AMD produces not only high-quality and well-known for their performance (albeit energy-intensive) processors, but also Radeon video cards, the characteristics of which are sufficient to run the most productive games.

This technique, especially released over the past 2 years, allows you to work with resource-intensive applications (3D graphics).

Although, in order to choose the right model and determine if its capabilities are enough to fulfill your tasks, it is worth considering the parameters of the graphics processors in more detail.

In order to get acquainted with the characteristics of the equipment it was easier, you can draw up a table indicating the main values ​​that affect the performance and functionality of the video card.

These include the parameters of the bus (frequency and bit width), the type of memory used in the manufacture of the GPU, the technical process, the speed of the data and the amount of memory.

You can also pay attention to the consumption of electricity, on which the power of the computer's power supply and the method of cooling the device depend.

Memory frequency and bus width

The memory frequency of a video card, first of all, affects the speed of its operation. The average of this figure is 1000 MHz for HBM memory and 6000-8000 for GDDR5.

At the same time, the dependence of the card's performance on its frequency is not always directly proportional, since the second indicator that affects the device's throughput is the bus width.

The characteristics of the tire, first of all, depends on throughput memory of the video card.

The wider its width, the faster the data is processed by the graphics processing unit (GPU).

So, 64-bit boards are practically not used in modern computers, although they are still available in online stores.

More modern models of video cards have a capacity of 128 and 256 bits, top versions - 512 bits and higher.

The ten best AMD models to date have the following bit depth:

• RX 470, 480 and 380 series - 256-bit;
• 390 series R9 - 512 bits;

• the latest models, R9 Fury and Nano, equipped with a new type of memory - 4096 bits;
• One of the models produced using the new technology with the 18 nm process technology, RX, has a capacity of only 128 bits, which is why it has a low data transfer rate, although it is relatively cheap, representing a budget option for gamers.

The high bit depth of the latest AMD graphics cards, obtained through the use of multilayer memory modules, allows for a lower frequency, providing more power.

At the same time, the specific power consumption of the equipment (1 W of power per 1 GB / s data transfer rate) becomes lower - R9 models with HBM memory consume less electricity compared to other cards.

The main feature of the Radeon Fury and Nano is the ability to run more graphics-intensive applications and resource-intensive games with a high FPS (frame rate).

Memory type and size

Once considered the best choice for a graphics card, GDDR5 memory is starting to age.

Moreover, manufacturers declare that its capabilities are approaching its limit, and are beginning to look for new solutions. One of them is HBM technology, which is different:

• increased productivity;
• less need for electricity;
• the peculiarity of the organization of the memory subsystem.

For this reason, modern and more expensive video cards R9 Fury, Fury X and Nano, having a low frequency of 1000 MHz, work 33% faster compared to the flagship of the previous generation R9 390X - 512 GB / s instead of 384.

The same relatively new, but budget model RX 460 at a good frequency of 1212 MHz has a 5 times slower operating speed compared to the most powerful model of the manufacturer, since it not only has GDDR5 memory, but also a bit depth of 128 bits.

The memory capacity of modern Radeon graphics devices is at the level of 4096-8192 MB.

At the same time, modern games require at least 4 GB of memory to run with normal settings.

Although this indicator is not so important for HBM memory, attention should be paid to the bandwidth, which is higher than that of GDDR.

Technical process

The main structural elements of a processor, including a graphics processor, are transistors that transmit or block electricity in a certain direction.

The performance of the video card depends on their number, and this indicator, in turn, depends on the size of the transistors and the technology used in their manufacture.

Most graphics card designers, including AMD, use a 28nm transistor process technology.

All modern models have this value, except for the RX 400 series.

Next-generation GPUs are built on 14nm technology. And in the future, Radeon cards will be manufactured using the 7nm process technology.

The 14nm technology is believed to provide the graphics core with 2 to 3x faster performance and support for up to 3 independently operating monitors.

Bandwidth

The speed of data transfer using video cards primarily depends on the product of their effective memory frequency and bit width.

The larger this value, the faster information is transmitted, and, therefore, the games work better.

At the same time, the new HBM memory has 8 times more bit depth, which means that the frequency may be lower.

For example, for the R9 Fury X model, the bandwidth is (4096/8) bytes * 1 GHz = 512 GB / s. This value is more than enough to run any game at maximum settings.

The RX 460 128-bit video card can transfer only 112 GB / s of information (= 7000 * 128/8).

Power consumption and cooling

The power consumption of different video cards depends on various factors:

• the technology used to create the processor;
• memory type;
• graphics card power.

At the same time, even in one series of cards, you can find models with high energy consumption and low.

For example, the R9 390 and 390X models consume up to 275 watts of power and require a power supply unit of at least 500 watts.

The same figure for the more productive R9 Fury and Fury X. While the R9 Nano consumes only 175 watts, although in terms of performance it is not inferior to the others and even surpasses them.

And the inexpensive RX 460 only consumes 75W with an optimal power-to-power ratio.

Power up to 75W is provided by one PCI Express slot.

Exceeding this value is compensated by additional 8-pin sockets, each of which can supply up to 150 watts.

This means that one PCI slot is not enough to power modern AMD cards, and additional power is required.

The design of the cooling system also depends on the power consumption of the GPU:

• less efficient models are cooled by a conventional fan system;

• processors capable of running modern games require more serious cooling - liquid. For example, the R9 Nano's ventilation system includes not only a cooler, but also a vapor chamber with heat pipes. And the R9 Fury has a metal plate under the radiator.

conclusions

AMD, like its main competitor Nvidia, continues to increase most of the characteristics of its video cards.

And the Fury series surpasses the previous generation in most indicators (except for power consumption).

Although this only applies to older versions - the budget RX graphics cards, created on the basis of the new 14 nm process technology, are inferior to the old flagships and are comparable to the inexpensive models of the previous generation.

Therefore, when choosing a card for your PC, the main attention will have to be paid to the financial side of the issue - high costs will allow you to get better characteristics.

10 Mar 2016

This page below has links to download the latest free drivers for AMD graphics cards Radeon HD 8200 / R3, which is part of the Radeon HD 8000 series. The installation files are taken from the official site and are suitable for: Windows 7, 10, 8, 8.1, XP, Vista 32/64-bit (x86 / x64).

For the convenience of choosing the necessary files, below is the version of your Windows and its bitness ("bitness").

Your computer runs on:

1. Download (153.5 MB / version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / released on 08/12/2016)

   For Windows 7 32-bit

2. Download (239.8 MB / Version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / Released 08/12/2016)

   For Windows 7 64-bit

3. Download (134.8 MB / Version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / Released 08/12/2016)

   For Windows 10 32-bit

4. Download (208.24 MB / Version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / Released 08/12/2016)

   For Windows 10 64-bit

5. Download (205 MB / Version 14.4 (Catalyst Software Suite) / Released 04/25/2014)

   For Windows 8 32-bit

6. Download (260MB / Version 14.4 (Catalyst Software Suite) / Released 04/25/2014)

   For Windows 8 64-bit

7. Download (154.21 MB / version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / released on 08/12/2016)

   For Windows 8.1 32-bit

8. Download (239.88 MB / version 16.8.2 (Crimson Edition 16.8.2 Hotfix) / release date 08/12/2016)

   For Windows 8.1 64-bit

9. Download (179 MB / Version 14.4 (Catalyst Software Suite) / Released 04/25/2014)

   For Windows XP 32 and 64-bit

10. Download (151 MB / Version 13.12 (Catalyst Software Suite) / Released 12/18/2013)

    For Windows Vista 32-bit

11. Download (209 MB / Version 13.12 (Catalyst Software Suite) / Released 12/18/2013)

    For Windows Vista 64-bit

Fallback - Get Drivers Using AMD Driver Autodetect

This option is convenient because the program AMD Driver Autodetect will pick and download the latest working drivers which are suitable for your AMD graphics card and for your version of Windows. The program does not need to be installed, it was created by AMD and the files are downloaded from their official servers.

Instructions:

1. Run AMD Driver Autodetect and it will automatically pick up required files to install drivers.
2. Click the Download Now button to download the files.
3. Wait for the files to download and start the installation.

Solid state drives with a volume of 1 TB in the domestic market are in very limited demand due to their cost - too expensive. Therefore, we rarely test solutions of this class. But still we are testing - selecting the most affordable models at the time of writing the review. And we try to do it in such a way as to cover exactly the hardware platforms in general, and not just the models.

advertising

• SmartBuy Ignition Plus - from 16.5 thousand rubles - a representative of a rare hardware platform that combines Phison S11 and MLC 3D V-NAND Micron. Although the family of these drives itself has been on sale for a long time, the modification for 960 GB has just appeared;
• Micron MTFDDAV1T0TBN - from 16.8 thousand rubles - this furious name hides the Marvell 88SS1074 controller and the TLC 3D 32L V-NAND Micron, the essence is the Micron M1100 (and also Crucial MX300). We tested this hardware configuration in January;
• WD Blue PC SSD - from 17.1 thousand rubles - Marvell 88SS1074 and 15 nm planar TLC NAND SanDisk. We studied this configuration a little over a year ago. The term is quite long, but no fundamental changes have taken place during this time, and therefore it can be considered that no special issues should arise with this solution;
• Samsung 850 Evo - from 17.2 thousand rubles. A beaten up and down model, tested by all and sundry. I don't see much sense;
• Crucial MX300 - from 17.5 thousand rubles - has already been mentioned a couple of paragraphs above;
• Intel SSDSCKKW010X6X1, aka Intel 540s at a price of 17.9 rubles - also reviewed a year ago;
• WD Blue 3D NAND SATA SSD - at a price of 17.9 thousand rubles - a new product that just went on sale, a week ago we tested a modification for 500 GB and got some idea of ​​the potential of this modification;
• AMD Radeon R3 - 18 thousand rubles - we have not studied this drive yet;
• Kingston SSDNow UV400 - from 18.5 thousand rubles - the same;
• Toshiba HDTS896EZSTA - from 19 thousand rubles - the combination of Phison S10 and 15 nm planar TLC NAND Toshiba is familiar from the Toshiba TR150 test a year ago, in fact, the remnants are already being sold, in the near future the drive should disappear from sale, since it has been discontinued;
• Corsair Force LE - at a price of 19.5 thousand rubles - is essentially the same Phison S10 and 15 nm TLC NAND Toshiba.

In total, only three drives are unfamiliar to us in this weight category and with price tags that can somehow still be called "budget" (up to 20 thousand rubles): SmartBuy Ignition Plus, AMD Radeon R3 and Kingston SSDNow UV400. With the first - alas, the question is still in limbo (administrative difficulties), the second was chosen for this material, and with the third we will try to resolve the issue.

Let me explain separately: I did not distinguish solutions in the M.2 form factor in any way. The fact is that such a drive easily "turns" into 2.5 "- corresponding housing-adapters with a shaft on AliExpress / eBay at a price of 200-250 rubles with delivery.

And the difference between M.2 and the "original" 2.5 "from a hardware point of view is vanishingly small. And even the seemingly denser layout with the emerging risk of overheating under heavy loads is actually no longer like that. In 2.5" packages, full-size PCBs have become vanishing view - in recent times manufacturers in order to reduce costs en masse have switched to shorter boards.

AMD Ryzen 3 2200G / Radeon Vega 8 vs. Intel Pentium G4560 / GeForce GT 1030: Which to Choose?

Test bench with:

• ASRock A320M-HDV
• MSI B350I PRO AC
• CHIEFTEC GPE-500S 500W
• Vinga CS207B

In contrast to it, there is a configuration based on a 2-core 4-thread with a Vinga CL-2001B cooler, an ASRock H110M-HDS motherboard and a low-profile MSI GeForce GT 1030 video card with 2 GB GDDR5 memory. It stands out for its slight GPU overclocking: 1265/1518 instead of the reference 1227/1468 MHz. The effective memory frequency is 6 GHz. The rest of the components of the test systems are the same.

Testbed with Intel Pentium G4560:

• ASRock H110M-HDS
• Vinga CL-2001B
• MSI GeForce GT 1030 2G LP OC
• 2х 4GB DDR4-2400 GOODRAM (GR2400D464L17S / 4G)
• SSD AMD Radeon R3 120GB (R3SL120G)
• HDD i.norys 1TB (INO-IHDD1000S2-D1-7232)
• CHIEFTEC GPE-500S 500W
• Vinga CS207B

At the time of preparation of the material, the total cost of the configuration based on AMD Ryzen 3 was about $ 384. A competitive system based on the Intel Pentium G4560 cost $ 435, or 13% more. For the purity of the experiment, we took almost all prices from the price list of one store, but we do not exclude that in other stores the price tags for individual goods may be higher or lower, therefore, these figures are very indicative. And, of course, we do not claim that the indicated assemblies are optimal, because everyone selects a system based on their own needs.

Now let's take a look at what these systems are capable of in versatile games at Full HD resolution. Graphics profiles were selected in such a way that the embedded video could handle the launch AMD core Vega 8.

Benchmark World of Tanks Encore with an average preset, it produces an average of 56 FPS with drawdowns up to 26 in a system with AMD Ryzen 3. The opponent's results are 30-50% higher. And the frame timeline is much smoother and quieter, so a system with a discrete graphics card looks better.

V Rainbow six siege I had to go down to a low profile in order to get playable performance on Ryzen 3: an average of 62 FPS with drawdowns up to 28. In turn, a bundle of Intel Pentium G4560 and GeForce GT 1030 produces on average only a little more - 66 frames / s. But the increase in the minimum frame rate, rare and very rare events exceeds 50%. That is, the comfort of the gameplay will be higher in a system with a discrete video card.

Watch dogs 2 is considered a processor-dependent game, therefore, even with a low preset, the Pentium is sometimes loaded to capacity. The processor part of Ryzen 3 does better - 4 full cores make themselves felt, but the video core does not pull out and drawdowns up to 14 FPS occur, while with the GeForce GT 1030 the speed does not drop below 21 frames / s. In general, the advantage of the second configuration can be estimated at 40-60%.

V PUBG I had to choose a very low profile and reduce the draw scale to 70%. But this also did not save from freezes up to 16 FPS in both cases. Moreover, very rare events in the system with the GeForce GT 1030 were lower than in the AMD Vega 8, but in other parameters it was 50-60% ahead. And the timeline of the frame is quieter.

Jogging in Novigrad in Third Witcher took place with low graphics and post-processing presets. On average, the system with the GeForce GT 1030 looks better: 34 versus 29 FPS, but the rest of the statistics speaks in favor of AMD Vega 8, although the gap is best case is only 2 FPS. The lack of processing power is clearly affecting.

Heavy Assassin's Creed Origins you can run it on iGPU at a very low preset, but the results will not please you - on average 27 FPS with a drawdown of up to 12. To complete it, you will have to switch to HD without fail. The combination with the GeForce GT 1030 also does not shine: an average of 33 FPS with drawdowns up to 13. But the statistics of very rare and rare events are much better: 22-25 against 12-17 FPS.

Network mode Battlefield 1 cannot be synchronized, so it is difficult to talk about repeatability of results. Nevertheless, with a low preset, the indicators of the minimum speed, rare and very rare events in both systems are at approximately the same level with an advantage of 1-3 FPS in favor of the GeForce GT 1030. At the average frequency, it goes ahead by 28%.

Completes the first test block benchmark Far cry 5 at a low preset. Here the processor load is not as big as in Battlefield 1, which allows you to feel the advantage of using a discrete graphics card in every statistic: the difference is in the range of 10-60%.

Nice 21.5-inch ASUS ET2230AGK monoblock based on AMD Beema

If you are looking for an All-in-One computer for office work, study or entertainment, then take a look at the ASUS ET2230AGK model. It uses a 21.5-inch Full HD screen with high-quality and natural color reproduction.

The novelty is built on the basis of 4-core energy-efficient APUs of the AMD Beema series, which are supplemented by a mobile video card entry level(AMD Radeon R5 M230 or Radeon R5 M320), DDR3L RAM and HDD capacity from 500 GB to 1 TB. Additionally, the package includes an optical DVD drive RW, a pair of stereo speakers supporting ASUS SonicMaster technology, a set of necessary network modules and external interfaces, and a webcam with a microphone. That is, the novelty is a device completely ready for work and entertainment, which occupies only the monitor space in the workplace.

It will go on sale with Windows 8.1 preinstalled. Technical specification monoblock ASUS ET2230AGK is presented in the following table:

Computer selection 2015. Winter

After a long break, we decided to continue publishing analytical materials on the choice of components. Of course, the situation in the country affected the domestic IT market and the purchasing power of citizens. However, judging by the comments to the reviews and messages on the specialized forums, the issues of assembling the optimal configuration still do not lose their relevance. In addition, exactly a year has passed since the publication of the article "Computer choice 2014. Winter". During this seemingly insignificant period of time, a lot of changes have occurred in the IT industry: several new platforms have appeared, promising technologies and standards have been released, many PC components have stepped to a higher level of performance. In such a whirlpool of events, and even with constant fluctuations in the course, it is sometimes difficult for even experienced users to keep track of all the changes. What then can we say about those who are interested in the world of digital technology only at the level of a common man in the street. Naturally, in such conditions, choosing the optimal PC for them can turn into a real horror. We hope that this material will help simplify this task at least a little, as well as assess the state of domestic market components at the beginning of 2015.

As before, when making configurations for certain tasks, first of all, the following set of components will be considered: motherboard + processor + video card + RAM + drives + power supply + cooling system + case. The rest of the components (monitor, keyboard, mouse, etc.) are deliberately not included in the list, because their choice is greatly influenced by the subjective factor. In this case, advising something specific is not entirely correct.

Also, we will continue to abstract from any brands, and if we meet somewhere specific titles, then they should be considered only as an example, and not as a call to purchase. However, if some model turns out to be much better than its counterparts, of course, this point will be surely noted in the article. We took all the indicated prices from popular online stores and deduced the average value. It is quite possible that it is in your city that the cost of some components will be higher or lower. And in today's conditions, such a situation is more than real, especially if we talk about the same components delivered to the country at different times. Therefore, being guided by this material when choosing a PC, you need to understand that the prices are approximate and are only indicative.

Well, we figured out the official part, now you can go directly to the computer configurations. In order to increase functionality and cost, they can be placed as follows:

• computer for study and surfing the Internet;
• office computer;
• HTPC;
• HTPC, which combines the functions of a mini-PC;
• home computer to run modern games at minimum / low graphics settings;
• home computer to run modern games at low / medium graphics settings;
• home computer for running modern games at medium / high graphics settings;
• home computer for running modern games at high / maximum graphics settings and high resolutions;
• home computer for running modern games at ultra-high graphics settings and in high resolutions;
• computer for multi-monitor systems and workstations.

Monoblocks MSI Adora20 5M, AE200 5M and AE220 5M based on AMD Beema APU

MSI has launched three all-in-one PC models at once: MSI Adora20 5M, AE200 5M and AE220 5M, which are based on different AMD Beema series APUs. So, the 19.5-inch MSI Adora20 5M is equipped with a 4-core SoC-processor AMD E2-6110, clocked at 1.5 GHz. But the 19.5-inch MSI AE200 5M and 21.5-inch MSI AE220 5M are based on the more powerful 4-core version of AMD A4-6210 clocked at 1.8 GHz.

The video subsystem of all new products is assigned to the graphics core integrated into the APU, and two SO-DIMM slots are available for installing RAM modules. The disk subsystem of the MSI Adora20 5M solution can use one 2.5-inch drive, while the MSI AE200 5M and AE220 5M come with pre-installed 3.5-inch HDDs of 500 GB or 1 TB.

All three models also boast support for the necessary network modules and external interfaces, a pair of 3-watt speakers, a Tray-in DVD Super Multi optical drive, a webcam and a card reader. Particularly noteworthy in the new products are the displays used, which have support for Anti-Flicker and Less Blue Light technologies to reduce eye strain.

MSIAE220 5M

comparison table technical specification new monoblocks from MSI is as follows:

Review and testing of the AMD Athlon 5150 processor

Not so long ago, the world was presented with a new energy-efficient AMD AM1 platform and a number of processors for it. We have already managed to get acquainted with three of them (, AMD Sempron 3850 and) in practice. In this review, we will continue to explore the capabilities of representatives of the AMD Kabini family and take a closer look at the model. It is, so to speak, a lite-version of the series flagship (AMD Athlon 5350 processor) and differs from it only in clock frequency.

Specification:

Packaging, delivery set and appearance

All AMD Kabini APUs, including the AMD Athlon 5150, ship in the same red and white box. The only difference lies in the emblem and information on the sticker, on which the manufacturer has traditionally placed only the main technical characteristics: clock frequency (1.6 GHz), L2 cache size (2 MB) and the number of processor cores (4). It is also noted there that the cooling system is already included in the package.

The box contains:

• processor packed for additional protection in a plastic blister;
• cooler;
• user guide;
• AMD Athlon series APU logo sticker.

Outwardly, AMD Athlon 5150 does not differ in any way from the solutions of the AMD Kabini family reviewed earlier. The heatsink cover contains the series name and model designation. The countries where the crystal was grown (Germany) and where the final assembly of the processor took place (Taiwan) are also indicated. The arrangement of contacts on the rear side corresponds to the socket AM1.

Standard cooling system

All solutions from the AMD Kabini family have the same TDP level (25 W), so it is quite logical that their stock coolers are identical. In addition, this versatility saves money when developing processors, since there is no need to re-calculate the cooling system parameters for each group of models.

Although the developers are unlikely to have spent a lot of money on the creation of this cooler, because its design is extremely simple: a small aluminum radiator consisting of four sections of aluminum fins is cooled by a low-profile 50 mm fan.

It is noteworthy that the height of the cooling system is only 40 mm, which will allow it to be used in very compact cases, which are often the basis for nettops and multimedia PCs (HTPCs). Recall that coolers with mounting for motherboards equipped with Socket AM3 / AM3 + / FM2 / FM2 + sockets, for AMD platforms AM1 will not work.

CPU heatingAMDAthlon 5150 idle

CPU heatingAMDAthlon 5150 at maximum load

In practice, the standard cooling system has proven itself quite well. During a prolonged stress test for processor cores and integrated graphics core, the AMD Athlon 5150 temperature did not rise above 47 ° C, but during idle time it was 33 ° C. At the same time, the fan rotation speed varied in the range of 1300 - 2600 rpm. The maximum value is 4000 rpm, which can be reached by activating the corresponding profile in BIOS menu motherboard. As for the noise characteristics, the cooler works quite quietly up to 3000 rpm, and after overcoming this threshold, a distinguishable background appears.

Analysis of technical characteristics

In normal operation, the AMD Athlon 5150 speed is 1600 MHz with a reference frequency of 100 MHz and a multiplier of "x16". At the time of taking the readings, the voltage across the core was 1.296 V.

In idle mode, the multiplier is reduced to "x8", thereby dropping the frequency to 800 MHz. The voltage is 1.092 V.

The AMD Athlon 5150 cache is allocated in the same way as in the previously reviewed 4-core AMD Kabini models:

• cache memory of the first level L1: for each of the 4 cores 32 KB are allocated for data with 8 associativity channels and 32 KB for instructions with 2 associativity channels;
• L2 cache memory: 2 MB for all cores with 16 associativity channels;
• there is no L3 cache.

The DDR3 memory controller operates in single-channel mode and is guaranteed to support modules with frequencies up to 1600 MHz. The maximum memory capacity can be up to 16GB.

The GPU-Z utility did not correctly determine the characteristics of the integrated graphics core, so for these purposes we used another popular diagnostic program - AIDA64.

AMD Athlon 5150 has a video core from the AMD Radeon R3 Graphics series, codenamed AMD Radeon HD 8400 which is built on the leading AMD GCN microarchitecture. It includes 128 stream processors, 4 rasterization modules and 8 texture units, and the clock speed is 600 MHz. To save energy when there is no heavy load on the iGPU, its frequency is automatically reduced to 266 MHz.

By the way, exactly the same graphics core is used in the flagship model of the AMD Kabini family. Therefore, we can assume that both APUs (AMD Athlon 5150 and AMD Athlon 5350) will show approximately the same results in games. However, for a more accurate answer, let's take a look at the test results.

AMD's new Kaveri desktop APUs start appearing on motherboard support lists

At the moment, only two representatives of the AMD Kaveri desktop APU line are available on the market: AMD A10-7850K and AMD A10-7700K. It is not known why AMD delayed the release of the rest of the models, but they have already begun to appear on the support lists of some motherboards, which indicates an imminent debut.

In particular, models AMD A6-7400K, AMD A8-7600 and AMD A10-7800 were seen on the websites of MSI and Biostar. The AMD A6-7400K version is equipped with two processor cores with a base frequency of 3.5 GHz. The volume of its L2 cache is 1 or 2 MB, and a solution of the AMD Radeon R3 or AMD Radeon R5 series is used as a video adapter. It is difficult to say more precisely, since the information is contradictory. It is known for sure that its TDP is 65 W.

One of the most interesting is the 4-core AMD A8-7600 model. In the nominal mode (TDP at 65 W), its processor cores operate at a base / dynamic frequency of 3.3 / 3.8 GHz, respectively. However, the user can put it into a power-saving mode of operation (TDP will be 45 W), while the speed indicators will drop to 3.1 / 3.1 GHz.

APU AMD A10-7800 will be of interest to those who want to get high performance without planning to use additional overclocking. The base frequency of its 4 processor cores is at 3.5 GHz. The video accelerator of the AMD Radeon R7 series consists of 512 stream processors and operates at a frequency of 720 MHz, which will allow it to demonstrate a fairly high level of performance. At the same time, the indicator of its TDP is set at around 65 W.

Technical Specification Summary Table new APUs AMD Kaveri series:

AMD Sempron 2650 processor review and testing

Ultra-budget processors are always in steady demand among buyers due to their undeniable advantages. They make it easy to build an inexpensive work or first school computer for a child that is powerful enough to run standard everyday applications.

Since 2004, the AMD Sempron family has been replenished with different processors, but all of them were united by a common attitude towards the lower price range. With the release of the new energy-efficient AMD AM1 platform, AMD has redesigned them and moved from classic CPUs to hybrid devices with an integrated graphics core - APU.

The new AMD Sempron APUs are based on the AMD Jaguar microarchitecture. In accordance with the SoC (System-on-Chip) design, they combine computing and graphics cores, a RAM controller, and a chipset. At the moment, the new series includes two models: AMD Sempron 2650 and AMD Sempron 3850, the summary table of the technical specifications of which is as follows:

This review will focus on the dual-core model, which has a good chance of success in the lower price bracket.

AMD Sempron 2650 comes in a small white box made of heavy cardboard. It has a small transparent plastic window that allows you to evaluate the appearance of the processor.

On one of the sides, the manufacturer noted the scope of the novelty (solving everyday tasks, in other words - working with documents and multimedia files, as well as surfing the Internet). On the opposite side there is a sticker with a security hologram and a serial number of the product.

The AMD Sempron 2650 package includes:

• cooling system;
• brief instructions for installing the processor;
• sticker on the computer case.

The instruction itself, step by step, with the help of visual pictograms, shows the entire process of not only installing the APU into the connector, but also correctly fixing the complete cooling system.

The cooler consists of a small heatsink, which is fixed to the motherboard with two spring-loaded clips, as well as a fan installed on it. In this case, a Foxconn PVA050E12L model with a diameter of 50 mm with an operating voltage of 12 V and a current strength of 0.16 A is used as a propeller.

It is curious that the area in contact with the processor through a thin layer of thermal paste has a round shape.

We also checked the efficiency of the standard cooling system on an open test bench. The operating speed range of the complete fan in automatic mode is between 1300 and 4000 rpm. Up to 3000 rpm it remains virtually silent, with only subtle background noise appearing at 4000 rpm. In the normal operating mode of the turntable, the GPU temperature does not exceed 28 ° C, and the processor cores - 40 ° C, so there is no need to worry about overheating.

On the AMD Sempron 2650 case, in addition to the marking, the countries of production are indicated: the crystal itself was grown in Germany, and the final assembly took place in Taiwan. back side contains a set of contacts compatible with the latest socket - Socket AM1.

We also remind you that when installing the APU into the connector, you need to be extremely careful not to damage the sufficiently long and thin copper contacts.

The main advantage of the AMD Sempron 2650, which can be highlighted in the specification table, is the rather low TDP level (25 W). This makes it possible to use not only a compact and low-noise active cooler, but also a completely passive cooling system.

When performing stress tests, the APU multiplier was at the maximum mark "x14.5", and the clock frequency at the time of taking the readings was at the level of 1447 MHz. The voltage across the core was 1.288 V.

In idle mode, the frequency dropped to 798 MHz with a multiplier of "x8" and a supply voltage of 1.072 V.

Now let's examine the cache allocation scheme. For data caching, 32 KB of L1 cache per core with 8 associativity channels are allocated, and 32 KB of L1 cache per core with 2 associativity channels are allocated for instructions. There is also 1024 KB of shared L2 cache with 16 associativity channels. This processor is not equipped with L3 cache memory.

The built-in RAM controller operates in single-channel mode and supports DDR3 modules with a frequency of up to 1333 MHz. Support for modules with a frequency of 1600 MHz and higher (with automatic downgrading to the nominal 1333 MHz) depends on the specific motherboard model with which this APU will be used.

Review and testing of the AMD Athlon 5350 processor for the AMD AM1 platform

“Even a slight flap of a butterfly's wings at one end of the world
can provoke a tsunami on another "
Butterfly Effect from Chaos Theory

In 2011, AMD began a transition to the active use of the APU design in the budget segment of AMD processors, which implies the integration of the cores of the central and graphics processors, as well as the memory controller on a single die. The first to hit the market were the AMD Zacate (AMD E) and AMD Ontario (AMD C) series, which were targeted at netbooks, nettops and entry-level notebooks. This approach made it possible to abandon the design of printed circuit boards using chips from Severny and South bridges... The first of them became part of the processor, and the second became known as the "Chipset". This greatly simplified the board layout and design of the cooling system, increased the speed of operation of individual components and reduced the total cost of production.

The next evolutionary step was the transition to SoC (System-on-Chip) design. It involves the integration of the chipset into the processor, that is, together with computing functions The CPU is performing and coordinating, ensuring the correct interaction of many internal interfaces. The result is increased ease of design and wiring for motherboards, and eliminates the need for many additional controllers. All this leads to a further decrease in the cost of production, which has a positive effect on the final price.

The first APUs with support for SoC design in the AMD lineup were the 28nm solutions of the AMD Temash and AMD Kabini series, which replaced the 40nm models of the AMD Ontario and AMD Zacate series. They are designed for use as part of budget tablets, nettops, all-in-ones and laptops. There are even desktop motherboards on the market with integrated AMD Kabini APUs that enable entry-level systems for everyday computing or multimedia entertainment.

The only controversial point in the first SoC-processors from AMD is the use of a BGA package, which involves factory soldering the CPU to a connector on the motherboard. On the one hand, this approach reduces the production cost, and on the other hand, the process of replacing such a processor becomes much more complicated. And if for laptops this is considered the norm and does not cause massive complaints, then many owners of desktop PCs highly value and value the ability to freely update the configuration by replacing the processor.

Therefore, AMD decided to create desktop versions of the AMD Kabini APUs, housed in a PGA package, which makes it easy to change the processor if necessary. It should also be added that AMD decided to use well-known brands - AMD Athlon and AMD Sempron for naming new APUs, thereby reviving the competition of these chips with solutions of the Intel Pentium and Intel Celeron series (Intel Bay Trail platform).

Now let's go through the key aspects of the AMD AM1 platform presentation and consider the main features of the new processors. To begin with, AMD decided to give a reasonable answer to the question: "Why release a new budget platform at all?"

According to data from IDC for the fourth quarter of 2013, the majority of the desktop market (38%) is occupied by entry-level solutions. Mainstream PCs account for 30%, while performance desktops account for 32%. So the market budget systems large enough, so AMD did not want to give it completely to the Intel Bay Trail platform and prepared its own alternative, which looks very worthy, given the specifics of requests in this area. Particularly high hopes for the AMD AM1 platform are placed in emerging markets where pricing is paramount.

That is why AMD decided to take advantage of the rather successful 28nm AMD Jaguar microarchitecture to create the next generation of AMD Sempron and AMD Athlon processors. As mentioned earlier, they combine four CPU cores, a graphics adapter with AMD GCN microarchitecture and a single-channel DDR3-1600 memory controller on a single die, supporting up to 16 GB of total capacity.

Additionally, they support a number of controllers that are part of the chipset in traditional systems. In particular, this applies to:

• SD memory cards up to 2 TB;
• two USB 3.0 ports;
• eight USB 2.0 ports;
• PS / 2 interface and various internal sensors (temperature, fan speed, etc.);
• video ports eDP, DisplayPort / HDMI and VGA;
• four lines of the PCI Express x16 interface for connecting a discrete video card;
• two SATA 6 Gb / s ports;
• four lines of the PCI Express x1 interface, one of which is used to connect a gigabit network controller.

The specialists of AMD did not forget to remind about the improvements brought by the 28nm AMD Jaguar microarchitecture. The 40nm AMD Bobcat was taken as a basis, but the transition to a new technical process allowed an increase in the number of structural elements and optimization of all key blocks. It is not worth blaming AMD for improving the microarchitecture instead of implementing a radically new one, because there is an unwritten rule: "when changing the technical process, you should not change the microarchitecture either, in order to avoid many mistakes." Therefore, we can expect more significant changes in future versions of the AMD AM1 platform processors. In this case, engineers improved the integer processing units (IEU) and fractional numbers (FPU), redesigned the load / storage queue, provided 128-bit access to the FPU, allocated more resources for the operation of prefetch blocks, added support for new instructions (SSE4. 1 / 4.2, AES, CLMUL, MOVBE, AVX, F16C and BMI1) and many more improvements.

Much similarity can be found in AMD Steamroller (AMD Kaveri APU) and AMD Jaguar microarchitectures: the same OOO (Out-of-Order) design, 28nm process technology, support for new instruction sets, etc. However, there are significant differences. The first is size: four AMD Jaguar processor cores take up an area equivalent to one AMD Steamroller dual-core module. Important differences between AMD Jaguar's energy efficient microarchitecture and AMD Steamroller are also: support for 32KB L1 data cache instead of 16KB, use of an FPU in each core, and shared access to L2 cache for all cores. Recall that AMD Steamroller assumes the use of one FP-block per dual-core module. The L2 cache is allocated on the same principle.

As a result of all the improvements, the AMD Jaguar microarchitecture IPC (Instruction Per Clock) is 17% higher than AMD Bobcat. Performance in single- and multi-threaded tasks has increased significantly, which is good news.

The integrated graphics adapter uses the already familiar AMD GCN microarchitecture, which is also present in the. Before us is the same structure of computing clusters CU (Compute Unit), which include four vector blocks and a scalar coprocessor. In turn, each vector block contains 16 stream processors, so the total number in one CU is 64. Since the first APUs of the AMD AM1 platform use a maximum of two CU clusters, the total number of stream processors in them is 128.

Another curious moment in graphic adapters, which is associated with their name, deserves attention. Initially unofficial sources indicated the use of the "AMD Radeon HD 8000" naming scheme. In the official presentation, the name "AMD Radeon R3" is used, which greatly simplifies the classification of the performance level of the graphics adapter in the current structure of AMD. As a reminder, the first AMD Kaveri APUs are equipped with AMD Radeon R7 graphics. As a result, the name AMD Radeon R5 remains free, which, most likely, will be used in less powerful APUs of the AMD Kaveri line. They should appear in the second half of 2014.

The results of comparative testing in popular synthetic and gaming benchmarks of the flagship model AMD Athlon 5350 look very impressive. It confidently outperforms its main competitor in the form of Intel Pentium J2900. In undemanding games, AMD Athlon 5350 even outperforms the Intel Celeron G1610 processor bundle and the discrete NVIDIA graphics cards GeForce GT 210.

The test results are even more impressive after comparing the cost of these models, since an APU from AMD together with a motherboard will cost less than one processor from Intel. But the cost for entry-level platforms plays a very important role.

It is in the APU of the AMD AM1 platform that a productive graphics adapter is a very important advantage, the capabilities of which are enough for fast and high-quality processing of the operating system interface, video playback high resolution(4K Ultra HD), wireless video transmission (Miracast), launching undemanding games, fast photo editing and other similar tasks. Considering that such systems usually do not rely on the help of a discrete video card, then AMD's APUs look very cool against the background of competitors. In addition, we continue to collaborate with many popular software developers to optimize their products for the microarchitectural features of AMD solutions.

At the end of the presentation, AMD recalled the positioning of all its desktop platforms: AMD AM1 - entry-level systems, AMD FM2 + - mainstream computers and AMD AM3 + - high-performance PCs.

A summary table of the technical specification of the first APUs of the AMD AM1 platform is as follows:

Now let's move on to review and test the flagship APU model of the AMD AM1 platform -. Is the performance level of the new product really as good as indicated in the presentation? Does it have any other hidden advantages or disadvantages? We will try to answer these questions further.

PlatformAMDAM1

We tested not only a representative of the AMD Kabini family, but also a whole system at once (processor + motherboard + RAM). This will give us the opportunity to fully assess the capabilities of the entire AMD AM1 platform, as well as allow us to understand what tasks it is best suited for.

Let's start with the motherboard - the basics of the entire computer. In our case, it is represented by the model ASRock AM1B-ITX, made in Mini-ITX format. This form factor will be the main one for the AMD AM1 platform, although solutions made in the microATX format will also appear on the market. At least all major motherboard manufacturers, including ASRock, have already announced at least one such model.

But back to our ASRock AM1B-ITX board. As you can see, its layout is quite standard for such compact solutions: the processor socket is in the middle; interfaces are located on the left edge of the PCB, and slots for RAM are located on the opposite side; the bottom part is reserved for the PCI Express x16 slot. Recall that it uses only 4 PCIe 2.0 lanes. But even this amount in this case will be quite enough, since the AMD AM1 platform is primarily positioned as a basis for office computers, nettops or HTPCs, and not game configurations... Therefore, most likely, the PCI Express slot will be occupied by some kind of card that expands the multimedia capabilities of the system, for example, an external audio card or TV tuner.

Some restrictions are imposed on the RAM: its volume can reach 16 GB, and the speed is 1600 MHz. In addition, there is no support for dual channel mode. However, for the tasks outlined above, these restrictions are not so critical, and in practice they will not play a special role.

Since the AMD Kabini family processors have taken over many functions of third-party controllers, the number of additional microcircuits on the motherboard has significantly decreased. First of all, the lack of a chipset is striking. Now support for SATA 6 Gb / s ports is carried out directly by the processor, however, in the amount of only two pieces. ASRock felt that this might not be enough and used an additional ASMedia ASM1061 controller, which implements support for two more SATA 6Gb / s ports. The exact same picture is observed with USB 3.0: 2 connectors. USB port 3.0, located on the interface panel, operate under the control of the processor, and the work of 2 more, which can be connected to the block on the motherboard, is provided by the ASMedia ASM1042A controller.

VGA, DVI and HDMI are taken out from video interfaces to the rear panel. In the latter case, there is support for a resolution of 4096 x 2160 at a refresh rate of 24 Hz. Also present here: LAN connector, LPT port, three audio connectors, pair USB 2.0 and one PS / 2 Combo for connecting a keyboard or mouse. The audio path is based on the Realtek ALC662 chip, and network interface operates under the control of the Realtek RTL 8111GR gigabit chip.

By functionality AMD AM1 platform is practically not inferior to other popular solutions equipped with Socket FM2 / FM2 + / LGA 1150 processor sockets.

TDP of AMD Kabini processors is declared at the level of 25 W, therefore, rather low requirements are imposed on its power subsystem. The 2-phase VRM module, which we can see on the ASRock AM1B-ITX board, is quite enough. Its operation is provided by the Richtek RT8179B PWM controller, which includes two phase drivers, and also has a number of protective technologies (according to the specification - OCP / OVP / UVP / SCP).

Such a simple configuration of the processor converter allows you to reduce the cost of manufacturing the motherboard and, as a result, reduce the final cost of the entire computer.

The system is powered via a 24-pin ATX connector. Although, given the low power consumption of AMD Kabini processors, it is quite possible that we will see motherboard models powered by an external adapter (DC 19V).

The RAM subsystem received for configuration testing consists of one module AMD AE34G1609U1S, which belongs to the proprietary AMD Radeon Memory series. According to the marking and inscription on the sticker, it has a volume of 4 GB and can operate at a nominal frequency of 1600 MHz with delays of 9-9-9-28 and a voltage of 1.5 V. Since nettops and HTPCs are assembled in compact cases, where, as a rule, it is difficult to organize good cooling, then the presence of additional heatsinks on the memory chips will definitely not be superfluous.

We did not check the overclocking potential of the AMD AE34G1609U1S module, since the memory controller built into the processor will not allow it to operate at a frequency of more than 1600 MHz. However, you shouldn't worry too much about this, because an increase in the speed of the memory subsystem has practically no effect on the performance of most real applications. A slight increase is observed only in highly specialized programs that are unlikely to run on configurations based on the AMD AM1 platform.

AMD Athlon 5350 processor

Packaging, delivery set and standard cooling system

And now let's move on to the most interesting thing - the AMD Kabini processor, which in our case is represented by flagship model... It ended up in the test laboratory as part of the system, so we omit the description of the box, and immediately consider the standard cooling system.

It differs from the usual coolers that come with AMD Trinity / Richland / Kaveri / Zambezi / Vishera processors, primarily by its compact dimensions. The length and width of this cooling system is 55 mm (excluding clips), and the height is only 40 mm. And these are the dimensions already with the installed fan.

Note that for the first time in many years, AMD has changed the mounting system: instead of the usual latches, the cooler is attached to the board using two spring-loaded plastic clips. As a result, cooling systems with mounts for motherboards equipped with Socket AM3 / AM3 + / FM2 / FM2 + will no longer fit here.

The radiator has a familiar design - an aluminum core, from which four sections of thin aluminum fins extend. A low-profile FOXCONN PVA050E12L fan with a size of 50 mm and a power of 1.92 W is used to blow them off. Power is supplied through a 3-pin connector with support for monitoring the rotational speed of its blades.

Despite its compact size, the standard cooling system does its job well. In idle mode, the processor temperature was 36 ° C, and at maximum load (created by the stress test built into the AIDA64 utility) - 43 ° C. The maximum fan speed during the experiment reached 2950 rpm. All measurements were carried out on an open stand.

Appearance and technical specification

The AMD Athlon 5350 model is made in a micro-PGA package and looks very similar to other processors released under the AMD brand. On the heat distribution cover there is a marking and the name of the country of manufacture (in this case, Taiwan). The processor got there for the final assembly. The crystal itself was grown in Germany, as indicated by the inscription "Diffused in Germany".

Specification and technical characteristics:

In normal operation, the AMD Athlon 5350 speed is 2050 MHz with a reference frequency of 100 MHz and a multiplier of "x20.5". At the time of taking the readings, the voltage across the core was 1.288 V.

In idle mode, the multiplier is reduced to "x8", thereby dropping the frequency to 800 MHz. The voltage is 1.024 V.

The AMD Athlon 5350 cache is allocated as follows:

• cache memory of the first level L1 - 32 KB for data with 8 associativity channels and 32 KB for instructions with 2 associativity channels are allocated for each of 4 cores;
• L2 cache memory - 2 MB for all cores with 16 associativity channels;
• cache memory of the third level L3 - absent.

The DDR3 memory controller operates in single-channel mode and is guaranteed to support modules with frequencies up to 1600 MHz.

ASUS 8200 Ti200 - Detonator 23.11
ATI RADEON 8500 OEM - 4.13.9009.

By the end of testing, these drivers were the recommended versions (release) for both cards for Windows 98 / ME.

The system, of course, by today's standards is clearly not "Top", rather average or slightly higher, but, on the other hand, video cards are also not the most expensive and fast, so there is some balance here.

The tests were carried out only in 32-bit color (games in 16-bit color were not considered), VSync was disabled in all tests. Other settings were taken by default, i.e. after installing the driver, no changes were made (except for disabling VSync). The sound was muted in all tests.

In all tests, only resolutions of 1024x768, 1280x1024 are considered, since at these resolutions these video cards provide quite a comfortable level of playability in all modern games. In addition, in our opinion, most users in our conditions use 15 "-17" monitors, for which only 1024x768 and 1280x1024 provide a completely "humane" refresh rate of 85-100Hz. And for LCD monitors, which are gaining in popularity in recent years, these resolutions are also the main ones.

We carried out almost all tests both in normal mode and with anisotropic filtering. Why, in fact, with anisotropic filtering? The fact is that in modern games in normal mode (with a trilinear) at these resolutions, quite playable fps are often given by weaker and, accordingly, less expensive cards, such as GeForce2 Titanium, ATI Radeon 7500, GeForce4 MX 440. Accordingly, many actively playing owners of video cards of the GF3 and Radeon 8500 class believe that playing on such powerful cards (even on their lower modifications) without additional load, which is anisotropic filtering, is like taking a taxi to a bakery, and as you know, "our people don't take a taxi to the bakery ":).

The question is actually only what type of load to choose. We deliberately excluded other types of load, because (again, in our opinion), not all today's games can be played simply with antialiasing in 4x mode (not to mention combining this mode with anisotropic filtering). The choice is actually between AA 2x and AF, of which we left the last mode (as giving, subjectively, a more pleasant picture in 3D in most cases). Of course, giving up something is not the best solution, but in the conditions of limited time for testing, you have to separate the main from the secondary.

For tests, the 8500 LE selected the mode with the maximum degree of anisotropic filtering, for the Ti200 anisotropic filtering level = 4 as a compromise between quality and performance.

To assess the effect of overclocking video cards, we tested the Radeon 8500 LE both in normal mode (250/250) and in overclocked mode (286/266). The Ti200 card, in turn, was also tested in the stock (175/200) and overclocked (200/230) states. Of course, overclocking is roulette, someone's video card will overclock and will work stably (and not only pass five tests) for more high frequencies especially if you use additional cooling or start soldering resistors. We stopped at these numbers, since, in our opinion, such overclocking is practically guaranteed for these video cards and without additional measures.

Due to time constraints and a large volume of tests, it was decided to limit ourselves (at this stage) to tests only under the Win98SE operating system. In the foreseeable future, we plan to prepare the 2nd part of this article with tests in Windows XP.