all, and power consumption is still a sore point. But AMD FX-8350 At under $200, it can definitely be used in a mid-range workstation.

Would we choose AMD FX-8350 For new system? Most likely not. Even though AMD's architects have done a great job over the past year, performance still varies greatly depending on the type of task. And given that electricity is constantly becoming more expensive, and productivity is at a similar level, we would choose a more efficient option.

What makes up processor performance? Previously, there was a formula in use that described performance as the product of the number of instructions executed per clock cycle and the frequency at which this processor operates. Now a third factor has appeared in this formula - the number of computing cores. Therefore, a processor developer who wants to release a fast product has several ways to do this.

However, it's not that simple. Increasing the number of instructions executed by a computing core per clock cycle is a rather difficult task. Classic x86 program code assumes sequential execution of instructions, and therefore, in order to achieve their parallel processing, the processor must be equipped with highly efficient branch prediction and instruction reordering units, the implementation of which requires considerable engineering effort. At the same time, the complication of microarchitecture affects the physical dimensions of the crystal and leads to restrictions when increasing the number of cores. So if a manufacturer is going to make a processor with a large number of cores, then the microarchitecture should, on the contrary, try to simplify. It's not easy with clock speed either. A bet on its growth will again require changes to indoor units processor and lengthening its execution pipeline. The result is the following: in order for a processor to win a medal for performance, its developers must work hard to simultaneously optimize a number of parameters.

The problem also lies in the fact that any of the chosen ways to improve processor performance can be successful only for special cases. Not all programs can work effectively with a large number of cores. Some algorithms do not allow you to correctly predict transitions and reorder instructions. And in some cases, performance does not increase even with an increase in clock frequency, because there are some other bottlenecks in the system.

Finding the optimal balance is not easy, and what is considered the optimal criterion? We can only compare the performance of processors in a finite number of programs and select the fastest one for a given specific case. However, this does not at all guarantee that, using a different set of test tools, we will not get completely opposite estimates. Such a lengthy introduction is given here because today we are going to get acquainted with the new series of AMD FX processors - the flagship product of AMD, widely known under the code name Zambezi. This processor is based on the very controversial Bulldozer microarchitecture, which has already managed to collect a considerable bouquet of unflattering reviews. But the point is not that this microarchitecture is completely bad. When choosing the best balance of characteristics, the developers incorrectly assessed the needs of the majority of users and placed the main emphasis on the wrong factor in the “basic formula”. As a result, the initial plan to release a high-performance solution of a new generation went wrong, and AMD adherents, intrigued by the promises of a breakthrough, received something completely different from what they expected. However, is this a serious and objective reason for disappointment? We will talk about this in this material.

⇡ Counting kernels: eight or four?

Working on a new design for productive processors, AMD decided to prioritize the number of cores. This is a completely logical choice, based on the fact that over the years, multithreading software is becoming larger and larger, and the development of a microarchitecture designed for long-term development should primarily take into account not the current state of the market, but the observed trends. Eight cores, provided in the basic version of the new processor, are what AMD was going to conquer the market, where so far only chips were presented, the maximum number of cores in which was limited to six. ( Here we are only talking about desktop computers. — approx. ed. )

At the same time, the developers did not want to take the cores of the old K10 microarchitecture. Not only are they too large in physical size, but also, as can be seen from Llano, they are not prone to operating at high clock speeds even after being converted to modern 32nm technology. In addition, they do not support many modern features, such as AVX instructions. Therefore, to assemble eight-core processors, AMD made a new microarchitecture - Bulldozer. Representatives of the company prefer to say that its development was carried out from scratch, but in fact, in the Bulldozer cores you can find many references to another microarchitecture presented this year - Bobcat, aimed at use in compact and energy-efficient devices. However, the relationship between Bulldozer and Bobcat is quite distant, and we mention it only so that the general idea becomes clear - Bulldozer combines many relatively simple cores.

At the same time, we are not talking about the primitive combination of eight simple cores on one semiconductor chip. In this situation, the resulting processor would have very low single-threaded performance, and this would become a rather serious problem, since there are not so few programs that do not split the load into several computational threads. Therefore, firstly, the cores were optimized for operation at high clock speeds. And secondly, they were paired into dual-core modules capable of sharing their resources to serve a single thread. The result is a rather interesting design: the input part of the execution pipeline of such a dual-core module is common, and further instruction processing is divided between two sets of execution devices.

The basis of the Bulldozer design is what is conventionally called a dual-core module

Let us recall that the data processing process in a modern processor includes several stages: fetching x86 instructions from cache memory, decoding them - translating them into internal macro-operations, executing them, recording the results. The first two stages in the Bulldozer module are performed for a pair of cores together, and then for integer instructions, execution is distributed over two cluster cores or, in the case of real arithmetic, it is carried out in a block of floating point operations common to two cores.

Bulldozer modules are designed to process four instructions per clock cycle, and, thanks to macro merging technology, some pairs of x86 instructions can be considered by the processor as one operation. This means that, in general, the dual-core Bulldozer module is similar in power to a single core of modern Intel processors, which can also process four instructions per clock cycle and also support macro merges.

However, there are significant differences between the Bulldozer module and the Sandy Bridge core that can call their approximately the same theoretical speed into question. Due to the fact that the module of the new AMD processors contains the remains of two equal cores, it can demonstrate maximum performance only when processing a pair of threads. If it bears a single-threaded load, then the speed of its service will be limited by the number of execution devices within one such cluster. And there are not so many of them, given AMD’s desire to simplify individual cores - one and a half times less than in processors with Sandy Bridge or K10 microarchitecture. That is, two arithmetic ALUs and two address AGUs.

This is what the functional structure of a module built on the Bulldozer microarchitecture looks like. From two cores there are only two sets of integer actuators left

The block of floating point operations common to the processor module is also relatively low in complexity. It includes two 128-bit FMAC execution units, which can be combined into a single unit to process 256-bit instructions. It would seem that there are not so many actuators here, especially considering that they are divided into a pair of cores. But they are more universal than in previous and competing microarchitectures, which use separate multipliers and adders. And thanks to this, in certain cases when working with real numbers, a dual-core Bulldozer module can provide comparable and even higher performance than, for example, a single Sandy Bridge core.

A similar idea of ​​combining 128-bit devices to work with 256-bit instructions is used in Sandy Bridge

However, their most strengths The Bulldozer module should show up under a dual-threaded load. One Sandy Bridge core is also capable of processing two computational threads, for this it has Hyper-Threading technology. However, all instructions are sent to one set of actuators, which in practice causes numerous collisions. The Bulldozer module contains two independent integer clusters that can execute threads in parallel, and the total number of execution devices in them exceeds the number of such devices in the Sandy Bridge kernel by one and a half times.

On the left is the Bulldozer module, on the right is some competing core with Hyper-Threading support. In fact, it doesn’t look much like Sandy Bridge, but the illustration conveys the essence of the problem

As a result, the Bulldozer module has higher peak performance than the Sandy Bridge core, but this performance is somewhat more difficult to unlock. The Sandy Bridge core intelligently loads its own resources thanks to advanced on-chip logic that independently parses single-threaded code and executes it in parallel on full set their actuators. In Bulldozer, the task of effectively using actuators is partially shifted to the programmer, who must split his code into two threads - full loading of all the module’s capacities will become possible only then.

And that’s what’s typical. When considering the dual-core Bulldozer processor module, we constantly compared it with a single Sandy Bridge core, and at the same time we were able to draw quite correct parallels. This makes us wonder: shouldn’t the “eight-core” nature of the new microarchitecture be considered a product of marketers’ imagination? AMD says cores should be counted by the number of integer clusters, arguing that the module can provide up to 80% of the performance of two independent cores. However, we should not forget that the cores on which Bulldozer is based are significantly simpler than the cores of other processors. Therefore, the number of dual-core modules is a characteristic that reflects the performance of Bulldozer much more adequately.

Find the maximum number of processor cores and get a job in the AMD marketing department

⇡ Cache memory

The organization of cache memory in Bulldozer processors is also “tied” not so much to individual cores, but to dual-core modules. In fact, each core is allocated only its own first-level data cache; all other levels of cache memory relate either to the module as a whole or to the processor:

• Each core has its own L1 cache for data. Its volume is 16 KB, and the architecture assumes the presence of four associative channels. This cache operates with a write-through algorithm, which means it is inclusive.
• The first level cache for instructions is provided in a single copy for each dual-processor module. Its volume is 64 KB, and the number of associativity channels is two.
• The second level cache is also implemented in a single instance per module. Its size is an impressive 2 MB, associativity is 16 channels, and the operating algorithm is exclusive.
• In addition, the eight-core processor as a whole has an 8-megabyte L3 cache with 64-channel associativity. The peculiarity of this cache is that it operates at a significantly lower frequency compared to the processor itself, which is about 2 GHz.

The following table describes the ratio of cache memory volumes for eight-core Bulldozer, four-core Sandy Bridge and Thuban processors (six-core Phenom II X6, built on the K10 microarchitecture).

As you can see from the table, AMD relied on capacious upper-level caches, which can be really useful in the case of a serious multi-threaded load. However, the cache memory in new processors is generally slower than that of previous and competing products. This is easily detected when measuring practical latency.

Large delays when accessing data in Bulldozer can only be compensated by the high clock speed of these CPUs. Which, however, was originally planned - in terms of frequencies, the new eight-core processors were supposed to exceed the Phenom II by 30%. However, AMD was never able to design semiconductor crystals capable of operating stably at such high frequencies. As a result, high cache latency can cause significant damage to Bulldozer-based systems.

Not only fans of the company's products, but also many users who follow IT progress, are openly waiting for AMD processors with a fundamentally new Bulldozer architecture. Over the past few years, offering interesting solutions in terms of price/performance ratio, AMD has mainly focused on entry-level and mid-range devices. By reviving the FX line, it is obvious that the company expects to attract the attention of more discerning enthusiasts who are ready to experiment and demand maximum speeds. We will study the capabilities of the new family using the example of the world's first eight-core desktop processor - AMD FX-8150. Let's see if the manufacturer manages to meet the expectations of its fans.

Unlike its main competitor, which can afford to follow the pendulum principle of CPU development, changing architectures and technological processes every year, AMD does not outline a specific time frame for its projects, relying on market instincts and its own technological potential. The story of the Bulldozer architecture began a long time ago. It was assumed that it would be presented back in 2009, but due to various circumstances the practical implementation of bold engineering solutions in silicon has become possible only now.

Bulldozer for AMD is serious and will last for a long time. Over the next few years, this microarchitecture will become the basis for future processors from various segments: server, desktop and mobile. This applies to both discrete CPUs and hybrid ones - the APU is also planned to be transformed over time into Bulldozer. Only for compact systems AMD is going to use chips based on the economical Bobcat and its upgraded versions. With the announcement of Bulldozer, the company decided to revive the legendary series by introducing processors from the AMD FX line, which received a new architecture and are manufactured using the most advanced 32-nanometer process technology.

Architecture Features

Bulldozer chips are based on modules with two x86 computing units. However, the latter are not completely autonomous - some resources are common to both cores. Specifically, the prefetch unit, instruction decoder, FPU, and L2 cache. A monolithic dual-core module allows two threads to run simultaneously, but with certain caveats. According to the manufacturer’s calculations, this approach is completely justified and allows us to obtain about 80% of the efficiency of full-fledged physical cores. However, this significantly reduces the number of transistors, and, accordingly, the area of ​​the crystal and its power consumption.

Taking into account the new structure, the internal architecture was very seriously redesigned, which actually affected all execution units. There are practically no similarities with K10, which was used for Phenom II and Athlon II chips. AMD has introduced support for AVX, SSE 4.2 and AES-NI instructions and added its own FMA4 and XOP sets.

Like the top Phenom processors, the FX chips received a three-level caching system. However, its organization is also noticeably different from that of its predecessors. The L1 data cache has been reduced from 64 KB to 16 KB, while at the same time its throughput has increased significantly. L2 with a capacity of 2 MB is shared by both cores of each module. Depending on the number of the latter, the total capacity of the second level cache in the AMD FX processor can range from 4 to 8 MB. Its latency is slightly increased - a price for optimization for operation at higher frequencies. Chips with Bulldozer architecture are also equipped with an 8 MB L3 cache. Considering the exclusive operating scheme, the total buffer volume is quite impressive for desktop models. An improved data prefetch algorithm allows us to hope that the speed of the memory subsystem will be increased. As for the RAM itself, the CPU FX supports DDR3-1866 modules in dual-channel mode.

AMD FX is manufactured using a 32nm process technology with SOI technology, similar to that used in the manufacture of the Llano APU. The chips are produced at the facilities of sister company GlobalFoundries. The CPU is based on an eight-core crystal with an area of ​​315 mm2. According to the topology, most of it is allocated to cache memory, so it is not surprising that the total number of transistors in in this case is an impressive 2 billion. For comparison, the six-core Phenom II X6 (Thuban) includes “only” 904 million transistors, but due to the 45-nanometer process technology, the die area is 346 mm2. Considering the difference in area, we can assume that the cost of FX chips is lower than that of their predecessors. However, the transition to 32 nm is not easy for GlobalFoundries. AMD has already reported difficulties with the production of suitable blanks, due to which the company cannot fully satisfy the demand for hybrid Llano. Let's hope that this will not affect the availability of FX for sale and everyone will be able to purchase them.

The same crystal will be used for four- and six-core models, which will make it possible to more efficiently manage chips that have certain defects. Meanwhile, it is likely that fully operational crystals with deactivated modules will be used to produce these CPUs. And in this case, you can count on the next lottery with the unlocking of disabled cores. It would be a great way to stir up interest in AMD FX processors.

Turbo Core

Turbo Core dynamic frequency increase technology was previously used by AMD for six-core Thuban and Llano APUs. FX processors have a new mechanism and algorithm for this function. In the case when, under load, the power consumption of the chip is within the limits of its TDP, and the temperature does not exceed a specified value, the frequency can automatically increase (100–300 MHz) even in a situation where all cores are active (All Core Boost). If at least half of the modules are idle, then AMD FX can switch to Max Turbo Boost mode, increasing the supply voltage and very significantly the clock frequency of the working units (up to 900 MHz).

AMD is also concerned with improving the efficiency of the new chips. Given the growth in the number of computing cores, it is impossible to rely only on the effect of using a more refined technical process. When there is no load on both processor cores within the same module and they go into the C6 power saving state, power transistors allow you to turn off power from this node, reducing overall CPU consumption.

Logical support

Like the previous desktop AMD platforms, the PCI Express 2.0 bus controller remained the prerogative of the northbridge of the chipset, and did not move under the processor cover. It is the number of supported lines of this interface, and, as a result, the ability to build configurations with multiple video cards, that have become the defining differences of the new logic sets for Zambezi chips. The top-end AMD 990FX has 42 links at its disposal, with the ability to be configured for graphics needs as 2x16x or 4x8x. AMD 990X has 26 lanes and allows you to pair only two video cards in CrossFireX or SLI mode in a 2x8x configuration. Well, AMD 970, with the same number of PCI-E links, offers to be content with one adapter. In all cases, the periphery is served by the SB950 south bridge, which does not bring any interesting innovations: six SATA 6 Gb/s ports with the ability creating a RAID(0,1,5,10), up to 14 USB connectors 2.0, work with PCI. Alas, unlike the AMD A75 chipset for the FM1 platform, there is no support for the high-speed USB 3.0 bus.

AM3+ platform

To work with FX series processors, you need a motherboard with an AM3+ socket. This can be either a model based on the “new” AMD 9xx chipset, or a product with logic from previous generations. Compatibility with AM3 is theoretically possible, but is not guaranteed either by AMD itself or by motherboard manufacturers. It is possible that the latter will release firmware for their top solutions, but these are rather isolated cases. And even in such situations, FX chips will operate with a reduced speed of switching between Turbo Boost and Cool’n’Quiet states. At the same time, everything possible problems the operation of the system will fall on the shoulders of users. Therefore, you can’t count on a problem-free upgrade in this case.

Boards with AM3+ are easily distinguished by the black color of the processor socket, while the AM3 connector is white. Fortunately, the design of the CO mounting elements has not changed, so any cooler compatible with AM2/AM2+/AM3 will be suitable for cooling the AMD FX.

Model range

3DMark 11, CPU test (Physics), points

3DMark Vantage points

PC Mark 7, Computation test, points

CineBench 11.5, points

x264 HD Benchmark 4.0, fps

7-Zip 9.20, MIPS

Far Cry 2, 1920×1080, DX10, high quality, fps

Hard Reset, 1920×1080, High mode, fps

Metro 2033, 1920×1080, DX11, PhysX, high quality, fps

Colin McRae: DiRT 3, 1920×1080, high quality, fps

Lost Planet 2, 1920×1080, DX11, high quality, test B, fps

Crysis 2, 1920×1080, DX9, high quality, Downtown test, fps

System energy consumption, W

Thanks to the modular structure of the company's processors, it is easy to build its model range, offering devices with different numbers of computing units and clock speeds. At launch, the line of desktop chips, called Zambezi, includes four CPUs. The flagship is the eight-core FX-8150 solution with a frequency formula of 3.6/3.9/4.2 GHz. 8 MB of L2 and L3 cache, as well as a TDP of 125 W. The FX-8120 is similar in equipment, the only difference is in the frequency mode of operation - 3.1/3.4/4.0 GHz. The six-core FX-6100 has 6 MB of L2 cache and the same 8 MB of L3, but its thermal package is 95 W. The most affordable version with two modules and four computing units x86 FX-4100 operates at 3.6/3.7/3.8 GHz, is content with 4 MB L2, capacious L3 (8 MB) and a TDP of 95 W. As for the cost of the devices, the recommended wholesale prices for the listed models are at $245/205/165/115, respectively.

Overclocking

The ability to seamlessly overclock processors is one of the key parameters of FX chips. AMD places special emphasis on this feature. The free multiplier is available to all models in the line, and the ability to change it will be present on any board with AM3+.

The FX architecture was originally designed to operate at high clock speeds. Craftsmen armed with vessels of liquid nitrogen were able to get a screenshot of CPU-Z in a situation where the processor was operating at almost 8.5 GHz. At the same time, however, it was necessary to leave only one module out of four active. All eight cores were made to function at 8.1 GHz. Previously, such frequencies were achieved only by the most lightweight versions Intel Celeron for LGA775. Now enthusiasts will have a much more interesting object for overclocking experiments.

In the case of an air cooling system, you will have to be content with more modest results. When the supply voltage was increased to 1.45 V, the CPU worked stably at 4.6 GHz. It may not be as impressive, but the potential is clearly better than the 45nm Phenom II chips.

Results

Performance test results are presented in charts. The picture is sufficiently indicative to form an overall opinion about the capabilities of AMD's new development. FX processors, as expected, received an increase in performance in multi-threaded tasks - archiving, HD video encoding, rendering. Here, the eight-core chip is quite capable of competing with both the Core i5-2500K and the more expensive Core i7-2600K. However, as soon as it comes to applications with unimportant optimization for parallel code execution, AMD FX lose ground - the specific performance of their x86 units is even slightly lower than that of products with the K10 architecture. In games that, at best, use 3-4 threads, Intel processors have a noticeable advantage. If you use maximum graphics quality settings, where the video card becomes a limiter, the systems' performance levels out, but it is impossible to assess the real potential of the CPU in such conditions.

The transition to a 32-nanometer process technology, rather, made it possible to keep power consumption at the same level with increased performance. The priority in this case was probably performance rather than improved CPU efficiency.

Even judging by the cost of AMD FX, it is obvious that the company primarily plans to gain a foothold in the middle market. price category, consciously giving Intel the segment of top-end expensive solutions. In the current conditions, objectively, the manufacturer cannot now perform adequately in the league of “super heavyweights”. Having relied on multi-core computing, it is very problematic to get outstanding results in poorly optimized software. At the same time, just five years ago we were sincerely surprised who might need a quad-core processor on a desktop and how to effectively use the resources of such a CPU. Today this is commonplace, and the advantages of chips with such a number of computing units no longer raise questions. Perhaps eight-core models will receive similar recognition some time later.

Fortunately, AMD won't sit back and watch what fate befalls its processors. Announced plans further development inspire, albeit restrained, but still optimism. The company will continue to actively refine the current architecture, improving both energy efficiency and CPU performance, but the indicated pace - 10-15% per year - is not very impressive. With such indicators, one can count on a radical change in the situation only if Intel slows down the development of its products, but there are no prerequisites for this - the “tick-tock” mechanism has not yet failed. Already in the spring of 2012, Ivy Bridge chips will be presented, made using 22-nanometer technology and using 3D transistors.

The final assessment of the considered architecture and the AMD FX-8150 processor based on it is ambiguous, and this already indicates that the revolution did not happen. At least at this stage, it is invisible to the end user. A qualitative leap in performance occurs in well-paralleled applications, while in single-threaded tasks no significant increase is observed. The high expectations placed on Bulldozer were only partly justified. AMD still has some work to do to offer interesting solutions and compete for a place in the hearts of demanding enthusiasts.

The recent announcement of the latest AMD processors has become one of the most exciting events of this year. The intense anticipation, fueled by numerous information leaks and secret slides, left not only fans of the white-green camp, but also adherents of the competing company’s products in peace. The performance data was mixed, ranging from an overwhelming advantage over competitors to complete failure. No one will argue with the statement that the Stars microarchitecture, which underlies all current desktop solutions from AMD, is quite outdated today. The capabilities of the successors of the legendary K8, AMD Phenom II and Athlon II processors, no longer meet modern realities. That is why the launch of processors based on the fundamentally new Bulldozer architecture was extremely necessary. This would make it possible to match or even surpass competitors' solutions in terms of performance and energy efficiency. The advantage in performance should be provided by a fundamentally new eight-core architecture, and the introduction of a thin 32-nm technological process, together with “advanced” capabilities for controlling voltages and frequencies of individual functional units, promises a significant reduction in power consumption compared to previous generation solutions.

Finally, on October 12, the veil of secrecy was torn off: it was then that the long-awaited announcement of AMD FX processors, based on the Bulldozer microarchitecture, took place. The chipmaker presented a whole line of CPUs based on this microarchitecture, which includes four-, six- and eight-core models. Among other things, AMD has revived the "FX" brand name, which in the past was the name of enthusiast products. Indeed, all AMD FX processors of the current generation have a multiplier unlocked for increase, which, in theory, should make them attractive to overclockers. By flexibly varying the number of functional units and operating frequencies, AMD was able to fill almost all major market niches, ranging from low-cost gaming systems to offerings for configurations in the upper price range. The full lineup of the latest AMD processors compared to the four- and six-core Phenom IIs looks like this:

Bulldozer microarchitecture: structure and operating features

First of all, it should be noted that AMD FX are purebred central processors and do not have a graphics core. Of course, in this regard, one can accuse AMD of inconsistency, because promoting the APU (Accelerated Processing Unit) market is one of the company’s main strategic initiatives. Instead of a built-in video adapter, users receive full AMD FX compatibility with performance Socket platform AM3/AM3+, for which there are many excellent motherboards and support for all current expansion options is provided. AMD has released an updated 9th series of system logic sets especially for FX processors.

Let us recall the main features of the flagship AMD 990FX chipset. So, it allows you to build AMD CrossFireX and NVIDIA SLI graphics configurations, thanks to the south bridge SB950 supports the SATA 6 Gb/s standard, but lacks connectivity USB devices 3.0. As for Socket AM3 motherboards based on system logic sets of previous generations, after updating the firmware microcode they will also need to support Bulldozer. But this depends on the specific model.

One of the key features of processors based on the Bulldozer microarchitecture was the transition to a 32-nm lithographic process, which has been very successfully used by its main competitor, Intel, for almost two years. In addition to potentially reducing power consumption and improving overclocking potential, this fact had a positive impact on the cost of producing semiconductor crystals. AMD can no longer be called a newcomer to the development of the 32-nm process technology: it is with this level of detail that quite successful Llano APUs are produced, which, although they have not won recognition among enthusiasts, are excellent for building inexpensive and compact universal PCs. Thanks to the application modern technologies The production chip (despite the fact that it contains almost 2000 million transistors) turned out to be very compact. The eight-core AMD FX 8150 has a core area of ​​only 315 mm², which is smaller than the previous generation flagship Phenom II X6, whose die occupies as much as 346 mm². However, AMD FX processors are still far from the performance of quad-core Sandy Bridge processors, since in the former the chip, despite the presence of a built-in graphics accelerator, occupies only 216 mm².

The main innovations that were made in the Bulldozer microarchitecture affected the algorithm for performing multi-threaded calculations. For a long time, central processing units were capable of executing a single computational thread at a time. The so-called simultaneous operation of several programs was carried out using an interrupt handler, that is, the computing tasks of different applications in turn received short-term access to processor resources. It was thanks to this that it became possible work multitasking operating systems. Needless to say, the speed of operation in this mode was low. At the same time, CPU developers began to notice that under load, different functional blocks of the processor could be idle while others were busy processing the computational thread. This is what prompted them to sharing the same processor resources to process multiple computational threads. Intel Company introduced such a feature called Hyper-Threading into its processors back in 2002. This principle gives some gain in a certain type of task. At the same time, AMD's approach to implementing multi-threaded computing for a long time remained unchanged: each thread must run on a separate core. Now, after optimizing the performance of individual processor nodes and carefully analyzing the load, AMD developers considered that the performance of some nodes is quite sufficient to service two independent computing threads at once. This approach made it possible to greatly reduce the number of transistors while maintaining high productivity. Now, in light of increasing performance requirements while maintaining acceptable power consumption parameters, developers are forced to look for ways to increase the number of instructions executed per clock cycle.

So, at the heart of all AMD FX central processors is a semiconductor chip consisting of four computing modules, each of which is equipped with its own array of Level 2 cache memory, a shared Level 3 cache of 8 MB, a dual-channel DDR3 memory controller, and bus controllers HyperTransport and built-in north bridge.

The compute module includes two integer units (ALUs), each capable of executing up to four instructions per clock cycle, equipped with its own Level 1 cache for data storage. All other blocks, such as the branch predictor, instruction decoder, instruction buffer memory, and 2 MB L2 cache array, are presented in a single copy. Obviously, the developers considered that the performance of these blocks is sufficient to service two ALUs.

As you can see, the Bulldozer microarchitecture has very advanced computing capabilities, especially compared to previous generations of AMD processors. However, this technological advantage comes at the cost of careful optimization. program code. IN otherwise, especially in older applications, performance levels may not be as good as expected.

A few words should be said about the organization internal memory AMD FX, which became champions not only in the number of cores, but also in the total cache size. As we have already said, each of the integer calculation blocks has a 16 KB buffer for storing data, and both buffers can be used to operate the FPU block. To store instructions, each computing module has separate cache L1 has a capacity of 64 KB, and intermediate data is accumulated in the second level cache, the size of which is an impressive 2 MB. The level 3 cache array common to all four computing units has a volume of 8 MB and has associativity of 64 lines per module. Thanks to the use of an exclusive organization of caches of the 2nd and 3rd levels, we can talk about their total volume of 16 MB. It is not surprising that the Bulldozer crystal turned out to be so complex; the lion's share of the transistor budget is allocated to organizing the internal memory of the processor. Note that the operating frequency of the L3 cache memory can be 2000 MHz or 2200 MHz depending on the processor model.

As can be seen from brief description kernel design, the Bulldozer microarchitecture, despite all its innovations, is not without some shortcomings. Still, each computing module has only one branch predictor, an instruction fetch unit and one instruction decoder, which, by the way, is capable of processing no more than four instructions per clock cycle. Let's see how AMD FX behaves in real applications, but intuition suggests that in applications that actively use the FPU, but do not have software optimization for the new SIMD instruction sets, latest processors will demonstrate the level of performance characteristic of quad-core models.

In addition to the architecture, energy management mechanisms have also undergone changes. Despite the larger number of transistors and high clock speeds, even the older eight-core AMD FX have a thermal package that does not exceed 125 W. Of course, the 32-nm technological process also played a certain role in this, thanks to which the standard supply voltage does not exceed 1.4 V, but the main merit still belongs to the advanced mechanisms for adjusting clock frequencies and supply voltages. The first generation of this concept was implemented in the Phenom II X6, where in the case of a computing load of no more than three threads, the frequencies of the three active cores could be increased by 400 MHz. AMD FX processors offer much more flexibility in controlling key performance parameters. Thus, thanks to the use of power gate transistors, the processor energy saving manager is able to turn off entire functional blocks. When there is no load, the compute module, along with the L2 cache array, can be completely turned off, freeing up part of the TDP budget. At the same time, the clock frequency and voltage of active computing modules can be increased, and the frequency increase in Max Turbo mode reaches a respectable 900 MHz. Agree, we have never seen such an aggressive automatic overclocking algorithm. Moreover, with a uniform load on all computing modules, it is possible to increase the clock frequency to about 300 MHz. Actually, this is the mode Turbo works Core, and it will be active as long as the processor’s power consumption does not exceed the thermal envelope. In other words, the very concept of “standard clock frequency” for AMD FX loses its original meaning.

Sysadmin(he's on English sysadmin, system administrator) - an abbreviated name of a profession, the full name of which in Russian sounds like system administrator. This profession in lately has become very popular for most young and not so young people, they teach it, they work on it, they get good money for it. This is due to the rapid development of various computer technologies and their penetration into all spheres of human life. The word system administrator is often used in colloquial speech, in vacancies and resumes when searching for a job, in a word - everywhere. Below we will talk about what the profession is system administrator.

In modern realities, a system administrator can be called virtually any person who is involved in servicing and maintaining the work of a certain computer network, including all its hardware and/or software components, which may include:

• Personal computers, both workstations and servers;
• Network equipment such as switches, routers, firewalls and more;
• Web servers, mail servers, database servers, and others.

Also, in certain cases, the system administrator may be responsible for ensuring proper information security.

Depending on his specialization, a system administrator can engage in the following activities:

• An administrator of workstations and servers most often repairs both hardware (failed motherboards, burnt-out power supplies) and software (Windows won’t load, commas won’t print in Word...).
• Administrator corporate network based on domain Active Directory. A very popular activity, given the prevalence of Windows operating systems, as well as the need to somehow centrally control them. Such a specialist must be able to create, distribute into groups, edit users, give them the appropriate rights in the AD domain, and also be able to manage group policies for users, their computers and the groups they all belong to.
• Administration of networks and network equipment. His responsibilities include knowledge of network topology, the ability to work with both non-configurable and customizable network equipment, planning local computer network, as well as the ability to combine several remote workplaces into one network by setting up NATs and VPNs. You should also not forget to control access within this network and outside it - setting up a proxy.
• A web server administrator, who must at least be able to install, configure and maintain one of the following web servers - Apache, IIS, Nginx, and monitor hosting (which can be located both within the organization’s network and outside it). In addition, a good administrator should be able to configure normal resource distribution under high loads, clustering and many other specific things.
• Administration mail server is also a common task for a system administrator; his tasks include working with such popular solutions as Exim, Microsoft Exchange, Postfix, Sendmail, or corporate postal solutions from Google or, for example, Yandex. In addition to the obvious control over accounts (creation, deletion, configuration), it is also necessary to be able to set up an antispam system, etc.
• Site administrator. These responsibilities may include simply filling the site with some content, but since we are talking about a system administrator, then in theory he should be able to set up hosting (including a web server, as mentioned above), install and configure the desired site, for example any content management system (CMS).
• Very rarely, the responsibilities of a system administrator may include the task of creating or maintaining a video surveillance system. The tasks include installing and configuring cameras, responding to various events, saving and playing back recordings. He has a weak attitude towards system administration, and often falls into his responsibilities in conjunction with some other responsibilities.

Left out of the system administrator's occupations described above are such possible things as database administration (Microsoft SQL, MySQL and its multiple branches, Oracle, etc.), 1C administration (not to be confused with “1C programmer”), PBX and much more .