Power supply system, cooling system. Number of universal processors

About a year has passed since the announcement of older video cards based on the Cypress graphics core. And as if according to the schedule of Japanese high-speed trains, the successor to successful architecture is presented to the world. Over the course of a year, AMD gathered its strength and made a small revolution, changing the very heart of its GPUs: five former scalar processors were optimized into a dual pair. As well as the film of the same name, Nvidia’s countermeasures, the successful GF104 and the ultra-fast castling in the upper segment - GF100 -> GF110 were taken into account.

After the launch of the Barts series, there is not much space left in the ranks of cards, and you have to carefully plan placement among both competitive solutions and your own graphics adapters. As a result, from now on we can consider that the Middle…High segment of AMD video cards has been completely updated. HD 5830/50/70, which left the stage, was replaced by HD 6850/70 and 6950/70.

Characteristics table

Characteristics	HD 5870	HD 6870	HD 6950	HD 6970	GTX 470	GTX 480	GTX 570	GTX 580
Codename	Cypress XT	Barts XP	Cayman Pro	Cayman XT	GF100	GF100	GF110	GF110
Technical process, nm	40	40	40	40	40	40	40	40
Size of core/cores, mm 2	334	255	389	389	~500	~500	~530	~530
Quantity transistors, million pieces	215,4	180	264	264	320	320	330	330
2D core frequency, MHz	157	100	150	150	50 / 100	50 / 100	50 / 100	50 / 100
3D core frequency, MHz	850	900	800	880	607 / 1215	701 / 1402	732 / 1464	772 / 1544
OC core frequency, MHz	1050	950	950	950	800 / 1600	875 / 1750	875 / 1750	850 / 1700
Voltage at core 2D, B	0,95	0,95	0,90	0,90	0,88	0,96	0,91	0,96
Voltage at core 3D, B	1,15	1,21	1,28	1,28	0,99	1,01	1,01	1,06
Number of shaders PC. (PS)	1600	1120	1408	1536	448	480	480	512
Number of blocks rasterization, PC. (ROP)	32	32	32	32	40	48	40	48
Number of textures blocks, pcs. (TMU)	80	56	88	96	56	60	60	64
Maximum fill speed, Gpix/sec	27,2	28,8	25,6	28,2	24,3	33,6	29,3	37,1
Maximum speed texture samples, Gtex/sec	68	50,4	70,4	84,5	32,4	42,1	43,9	49,4
Version pixel/ vertex shaders	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0	5.0 / 5.0
Memory type	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Efficient frequency 2D memory, MHz	1200	300	300	300	67	67	67	67
Efficient frequency 3D memory, MHz	4800	4200	5000	5500	3360	3700	3900	4008
Efficient frequency memory OC, MHz	5200	4800	5800	5800	3600	3900	4000	4100
Voltage at memory 2D, B	1,60	1,61	1,60	1,60	1,54	1,58	1,34	1,36
Voltage at 3D memory, B	1,60	1,63	1,60	1,60	1,55	1,58	1,56	1,62
Memory size, MB	1024 / 2048	1024	2048	2048	1280	1536	1280	1536
memory bus, bit	256	256	256	256	320	384	320	384
Bandwidth memory, GB/sec	153,6	134,4	160	176	133,9	177,4	152	192,4
Consumed 2D power, Watt	27	19	20	20	nd	nd	nd	nd
Consumed 3D power, Watt	188	151	250	250	215	250	219	244
Crossfire/Sli	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Card size LxWxH, mm	282x100x38	248x100x37	270x100x37	270x100x37	270x100x38	270x100x38	270x100x38	270x100x38
Recommended price, $	399	239	299	369	249	499	349	499

Judging by preliminary data, the price war should flare up with renewed vigor. An unprecedented number of graphics cards are located in a narrow price range. As a result, one of the producers (or maybe both) will suffer losses, and the buyer will ultimately benefit.

Architecture and improvements

Due to limited time frames, preference was given to graphical tests, and a detailed description of the architecture, comparison with existing analogues and will be competitors in the second part. Now I offer you a quick review of significant and original modifications.

As you may have guessed, the GPU master plan remains the same. Superscalar processors are grouped and split into two halves.

In total, the core has 24 SIMD blocks for the older card and 22 for the younger card.

An important innovation was further optimization of SIMD loading. “Cayman” received two full-fledged “conductors”. And in addition to quantitative improvements, the internals of the Graphic Engine have undergone qualitative improvements: the tessellation module has grown to the 8th generation (according to the AMD developers themselves).

Refusal from one processor (now there are 4 instead of 5) made it possible to save about 10% of the area of ​​​​a similar graphics core and reduce the complexity of management.

ROP has been updated compared to Cypress. The difference in speed reaches twofold, fourfold superiority.

The memory controller also did not go unnoticed by the developers. Its main functions have been revised to further increase the speed of work.

But the company did not stop at the modifications described above, and continued to break the stereotype of the impossibility of improving what is already good. From now on, AMD flaunts a long list of various anti-aliasing modes, not giving its competitor a moment's peace.

In addition to the new Morphological filter, Enhanced Quality Anti-Aliasing has appeared, which in Russian means smoothing with improved quality. Represents the usual MSAA mode, but with twice the big amount overlap values.

We remembered, in addition to caring about the quality of the output image, and about energy consumption. Russian prices electricity still allows us not to worry about our bills, but in America and Europe the situation is different. Nevertheless, a certain kind of progress towards the appropriate use of energy is definitely encouraging.

According to AMD internal research, existing algorithms Power management is not ideal. New graphics accelerators offer us a more flexible system. According to representatives, there are more than a hundred sensors inside the GPU that monitor workload different parts graphics core and deactivate the GPU area at the right time, which contributes to lower electricity consumption. The moment of turning on this technology remains unclear, but they promised to solve the issue of turning it off, which excites the minds of enthusiasts and overclockers.

An example demonstrates the effectiveness of the newfangled option. During testing, the GPU frequency smoothly varies sufficiently wide range, but the final result remains within the promised performance.

Already now, the user has a tool in the drivers that controls the power consumption of the video card. It is enough to move the slider to the +10% position to indirectly affect the result. The multi-stage adjustment scheme used is more of a complex module for adaptive adjustment of the speed/watt ratio and officially legitimizes the overclocking of the card.

The photos of the Cayman family of graphics adapters that have appeared have aroused increased interest in the small switch at the end. The essence of its existence is simple: from now on, all cards manufactured according to the reference design are equipped with two BIOSes, the main one and the backup one. The first can be reprogrammed, while the second serves as a backup and is write-protected.

Appearance

Both cards are based on the same PCB design, like twins straight off the cover of a glossy magazine. At first glance, it is difficult to unambiguously determine which of them bears the HD 6950 index and which HD 6970.

The tradition of using the black rectangle design continues, but in different sizes.

As expected by rank, from small to large, the size of video cards also grows. The closest HD 6850 is a couple of centimeters smaller than the HD 6870, which in turn is shorter than both Caymans.

The best way to check how successful it was in pitting the GF110 against the AMD HD 6950/70 is in gaming applications, which is what I propose to do.

Test bench

Test bench configuration:

• Motherboard: MSI Big Bang-XPower (BIOS 1.4B7);
• CPU: Intel Core i7 920 @4305 MHz (205 x 21, 1.31 V);
• Cooling system: Water cooling system;
• Thermal interface: Arctic Cooling MX-2;
• RAM: Kingston HyperX KHX2000C9AD3T1K3/6GX 6 x 2 GB DDR3 1640 MHz (7-7-7-19-1T, 1.65 V);
• Hard drive: Seagate 7200.11 320 GB;
• Power supply: Tagan TG1100-U95 1100 W;
• Audio card: ASUS Xonar HDAV 1.3;
• Operating system: Microsoft Windows 7 x64;
• Driver version for AMD/ATi Catalyst 10.11, exception HD 6950/70 - special version Catalyst 10.11, Nvidia - nforce 261.00, exception GTX 570, nforce 263.09.

List of measuring and control instruments and instruments used:

• Sound level meter: Center 320;
• Multimeter: Fluke 289;
• Electricity tarifficator: E305EMG;
• Microphone: Philips SBC ME570.

Testing tools and methodology

To correctly measure temperature and noise, the following conditions were used:

A closed room with an area of ​​4 m 2, inside of which there is a system for automatic support of climatic conditions. In this case, the temperature level was set at 26°C +/- 1°C. The accuracy of compliance with the specified parameters was monitored by four sensors, one of which was located 5 cm from the fan of the video card cooling system and was the leading one. It was used for the main correction of the room temperature.

The noise was measured at a distance of 50 cm to the video card. The background level was 22 dBA. As hard drive an SSD was used, and the power supply, pump, radiator with fans were located outside the room during measurements. There were no other components at the stand that made any noise.

Sound recording of the cooling system was made at a distance of 10 cm from the fan. The first 10-20 seconds without load in idle mode, then 100% load was turned on using the Furmark program. The highest noise level is reached almost at the end of the audio recording. The temperature conditions and noise were determined in advance so that during the recording of the audio track you could hear exactly the maximum noise.

The level of electricity consumption [watts] during idle time was assessed according to the readings of the E305EMG tarifficator immediately after loading the operating system. The values ​​displayed in the graph correspond to the minimum achieved figures from the device. The video cards were tested under load with Furmark version 1.8.0. After 10-15 minutes, the temperature and fan speed reached their theoretical maximum. After which the data was entered into a table.

For a visual comparison of video cards, all games used as test applications were launched at resolutions of 1680x1050 and 1920x1200.

In some games, where possible, built-in performance measurement tools were used:

• 3DMark Vantage – Extreme, High, Performance
• Aliens vs Predator DX 11 Benchmark v1.03
• Colin McRae DIRT 2
• Far Cry 2
• Unigine Heaven Benchmark v 2.0
• Call of Duty World at War
• Mafia II
• S.T.A.L.K.E.R.: Call of Pripyat (total values ​​– averaged over four modes)

For the following games, performance was measured using the FRAPS v3.2.5 utility:

• Battlefield Bad Company 2
• Metro 2033.

VSync was disabled during tests.

To avoid errors in measurement error, all tests were performed three times. When calculating the average FPS, the arithmetic average of the results of all runs was taken as the final result.

Temperature conditions, noise levels and electricity consumption

Operating temperatures

Noise level

Please enable JavaScript to see graphs

Electricity consumption

Please enable JavaScript to see graphs

Listen to the cooling system:

http://trash..wma

http://trash..wma

And compare:

http://trash..wma

http://trash..wma

http://trash..wma

http://trash..wma

http://trash..wma

http://trash..wma

Overclocking to standard cooling

Introduction

A little more than a month has passed since Advanced Micro Devices introduced mid-range video cards price segment Radeon HD 6800, based on Barts GPUs, and today - December 15, 2010 - the company releases two of its most powerful single-processor solutions on the Cayman GPU.

So far, the Caymans have only two single-chip video cards: Radeon HD 6950 and HD 6970. At the same time, we do not rule out the possibility of a third graphics card appearing, with more modest characteristics and at a more attractive price, as happened not so long ago with Radeon HD 5830. In addition, in the first quarter of 2011, a dual-processor flagship based on the Cayman is promised to be presented, which will replace the Radeon HD 5970.

As for the heroes of today's material - the Radeon HD 6950 and HD 6970, their main task, according to AMD, is to fight against the flagships of NVIDIA - the GeForce GTX 570 and GTX 580 video cards. What is most interesting is that this fight will have to be imposed not so much as a pure performance, price factor and technological and operational advantages of new video cards over competitors. An interesting approach, I must admit. Well, let's try to get acquainted with the new graphics processor and video cards based on it.

Architectural Features of the New Cayman GPUs

Despite the fact that the new Cayman GPUs belong to the HD 6xxx series, they are architecturally different from the HD 68xx solutions and, of course, from the HD 58xx. These differences are both quantitative and qualitative, representing an important step in the development of the new AMD graphics architecture.

The most noticeable, and, in our opinion, significant change was the transition from VLIW 5-way processors, which debuted in the R600, to VLIW 4-way:

The previous VLIW 5-way processor contained, as the name suggests, 5 computing units, four of which were capable of performing only the simplest operations, such as addition and multiplication, and the fifth, in addition to these functions, also performed the so-called transcendental, complex operations. These include such as sine, cosine, extraction square root and so on. In the new VLIW 4-way processor, this block for complex operations has been eliminated, and its functions are now assigned to a pair of “simple” or “universal” blocks. Thus, each VLIW 4-way processor is still capable of performing four simple or one complex operation, which is required in the vast majority of cases. A conditional loss in performance to the old VLIW 5-way processor is possible only in rare, and, so to speak, “rigged” cases, when it is necessary to simultaneously perform one complex and more than two simple operations. In real applications, such cases are very rare, so you can ignore them, but the freed up space and increased efficiency of blocks lead to an overall increase in the performance of new GPUs at a lower cost.

In addition, the new GPUs have increased the speed of working with double precision operations and optimized the exchange mechanism with system memory:

Other equally important improvements to the Cayman include an increase in the performance of raster operations units (ROPs) in a number of modes, as well as the traditional improvement of the tessellation unit, which, according to AMD, has become noticeably faster than its predecessor:

The remaining changes to the Cayman's GPUs are quantitative in nature. Thus, the number of texture sampling and filtering units was increased from 80 to 96, the number of VLIW processors increased, compared to Cypress, from (1600/5) 320 pieces to (1536/4) 384 pieces, that is, by 20%.

In addition to the morphological anti-aliasing previously introduced in the Radeon HD 68xx series, the Radeon HD 69xx has a new anti-aliasing algorithm - the so-called “Enhanced Quality AA”, which is an addition to MSAA and is activated in the Catalyst (EQ) driver control panel:

Schematically, its work looks like this:

The bottom line is that to two, four or eight samples, the same number of interpolated points are added, requiring much less computing power of the GPU. Thanks to this, the picture is of higher quality, and the performance reduction compared to the conventional anti-aliasing algorithm, according to AMD, should not exceed 10%:

At the same time, Cayman GPUs have introduced new energy-saving technology, which can be controlled from the Catalyst driver control panel from the Overdrive tab:

A new “slider” has appeared here, called “Power Control settings”, with the ability to adjust from minus 20 to plus 20%:

Using this setting, you can change the peak power limit of the video card, and thereby either save energy, or, more simply, increase the likelihood of successful overclocking of the GPU. The new Cayman GPUs now feature multiple sensors that constantly monitor GPU health and load levels, allowing Caymans to operate more energy-efficiently. In addition, these sensors in the GPU, when they detect programs like FurMark and the like, will activate the protection mode, putting the card into a kind of throttling mode, periodically reducing the frequency of the chip. Here's what it looks like using the Perlin Noise test from the 3DMark Vantage synthetic package:

It can be clearly seen that the GPU frequency changes depending on the load during the test, but the frame rate per second remains unchanged, which we personally doubt. No, not in the results of this particular test, but in the fundamentally the same operation of the energy saving mechanism and in other very resource-intensive applications. Most likely, there will be a drop in performance. True, AMD has counted no more such applications than there are fingers on one hand. Well, time will tell.

To push back the response limit of this “officially approved throttling”, in drivers it is enough to move the slider to the maximum +20%. Using the same “Perlin Noise” test as an example, the impact of this setting on performance is as follows:

Or you can do the opposite - lower the maximum energy capacity of the video card by setting this setting to minus values. In this case, there will be a drop in performance. However, again, according to AMD, it will not be critical:

The power control limits are programmed in Video card BIOS, and we hope that over time, tools will appear to expand them (primarily in the direction of increase, of course). In this case, when organizing efficient cooling graphics processor, it will be possible to avoid limiting the overclocking potential of the video card by the upper limit of throttling.

At the end of the brief theoretical part, without comments, we present a couple of performance comparison slides AMD Radeon HD 6950 with NVIDIA GeForce GTX 470, and AMD Radeon HD 6970 with NVIDIA GeForce GTX 480:

AMD specifically notes that, as in the case of Radeon HD 68xx, on Radeon HD 69xx series video cards, the efficiency of CrossFireX technology has been increased:

Whether this is true or not, we will check in one of the following articles, but for now let’s move on to the specifications table and a direct study of the reference AMD Radeon HD 6950 and HD 6970 video cards.

Technical characteristics of AMD Radeon HD 6950 and HD 6970 video cards and their competitors

In the table of characteristics, video cards are arranged from left to right in descending order of their recommended cost:

AMD Radeon HD 6950 and HD 6970 2 GB video cards

We were provided with two reference video cards AMD Radeon HD 6950 and HD 6970 for testing (in fact, non-reference video cards do not yet exist at all, and it is not a fact that they will appear). Externally, video cards differ from each other only by a sticker on the plastic casing of the cooling system with the video card model and power connectors. The front sides of the video cards are covered with a cooling system, and the back sides are covered with a decorative aluminum cover:

“Decorative”, since it does not come into contact with any elements of the printed circuit board, and the protective function in this case is very doubtful. The length of the video cards is 276 mm and the thickness is 39 mm. The air heated by the video card is partially exhausted through the grille in the upper part of the casing, thus remaining in the body of the system unit. But, nevertheless, the bulk of its mass is removed from the housing through the grille on the panel with outputs:

By the way, there are now as many as five outputs: two DVI, two miniDP and one HDMI:

In addition to this “gentlemanly set” of outputs, we can say that now there will be no special Eyefinity Radeon HD 69xx video cards, and six monitors can be connected to any of these video cards using a special video hub.

Both models of video cards are equipped with two connectors for combining them in CrossFireX modes up to four video cards:

Next to them is a small switch for selecting the BIOS from which the video card can boot. Yes, yes, everything is fine with your eyes - now every reference Radeon HD 6950 and HD 6970 comes standard with two BIOS chips, one of which is reprogrammable, and the second is a backup, just in case of overly optimistic frequencies or voltages hardwired to the first BIOS chip:

Now let's look at the differences between the reference AMD Radeon HD 6950 (pictured here and below - on the left) and AMD Radeon HD 6970 (here and below - on the right). The first is the already mentioned additional power connectors. If the Radeon HD 6950 has two six-pin connectors, then the HD 6970 already required six- and eight-pin connectors:

Not surprising, since the peak power consumption of the older video card is stated at 250 W, while the Radeon HD 6950 is 200 W. Interestingly, when announcing the Caymans, AMD for the first time introduces the concept of “typical gaming power consumption”, which for the Radeon HD 6970 is at 190 W, and for the Radeon HD 6950 - 140 W. The power consumption of the new video cards in idle mode is the same and should be only 20 W.

Printed circuit boards of video cards, with the exception of three or four elements, are practically indistinguishable...

...as well as their power units:

With more detailed consideration There are visible differences in the markings of the chips used that are responsible for powering the video cards. Let's leave them to the judgment of experts.

The die area of ​​the Cayman graphics processors of the Radeon HD 6950 and HD 6970 video cards has increased significantly compared to Barts and is 389 mm². However, they still do not have a heat spreader. Note that in our case the GPU Radeon video cards HD 6970 was released 8(!) weeks later than the Radeon HD 6950 graphics processor:

The AMD Radeon HD 6950 GPU has 1,408 unified shader processors and 88 texture units (ROPs), while the Radeon HD 6970 graphics card has 1,536 unified shader processors and 96 texture units. The number of raster operations blocks for both video cards is the same and equal to 32. The nominal frequency of the Cayman on the HD 6950 is 800 MHz, and on the HD 6970 - 880 MHz.

In power saving mode (read: inactivity), GPU frequencies are reduced to 250 MHz, and voltages are reduced from 1.1 V to 0.9 V on the HD 6950, and from 1.175 V to 0.9 V on the HD 6970. Other characteristics of the GPUs have already been discussed above in the theoretical part and the table of characteristics. Let's add that on both video cards the diagonal distance between the mounting holes for the GPU radiator is 75 mm, the same as on the HD 5850 or HD 5870.

Both video cards have the same memory capacity of 2048 MB of GDDR5 standard, equipped with eight FBGA chips manufactured by Hynix Semiconductor Inc. The Radeon HD 6950 has chips marked T2C ">H5GQ2H24MFR T2C with a nominal voltage of 1.5 V and a theoretical effective frequency of 5000 MHz, while on the Radeon HD 6970 the memory is labeled H5GQ2H24MFR R0C, voltage 1.5 V and theoretical effective frequency 6000 MHz:

As you can already see from the characteristics table, the Radeon HD 6950 has memory operating at an effective frequency of 5000 MHz, and the Radeon HD 6970 - at 5500 MHz, which leaves decent operational space for overclocking enthusiasts. In idle mode, the effective video memory frequency of both video cards is reduced to 600 MHz. The width of the memory exchange bus is 256 bits. All memory chips are located on the front side of the printed circuit boards.

Thus, the specifications of the new AMD reference graphics cards are as follows:

Both video cards are equipped with a new dual-slot cooling system, consisting of a massive radiator with an evaporation chamber and thin aluminum fins, a metal base, a centrifugal fan and a plastic casing covering the whole thing:

By the way, this casing can now be removed without removing the cooling system itself from the video card, simply by unfastening the latches on the sides.

The vaporization chamber (according to AMD, already fifth generation) is located under the entire base area of ​​the radiator:

The base of the camera does not come into contact with the video memory chips; a metal plate at the base of the cooling system is responsible for cooling them. Contact is made using thermal pads:

In turn, the GPU is in contact with the vapor chamber through a thick gray thermal interface.

The cooling system for AMD Radeon HD 6950 and HD 6970 video cards is equipped with an 80-mm centrifugal fan manufactured by Foxconn (model PVB070G12N), which is non-trivial for video cards peak power 24 W:

The rotation speed of the centrifugal fan is automatically adjusted using the pulse-width modulation method in the range from 1200 to 5800 rpm. Let us recall that for video cards of the Radeon HD 58xx and HD 68xx series, the maximum rotation speed of the centrifugal fan did not exceed 5000 rpm (even more often - no more than 4500 rpm), so we can assume that the new Radeons will not be quieter than their predecessors. However, this is not true for both new video cards. Right now we will check the temperature conditions of the new products.

Temperature tests of all video cards in today's article were carried out in a closed system case at a room temperature of 25 ° C. You will find detailed configuration of the system unit in the next section of the article. The Aliens vs. test was used as a typical gaming load. Predator (2010), running for five to six cycles at 1920x1080 resolution with 16x anisotropic filtering. As ultimate load used FurMark test version 1.8.2, launched by a renamed exe file with the “Xtreme Burning Mode” option in a resolution of 2560x1600 with 16x anisotropic filtering activated in the Catalyst and GeForce/ION drivers. Monitoring was carried out using programs GPU-Z versions 0.4.9 and MSI Afterburner version 2.1.0 beta 5.

First, let's look at the temperature conditions of video cards under a typical gaming load in the automatic mode of operation of centrifugal fans:

AMD Radeon HD 6950AMD Radeon HD 6970
(auto mode)(auto mode)

As for the Radeon HD 6950 video card, its operating temperature and noise level are within acceptable limits, since the graphics processor warmed up to only 81 degrees Celsius, and the centrifugal fan of the cooling system spun up only to 2260 rpm. It became audible, but the noise level of the video card still could not be called uncomfortable. As the number of gaming test cycles increased, the temperature did not increase. Note that Aliens vs. Predator (2010) is one of the most difficult games in terms of warming up the video card.

Before analyzing the temperature conditions of the Radeon HD 6970 video card, you need to say the following. The fact is that the copy of the AMD Radeon HD 6970 provided to us for testing initially turned out to be not entirely functional. The video card could operate stably only in 2D mode, and when starting any 3D application, pandemonium began with the rotation speed of the centrifugal fan and temperature jumps. Assuming that the same notorious throttling was triggered, we changed the power limit adjustment to the maximum +20%, but this did not lead to the correction of such an unpleasant situation. Therefore, assuming that the graphics processor of the video card was simply overheating, the Radeon HD 6970 was rebuilt, replacing the standard dried-out thermal interface with Arctic Cooling MX4. Only after this did the video card work stably in 3D applications. However, its temperature conditions and noise level left much to be desired. Thus, the GPU temperatures reached 104 degrees Celsius, and the centrifugal fan speed increased to 4760 rpm! Naturally, in this case there is no need to talk about any comfort. It's good that Aliens vs. Predator (2010) was the only test on our list that warmed up the video card so much. However, we are still inclined to believe that not everything is entirely in order with the copy of the AMD Radeon HD 6970 provided to us for testing.

Despite this, it was decided not to abandon testing video cards in FurMark, and here are the results obtained in the automatic mode of operation of centrifugal fans...

AMD Radeon HD 6950AMD Radeon HD 6970
(auto mode)(auto mode)

...and at their maximum speed:

AMD Radeon HD 6950AMD Radeon HD 6970
(maximum speed)(maximum speed)

It is again difficult to make any complaints about the Radeon HD 6950 - here we see 90 degrees Celsius and only 2800 rpm in automatic mode, as well as 71 degrees Celsius at a maximum of 5770 rpm. In turn, the Radeon HD 6970 failed the FurMark test even at maximum centrifugal fan speed, reducing the GPU frequency to 500 MHz.

Considering such terrible results of checking the temperature conditions of the Radeon HD 6970 and the impossibility of organizing alternative cooling (primarily due to the lack of radiators on the VRM), only the AMD Radeon HD 6950 video card was tested for overclocking. And here are the results obtained:

The frequency potential of the GPU turned out to be very modest, since the frequency was increased by only 70 MHz (+8.8%), and the video memory did not disappoint, accelerating to an impressive 5880 MHz effective frequency. It should be noted that after overclocking the video card, its temperature did not change. That is, it hasn't changed at all. Only the centrifugal fan of the video card cooling system increased the speed by 100-150 rpm, depending on the type of load.

At the end of the subsection, we provide links to the main BIOS of the video cards reviewed: AMD Radeon HD 6950 2 GB And AMD Radeon HD 6970 2 GB.

Test configuration, tools and testing methodology

All performance tests of video cards were carried out in a closed system case with the following configuration:

Motherboard: Gigabyte GA-X58A-UD9 (Intel X58 Express, LGA 1366, BIOS F5g);
CPU: Intel Core i7 Extreme Edition i7-980X 3.33 GHz(Gulftown, B1, 1.225 V, 6x256 KB L2, 12 MB L3);
CPU cooling system: ThermoLab baram2010 (1x120 mm, 1100 rpm);
Thermal interface: Arctic Cooling MX-2;
RAM: DDR3 3x2 GB OCZ Platinum Low-Voltage Triple Channel (1600 MHz / 7-7-7-24 / 1.65 V);
System disk: RAID-0 2xSSD Kingston V-series SNV425S2128GB(SATA-II, 2x128 GB, MLC, Toshiba TC58NCF618G3T);
Disk for programs and games: Western Digital VelociRaptor (SATA-II, 300 GB, 10000 rpm, 16 MB, NCQ) in a Scythe Quiet Drive 3.5" box;
Archive disk: Samsung Ecogreen F4 HD204UI (SATA-II, 2 TB, 5400 rpm, 32 MB, NCQ);
Case: Antec Twelve Hundred (front wall - three Noiseblocker NB-Multiframe S-Series MF12-S2 at 900 rpm; back - two Noiseblocker NB-BlackSilentPRO PL-1 at 900 rpm; top - standard 200 mm fan at 400 rpm);
Control and monitoring panel: Zalman ZM-MFC2;
Power unit: Xigmatek "No Rules Power" NRP-HC1501(1500 W), 140 mm fan;
Monitor: 30" Samsung 305T Plus.

32nm six-core processor with a multiplier of 24 and activated function"Load-Line Calibration" was overclocked to 4.42 GHz by increasing the voltage in the motherboard BIOS to 1.425 V:

Moreover, 6 GB random access memory DDR-3 operated at a frequency of 1.47 GHz with timings 7-7-7-16-1T at a voltage of 1.64 V. Technologies " Turbo Boost" and "Hyper-Threading" were disabled during testing.

For comparison with the AMD Radeon HD 6950 and HD 6970, video cards were included in testing AMD Radeon HD 6870 1 GB And Inno3D GeForce GTX 570 1.28 GB:

Unfortunately, we were not able to include the NVIDIA GeForce GTX 580 in testing due to the lack of such a video card, as well as due to the tight time frame for testing and preparing the article (4 days).

Testing, which began on December 10, 2010, was conducted under the Microsoft Windows 7 Ultimate x64 operating system with all critical updates to specified date and with the following drivers:

motherboard chipset Intel Chipset Drivers - 9.2.0.1015 WHQL;
DirectX End-User Runtimes libraries, release date: November 30, 2010;
video card drivers for AMD GPUs - Catalyst 10.12 RC2 (12/07/2010);
Video card drivers for NVIDIA GPUs - GeForce/ION 263.09 WHQL (12/07/2010) including PhysX drivers version 9.10.0514.

Testing of video cards in games was carried out in two resolutions that are relevant for video cards tested today: 1920x1080 and 2560x1600 pixels. For the tests, two graphics quality modes were used: “High Quality + AF16x” - maximum texture quality in the drivers with 16x level anisotropic filtering enabled, and “High Quality + AF16x + AA 4(8)x” with 16x level anisotropic filtering enabled and full screen anti-aliasing (MSAA) of 4x or 8x, if the average number of frames per second remained high enough for comfortable gameplay. Anisotropic filtering and full-screen anti-aliasing were enabled directly in the game settings. If these settings were missing in games, then the parameters were changed in the control panels of the Catalyst and GeForce/ION drivers. Vertical sync is forcibly disabled in driver control panels.

Once again, the list of test applications and games has been updated. In connection with the emergence of the new 3DMark 2011, the old “parrot meter” 3DMark 2006 was excluded from the list of applications used for testing. In addition, the game Colin McRae: DiRT 2 was excluded from the list, which was replaced by the new and, one might say, related to it F1 2010, and NBA 2K11 was also temporarily removed, as its results often defied any logical explanation. All the latest patches for games have not been forgotten either. As a result, the test list consisted of two semi-synthetic packages, one techno demo and 17 games. This is what it looks like with brief description methods (hereinafter the games are arranged in the order of their release):

3DMark Vantage(DirectX 10) - version 1.0.2.1, “Performance” and “Extreme” settings profiles (only basic tests were tested);
3DMark 2011(DirectX 11) - version 1.0.0.0, “Performance” and “Extreme” settings profiles;
Unigine Heaven Demo(DirectX 11) - version 2.1, maximum quality settings, tessellation at the “normal” level;
Crysis(DirectX 10) - version 1.2.1, “Very High” settings profile, double cycle of the “Assault Harbor” demo recording from Crysis Benchmark Tool version 1.0.0.5;
Far Cry 2(DirectX 10) - version 1.03, “Ultra High” settings profile, double test cycle “Ranch Small” from Far Cry 2 Benchmark Tool v1.0.0.1;
BattleForge: Lost Souls(DirectX 11) - version 1.2 (11/18/2010), maximum graphics quality settings, shadows enabled, SSAO technology enabled, double run of the test built into the game;
Resident Evil 5(DirectX 10.1) - version 1.2, testing the variable test with maximum graphics settings without motion blur, the result was taken as the average value of the third scene of the test, as the most resource-intensive;
(DirectX 11) - version 1.6.02, settings profile “Improved dynamic lighting DX11” with additional manual setting of all parameters to the maximum, tested our own demo recording “cop03” at the “Backwater” level;
Borderlands(DirectX 9) - game version 1.2.1, testing “timedemo1_p” with maximum quality settings;
Grand Theft Auto IV - Episodes From Liberty City(DirectX 9) - version 1.1.2.0, test from the part “The Ballad of Gay Tony”, settings “Very High”, “View Distance” = 23%;
Left 4 Dead 2: The Sacrifice(DirectX 9) - game version 2.0.5.3, maximum quality, tested our own demo recording “d53” (two cycles) on the “1. Docks”, stage “Victim”;
Metro 2033: The Last Refuge(DirectX 10/11) - version 1.2, official test was used, quality settings “High”, tessellation, DOF and MSAA4x disabled, AAA anti-aliasing was used, double sequential pass of the “Frontline” scene;
Just Cause 2(DirectX 11) - version 1.0.0.2, maximum quality settings, “Background Blur” and GPU Water Simulation techniques activated, double sequential pass of the “Dark Tower” demo;
Aliens vs. Predator (2010)(DirectX 11) - “Texture Quality” in “Very High” mode, “Shadow Quality” in “High” mode, SSAO technology enabled, two test cycles in each resolution;
Lost Planet 2(DirectX 11) - game version 1.0, maximum graphics quality settings, motion blur enabled, performance test “A” was used (average of all three scenes);
StarCraft 2: Wings of Liberty(DirectX 11) - game version 1.0, all graphics settings to “Ultra” level, “Ultra” physics, reflections included, double two-minute test of our own demo “jt1”;
Mafia 2(DirectX 11) - game version 1.0.0.3, maximum graphics quality settings, double run of the test built into the game;
Sid Meier's Civilization V(DirectX 11) - game version 1.0, maximum graphics quality settings, double run of the “diplomatic” test of the five most difficult scenes;
F1 2010(DirectX 11) - game version 1.01, built-in Ultra-quality test, consisting of one lap along the “Silverstone” track;
Tom Clancy's H.A.W.X. 2(DirectX 11) - version 1.04, maximum graphics quality settings, shadows activated, tessellation enabled, double run of the test scene.

A more detailed description of testing methods for video cards and graphics settings in some of the listed games can be found in the specially created thread of our conference, as well as participate in the discussion and improvement of these techniques.

If games implemented the ability to record a minimum number of frames per second, then this was also reflected in the diagrams. Each test was carried out twice; the best of the two values ​​obtained was taken as the final result, but only if the difference between them did not exceed 1%. If the deviations of the test runs exceeded 1%, then the testing was repeated at least once more to obtain the correct result.

Video card performance test results and their analysis

In the diagrams, the test results for the AMD Radeon HD 6870 1 GB video cards are highlighted in purple, the AMD Radeon HD 6950 2 GB video cards in nominal and overclocked modes, as well as the Radeon HD 6970 2 GB are marked in red, and the GeForce GTX 570 in green (the frequencies of the Inno3D card are given to the GTX 570 reference frequencies). The video cards in the diagrams are arranged in descending order of their recommended cost.

3DMark Vantage

The new Radeon HD 6970 and HD 6950 look good in this test, but only compared to the Radeon HD 6870, which they are up to 40% ahead of. The new products fail to compete with the GeForce GTX 570; the “green” video card is ahead in both quality modes. The Radeon HD 6970 is 11-15% faster than the Radeon HD 6950 in this synthetic test, but the latter, when overclocked, reduces the lag to 3-5%.

3DMark 2011

The new products look better in the new 3DMark 2011 test, where the Radeon HD 6970 successfully competes with the GeForce GTX 570, and the HD 6950 is not far behind them. However, many probably expected a more convincing victory from AMD’s single-processor flagship, because AMD is positioning it as a competitor to the GeForce GTX 580, but at a more attractive price. As for the advantage of the Radeon HD 6970 and HD 6950 over the Radeon HD 6870, it reaches 28%. The difference between the two new video cards does not exceed 12% and is significantly compensated by overclocking the Radeon HD 6950.

Unigine Heaven Demo

Unigine Heaven - a test that actively uses tessellation, did not reveal a clear leader, since in the mode without anti-aliasing and a resolution of 1280x1024 the GeForce GTX 570 video card is in the lead, and in a more difficult graphics mode the Radeon HD 6970 is ahead. The difference between the Radeon HD 6970 and HD 6950 is 11% in favor of the older video card, but when overclocking the younger one it is reduced to 2-3%. The advantage of the Radeon HD 6970 over the Radeon HD 6870 in this test is the most impressive among all other tests and is 59% in light graphics mode and 66% in anti-aliasing mode. Very good.

Crysis

Crysis still remains one of the most resource-intensive games, in which even top-end AMD and NVIDIA video cards can provide gamers with a comfortable gaming mode only without the use of anti-aliasing and in resolutions no higher than 1920x1080. However, if we compare the video cards in today’s article, it should be noted that the GeForce GTX 570 in modes without anti-aliasing is on par with the Radeon HD 6970, and when anti-aliasing is activated, it is at the level of the Radeon HD 6950. The latter video card does not lag behind its older one when overclocked. sisters." The advantage of the Radeon HD 6970 over the Radeon HD 6870 ranges from 24 to 56% depending on the resolution and quality mode.

Far Cry 2

Frankly speaking, in the game Far Cry 2, the new Radeon HD 69xx looks unconvincing against the background of the GeForce GTX 570. A video card based on an NVIDIA GPU is at least not inferior to the older AMD video card in anti-aliasing modes, and surpasses it in light graphics modes, not to mention the Radeon HD 6950. Overclocking the latest video card allows you to reduce the gap from the Radeon HD 6970 from 7-12% to 1-2%.

BattleForge: Lost Souls

Only in the third game test - BattleForge: Lost Souls - did the new Radeon HD 6970 and HD 6950 demonstrate an advantage over the GeForce GTX 570, but only in mode without anti-aliasing. When MSAA degree 8x is activated in the game, the NVIDIA video card is again ahead. The Radeon HD 6970 is 10-11% ahead of the HD 6950, but this gap is easily reduced after overclocking the HD 6950. The advantage of the older Cayman over the older Barts is at the 25% mark.

Resident Evil 5

The unconditional victory of the GeForce GTX 570 is obvious. In addition, new Radeons HD 69xx in this game there are some problems with performance in anti-aliasing mode, where new video cards cannot break away from the Radeon HD 6870.

S.T.A.L.K.E.R.: Call of Pripyat

In the game S.T.A.L.K.E.R.: Call of Pripyat, the new Radeon HD 6970 turns out to be the fastest among those tested today, followed by the GeForce GTX 570, and then the Radeon HD 6950. It should be noted that the minimum number of frames per second on a video card with an NVIDIA GPU is higher than on AMD. The advantage of the Radeon HD 6970 over the HD 6950 is 13-17%, but when overclocking the junior video card on the “Cayman” it is reduced to 3-4%. In turn, the Radeon HD 6870 lags behind the Radeon HD 6970 by 35-36%.

Borderlands

It is noteworthy that in modes without anti-aliasing, the new Radeon HD 69xx cannot keep up with the GeForce GTX 570. However, as soon as MSAA 8x is activated in the drivers, the Radeon HD 6970 becomes the leader, and the GTX 570 can only compete with the Radeon HD 6950 (by the way, how is this and was intended by AMD). The Radeon HD 6970 is 10-12% faster than the HD 6950, and 22-26% faster than the Radeon HD 6870 in three out of four modes. But in the most resource-intensive fourth mode with anti-aliasing and at a resolution of 2560x1600, the advantage of the Radeon HD 6970 over the Radeon HD 6870 reaches 82%! Most likely, this advantage was achieved due to the doubled memory capacity of video cards based on Cayman GPUs. And today we will encounter such a situation more than once.

Grand Theft Auto IV: Episodes From Liberty City

In the game Grand Theft Auto IV: Episodes From Liberty City, the GeForce GTX 570 video card is in the lead. There is little difference between all video cards on AMD GPUs.

Left 4 Dead 2: The Sacrifice

Considering the high FPS in the low-resource, but very popular game Left 4 Dead 2: The Sacrifice, the current mode is the anti-aliasing mode, in which the Radeon HD 6970 is faster than the GeForce GTX 570, and the Radeon HD 6950 is almost equal to it. The advantage of the older Cayman over the younger one is 12-13% and is completely “eaten up” by overclocking the Radeon HD 6950. Let’s add that in this game the Radeon HD 6970 is 24-30% faster than the Radeon HD 6870.

Metro 2033: The Last Refuge

Crysis's friend, the game Metro 2033: The Last Refuge, did not reveal the leaders of today's test. The performance of all video cards is extremely low. Here we can only note a good increase in the performance of the Radeon HD 6970 compared to the Radeon HD 6870, ranging from 24 to 36%.

Just Cause 2

In the game Just Cause 2, the GeForce GTX 570 video card can compete with the new Radeon HD 69xx in only one test mode out of four. When increasing the resolution or enabling anti-aliasing, the video card on the NVIDIA GPU significantly lags not only behind the Radeon HD 6970, but also cannot catch up with the Radeon HD 6950. The advantage of the Radeon HD 6970 over the HD 6950 in this game is 12-13%, and when overclocking the HD 6950 is reduced to 0-4%.

Aliens vs. Predator (2010)

We can see a picture similar to Just Cause 2 in the game Aliens vs. Predator. In addition, what is noteworthy here is the impressive jump in performance when moving from the Radeon HD 6870 to the Radeon HD 6970, which in non-aliasing mode is 33%, and when MSAA 4x is activated reaches as much as 54-69%!

Lost Planet 2

The GeForce GTX 570 takes revenge on the previous two games in the new Lost Planet 2, outperforming both new Radeons. Here I would also like to draw attention to the significant difference in performance between the Radeon HD 6870 and the Radeon HD 6970, when the advantage of the video card on the “Cayman” reaches 94% in the most “heavy” graphics mode, and in the “lightest” it does not fall below 31%.

StarCraft 2: Wings of Liberty

In StarCraft 2: Wings of Liberty, all AMD video cards have performance issues in anti-aliasing mode. If without it all Radeon video cards are ahead of the GeForce GTX 570, then when activated in the MSAA 8x drivers they literally fail, demonstrating an extremely low number of frames per second. At the same time, NVIDIA, as we see, does not have such problems in this game.

Mafia 2

In the game Mafia 2, the GeForce GTX 570 video card successfully competes with the Radeon HD 6950, and the older HD 6970 is faster in three out of four test modes. True, the difference in these three is small.

Sid Meier's Civilization V

The results of testing video cards in the game Sid Meier's Civilization V do not stand out against the general background, but are interesting in that at a resolution of 2560x1600 the GeForce GTX 570 video card is inferior to both new AMD products, although we do not see such a picture in the less resource-intensive 1920x1080. Is it really a video card on a GPU NVIDIA running out of memory at the highest resolution?

F1 2010

In the new game F1 2010, the Radeon HD 6950 competes with the GeForce GTX 570, and the Radeon HD 6970, with the exception of the “lightest” graphics mode, is slightly ahead of both of them. By overclocking the HD 6950, this difference is reduced to a minimum or disappears altogether. The Radeon HD 6870 lags behind the Radeon HD 6970 the more, the higher the load on video memory and the requirements for its volume. Thus, at a resolution of 2560x1600 and with 8x anti-aliasing, the advantage of the HD 6970 over the HD 6870 reaches an impressive 64%.

Tom Clancy's H.A.W.X. 2

We think that comments are unnecessary here, since NVIDIA simply has no one to compete with in this test.

In this subsection, all that remains is to attach , and you can move on to the summary diagrams.

Performance comparison summary charts

Using the first two pairs of summary diagrams, we invite you to evaluate the advantage of the AMD Radeon HD 6970 2 GB over the HD 6950 2 GB in nominal operating mode and when overclocking the latter to frequencies of 870/5880 MHz. The results of the Radeon HD 6950 are taken as the zero axis, and the results of the Radeon HD 6970 are shown as a percentage of them:

On average, the Radeon HD 6970 is faster than the Radeon HD 6950 by 10-11% in modes without anti-aliasing, and by 11-13% in modes with anti-aliasing. It is noteworthy that both the maximum (20%) and minimum (2%) advantage of the older video card was obtained in one game - StarCraft 2: Wings of Liberty. The reason is simple - inadequate behavior of AMD Radeon video cards in anti-aliasing modes. Overclocking the Radeon HD 6950 allows you to reduce the lag behind the Radeon HD 6970 to a minimum, and even get ahead of it in some games and quality modes.

The following summary charts compare the nominal Radeon HD 6970 and HD 6950 with the GeForce GTX 570, the results of which are taken as the zero axis, and the advantage or lag of the newly minted competitors is shown as a percentage of this axis:

The fight is being carried out with varying success. Given the small performance difference between the Radeon HD 6970 and Radeon HD 6950, the GeForce GTX 570 manages to compete with both of them. In many games, there is parity between AMD and NVIDIA, or a very slight advantage of one or another video card. In some, such as Tom Clancy's H.A.W.X.2, Lost Planet 2, Borderlands and high-quality modes of StarCraft 2: Wings of Liberty, the advantage is clearly on the side of the GeForce GTX 570. In other games, the Radeon HD 69xx is faster. It is impossible to draw any general conclusion here , because after the game engines themselves, quality and resolution modes are of paramount importance.

Finally, the third pair of summary diagrams is dedicated to comparing the Radeon HD 6970 and Radeon HD 6870, as the older video cards in their classes. The advantage of AMD's single-processor flagship is shown as a percentage of the HD 6870 results:

On average across all games, the Radeon HD 6970 outperforms the Radeon HD 6870 by 26-31% in modes without anti-aliasing, and by 27-45% in modes with anti-aliasing. The usefulness of having 2 GB of video memory on the Radeon HD 69xx is clearly demonstrated by the last diagram, when in some games the difference between the HD 6970 and HD 6870 is more than impressive.

Energy consumption and noise level

energy consumption measurement

The energy consumption of the system as a whole (excluding the monitor) with different video cards installed in it was carried out using specially modified for these purposes, a power supply for two types of load. We emulated a typical gaming load by running two cycles of the Aliens vs. test. Predator in 2560x1600 resolution with 16x anisotropic filtering activated and MSAA 4x anti-aliasing. The theoretical maximum load was created by running FurMark version 1.8.2 in stability test mode and a resolution of 2560x1600 (with AF16x), as well as Linpack x64 (LinX 0.6.4, 4750 MB, 5 threads). Considering that both of these programs generate the maximum load on, respectively, the video system and the central processor, in this way we can find out the peak power consumption of the entire system and determine the power supply required for it (taking into account efficiency).

The results obtained are shown in the diagram:

All test participants included test system demonstrated a level of energy consumption that corresponds to the order of their recommended cost :) The more expensive the video card, the more electricity the system in which it is installed consumes. One of the interesting features here is the 100-watt difference in peak consumption of systems with Radeon HD 6970 and Radeon HD 6950 video cards (in nominal operating mode). At the same time, if you believe the maximum power consumption level stated by AMD, then the HD 6970 should consume only 50 W more than the HD 6950. However, in our case we are talking about the power consumption of the entire system, and not the individual video card included in it, Therefore, it is incorrect to draw such parallels. Separately, the maximum power consumption of centrifugal fans of video card cooling systems was measured, and for the Radeon HD 6970/HD 6950 it was 12.3 W, and for the GeForce GTX 570 it was 9.2 W.

noise level measurement

The noise level of video card cooling systems was measured using an electronic sound level meter CENTER-321 after one in the morning in a completely closed room of about 20 m² with double-glazed windows. The noise level of each cooler was measured outside the system case, when the only source of noise in the room was the cooler itself and its fan(s). The sound level meter, fixed on a tripod, was always located strictly at one point at a distance of exactly 150 mm from the rotor of the centrifugal fan of the cooler. The motherboard, into which a video card was inserted with a cooling system installed on it, was placed at the very corner of the table on a polyurethane foam backing. Bottom line sound level meter measurements are 29.8 dBA, and the subjectively comfortable (not to be confused with low) noise level of coolers when measured from such a distance is around 36 dBA. The rotation speed of the centrifugal fans of video card cooling systems was changed over the entire range of their operation using the controller by changing the supply voltage in 0.5 V steps.

Based on the results of measuring the noise level of video cards from today's testing, the following graph was constructed:

Vertical lines indicate the boundaries of operation of centrifugal fans of video cards in automatic mode when the video cards are loaded using FurMark. Red dotted lines refer to the Radeon HD 6950, red solid lines refer to the Radeon HD 6970, and green dotted lines refer to the GeForce GTX 570.

As you can see, the standard cooling system of the reference GeForce GTX 570 video card not only demonstrates a lower noise level throughout the entire range of operation of the centrifugal fan among the video cards tested today, but also operates at lower speeds, regardless of FurMark. The maximum speed of its centrifugal fan after 12 minutes of load was 2430 rpm. The new Radeon HD 69xx, alas, cannot boast of such a modest noise level. Thus, the centrifugal fan of the Radeon HD 6950 spun up to 2820 rpm during testing, and the centrifugal fan of the Radeon HD 6970 even reached 4760 rpm. Although, in the latter case, as mentioned above, we are still inclined to blame the problematic copy of the video card rather than draw final conclusions about the unacceptably high noise level of the Radeon HD 6970. We add that under gaming load, all video cards produced sounds that were more comfortable for the ear.

Let's move on to summing up the results of our first acquaintance with new video cards based on AMD “Cayman” GPUs.

Conclusion

Despite the fact that in terms of pure performance, the AMD Radeon HD 6950 and Radeon HD 6970 could not unconditionally outperform the NVIDIA GeForce GTX 570 and, with the exception of certain games and tests, of course, will not be able to surpass the GeForce GTX 580, the new video cards should attract the attention of computer users who decided to upgrade their video card, or who are purchasing it again. The reason for this, first of all, is the lower recommended price, which for the Radeon HD 6950 is $50 lower than the GeForce GTX 570, and for the Radeon HD 6970 it is immediately $130 lower than the GeForce GTX 580. In this regard, you need to understand that market realities are sometimes very far from the prices predicted or recommended by marketers, and both top NVIDIA video cards have a head start, since they are already present on the market, while we are still awaiting the appearance of the new Radeon HD 69xx in retail. Therefore, time will tell how it will actually turn out.

If we talk directly about video cards, then the Radeon HD 6950, with its 10-13% lag from the Radeon HD 6970 and its almost complete reduction during overclocking, and also taking into account the 70-dollar difference in price and the same amount of video memory 2 GB, looks more attractive, than the older video card. In addition, the Radeon HD 6950 operates at lower temperatures and, as a result, at lower noise levels. Although, of course, calling this video card cold and quiet would not be entirely true. Connoisseurs of quiet system units are probably already looking at replacing the standard cooling system of the Radeon HD 69xx, and overclockers are eyeing the possibility of installing water blocks. We consider the appearance of a backup BIOS on video cards to be a positive development. Now, if the working BIOS is damaged, another video card will not be needed to restore it. The new interesting type of anti-aliasing “Enhanced Quality AA” and energy saving modes will also not go unnoticed in the following materials.

What's next? Next, in our opinion, we will witness a response from NVIDIA, which, as we would like to believe, will not only reduce prices for the GeForce GTX 570 and GTX 580, but also release an attractive GeForce GTX 560 in terms of price-performance ratio. And from this, of course, we will all only benefit, so the appearance of “Caymans” in our networks is quite difficult to overestimate.

Thank you:
Russian representative office of AMD and personally Kirill Kochetkov,
Inno3D company and Natalia Lagunenko personally
for video cards provided for testing.

Other materials on this topic

Nvidia GeForce GTX 580: born to be first
Fallout: New Vegas and modern video cards
Simple arithmetic: AMD Radeon HD 68xx video cards in CrossFireX modes

The release of a new version of the DirectX API raises hopes for another revolution in graphics, steadily bringing it closer to the bright future promised seven years ago in the form of cinematic quality. Considering the shift in the priorities of game developers towards console projects, we won’t have to wait for another PC masterpiece any time soon. In the meantime, GPU developers are on the defensive, releasing a second generation of desktop video cards with support for DirectX 11.

The transition to this API for NVIDIA turned out to be difficult: the cards, which were six months late, were expensive to produce, had high power consumption and disabled some of the functional blocks of the company's flagship. Only now has it been able to release more or less decent solutions. AMD, on the contrary, is steadily following its strategy of developing simple single-chip products and hi-end accelerators based on two graphics processors. And if it weren’t for her, interest in multi-chip configurations would hardly be at the proper level.

AMD decided to change generations of video cards by expanding the Radeon line, thereby introducing confusion into the company's lineup. The new HD 6800 series actually turned out to be slower than their predecessors, but due to the price they should occupy a niche between the Radeon HD 5800 and HD 5700, competing with the GeForce GTX 460, which is good in terms of its characteristics.

High-performance solutions based on a single GPU are destined for a place where dual-chip products usually reigned. Thus, the Radeon HD 6900 series will now include all cards based on the Cayman core - the Cypress receiver.

Cayman

The Cayman GPU, on which the Radeon HD 6970 and HD 6950 video cards are based, is made using the same 40-nm process technology as its predecessor. But unlike Cypress, the new kernel has undergone some changes. Thus, the number of SIMD cores was increased from 20 to 24 (16 superscalar stream processors each), the number of texture units accordingly increased to 96. The rendering units were slightly improved to work with anti-aliasing, but their number remained unchanged, i.e. 32 pieces. The smoothing algorithm has also undergone some improvements. The “graphics engine” is now divided into two, each part serving its half of the core. Some blocks in it were twice as large, which made it possible to increase the speed of working with geometry and tessellation (the latter is almost three times faster than Cypress). In addition, each half of the kernel has its own thread manager (Ultra-Threaded Dispatch Processor), which is responsible for distributing the load among the execution units.

But the most important thing is that AMD abandoned the fifth ALU (T-unit) in superscalar processors, capable of executing a complex instruction (Special Function) - now its role is distributed among three of the four remaining ones. This approach, according to the developer, allows you to increase productivity by up to 10% per square millimeter of chip area.

Ultimately, the number of stream processors was 1536, while Cypress had 1600. But this is unlikely to negatively affect the performance of the new GPU, especially since its frequency has become slightly higher - 880 MHz versus 800 MHz. The GDDR5 memory frequency has also been increased - up to 5500 MHz, although the bus width is the same 256 bits. But the volume of the video buffer, without hesitation, was doubled, and now it is equal to 2048 megabytes, which will be especially useful at high screen resolutions and in CrossFireX mode.

The characteristics of the younger model are a little more modest: 1408 stream processors, 88 TMUs, the same number of ROPs, core and memory frequencies are 800 and 5000 MHz, respectively. The memory capacity is also 2048 MB.

Video adapter	Radeon HD6970	Radeon HD6950	Radeon HD6870	Radeon HD6850	Radeon HD5870	Radeon HD5850
Core	Cayman XT	Cayman Pro	Barts XT	Barts Pro	RV870 (Cypress)	RV870 (Cypress)
Architecture	VLIW4	VLIW4	VLIW5	VLIW5	VLIW5	VLIW5
	2640	2640	1700	1700	2154	2154
Technical process, nm	40	40	40	40	40	40
Core area, sq. mm	389	389	255	255	334	334
	1536	1408	1120	960	1600	1440
Number of texture blocks	96	88	56	48	80	72
Number of rendering units	32	32	32	32	32	32
Core frequency, MHz	880	800	900	775	850	725
Memory bus, bit	256	256	256	256	256	256
Memory type	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Memory frequency, MHz	5500	5000	4200	4000	4800	4000
Memory capacity, MB	2048	2048	1024	1024	1024	1024
	11	11	11	11	11	11
Interface	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1
Declared maximum power consumption, load/idle, W	(250) 190/20*	(200) 140/20*	151/19	127/19	188/27	170/27

* —The maximum power is indicated in brackets

Naturally, with an increase in the number of functional blocks, while the technological process remained unchanged, the area of ​​the Cayman crystal increased to 389 mm 2 - after all, it became “thicker” by 500 million transistors. In this case, we should expect an increase in power consumption of new cards. But with the release of Cayman, AMD changed the presentation of this indicator. Now, thanks to the new PowerTune technology, the GPU frequency, and not just its voltage, can dynamically change depending on the load and power consumption level. The latter percentage can be configured independently in the Catalyst Control Center (ATI Overdrive tab).

With standard settings, the power consumption of the Radeon HD 6970 in typical gaming applications is 190 W, while the peak in the OCCT benchmark can reach 250 W. For the Radeon HD 6950, these values ​​are smaller and are 140 and 200 W, respectively. When idle, the cards consume only 20 W.

If you move the slider in ATI Overdrive to -20%, the GPU frequency will change within 880-500 MHz (with a higher load it will decrease, with a lighter load, on the contrary, it will increase), thereby reducing the power consumption and heat dissipation of the card.

If you move it to +20%, then instead of changing the frequency, the user will have access to a higher overclocking level.

Among the pleasant moments for enthusiasts, we note support for Dual BIOS and a special switch near the CrossFire connector, which allows you to select one or another chip with microcode. One version is designed for constant updating, and the second is write-protected to prevent information corruption. This will be especially useful for those who like to reflash cards - if one chip fails, then you can continue to use the accelerator.

The cooling system of Radeon HD 69xx video cards is based on an evaporation chamber, like the Radeon HD 5970 and the new GeForce GTX 580. Previously, this design was found in Sapphire products, and as practice has shown, the efficiency of such solutions is higher than that of standard heatpipe coolers.

The number of connectors for connecting monitors has increased and now, instead of one standard DisplayPort, users will have access to two miniDP versions 1.2. Eyefinity support remains in place.

As we can see, the changes in the new Cayman GPU are enough to talk about a new generation of video cards supporting DirectX 11. But how will these changes be reflected in real applications? This is what we will find out below.

Now let's look directly at the cards themselves. Before us are almost complete twins. Instead of the smooth lines and rounded shapes characteristic of the previous Radeon series, the newcomers have a “brick” appearance with red accents on the sides.

Radeon HD 6970 is equipped with six- and eight-pin power connectors.

The younger Cayman has two six-pin sockets.

The back side of the video cards is covered with a black metal plate.

On back panel Output via a pair of DVI and mini-DisplayPort connectors, plus one HDMI.

The cooling system of both models is the same.

Plate on back side The board does not cool the memory chips, as was the case with the Radeon HD 5870 or Radeon HD 5970. All GDDR5 chips are soldered on the front side. The design of the turbine is quite typical and includes a main radiator for cooling the core and a massive metal base that acts as a heat sink for memory and power elements of the power supply system. Only instead of traditional heat pipes, reference cards based on Cayman are equipped with a radiator with an evaporation chamber.

The original CO looks quite impressive.

A Foxconn PVB070G12N fan with a diameter of 70 mm is responsible for blowing the entire structure.

The lower card uses a different fan, but of the same diameter.

Let's now take a look at the design of the boards. Below are photos of the Radeon HD 6970

But the Radeon HD 6950:

No difference! And even with the Radeon HD 5870 the differences are minimal.

Video cards differ in different GPUs and operating frequencies.

Radeon HD 6970 (left) and Radeon HD 6950

Both adapters are equipped with 2 GB of GDDR5 video memory. The older model uses Hynix H5GQ2H24MFR R0C microcircuits, designed for an operating frequency of 6 GHz.

The Radeon HD 6950 uses Hynix H5GQ2H24MFR T2C chips, rated at 5 GHz.

The GPU-Z utility correctly determines the frequencies of new video cards, making an error only in the number of computational units.

Monitoring and overclocking in MSI Afterburner also works great. So far there is just no support for softvolt mod. The power converters on the cards are built on Volterra VT 1556MF controllers, which support I2C and software voltage control is possible on them. So we expect this feature in future versions of Afterburner.

A little about temperature conditions. The Radeon HD 6970 easily warms up to 90°C in gaming applications (at 23°C indoors). Almost like the GeForce GTX 480, but the noise level of the new AMD product is much lower, the hum is noticeable, but tolerable.

Furmark's stress test warmed the card up to 92°C, but the noise level was already very high.

Let's comment a little on the Power Control function, which allows you to switch the card to a more energy-efficient mode. At a value of -5% we did not see any drop in performance, but at -20% the core frequency jumped from 500 to 800 MHz and in the Crysis Warhead test the result dropped by 26%.

Increasing Power Control helped a little with overclocking. The card was able to operate at higher frequencies, but stability was still not achieved at them. Plus, we also noticed a miniscule drop in performance in some benchmarks (up to 0.2%) when increasing the Power Control limit, even at low overclocking. In the end, we still settled on a frequency of 965 MHz, which the core obeyed even at the nominal voltage. This is also a very good result. The memory was overclocked to 1565 (6260) MHz. Stability in this mode was possible only at increased turbine speeds.

The younger Radeon HD 6950 turned out to be cooler and quieter. In game mode, it did not heat above 80 °C.

Even at Furmark, the peak temperature did not exceed 87 °C.

From the nominal 800/5000 MHz we managed to overclock this copy to 860/5640 MHz. The memory results can be explained by the use of other chips, but we expected more from the GPU.

Remembering the results of overclocking the older card and the identical design of the boards for both Radeons, users can only wait for utilities that will allow them to control the voltage. Power Control in this case had no effect on overclocking at all. Characteristics of tested video cards

For comparison with the AMD Radeon HD 6970 and HD 6950, we used the AMD Radeon HD 5970, Force3D Radeon HD 5870, Zotac GeForce GTX 580 and Inno3D GeForce GTX 480 cards that participated in the recent testing of the GF110-based solution. For comparison, their characteristics are listed in the following table:

Video adapter	Radeon HD6970	Radeon HD6950	Radeon HD5970	Radeon HD5870	GeForce GTX 580	GeForce GTX 480
Core	Cayman XT	Cayman Pro	RV870 x2 (Hemlock)	RV870 (Cypress)	GF110	GF100
Number of transistors, million pieces	2640	2640	2154 x2	2154	3000	3200
Technical process, nm	40	40	40	40	40	40
Core area, sq. mm	389	389	334 x2	334	512	526
Number of stream processors	1536	1408	1600x2	1600	512	480
Number of texture blocks	96	88	80 x2	80	64	60
Number of rendering units	32	32	32x2	32	48	48
Core frequency, MHz	880	800	725	850	772	701
Shader domain frequency, MHz	880	800	725	850	1544	1401
Memory bus, bit	256	256	256	256	384	384
Memory type	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Memory frequency, MHz	5500	5000	4000	4000	4008	3696
Memory capacity, MB	2048	2048	1024	1024	1536	1536
Supported DirectX Version	11	11	11	11	11	11
Interface	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1	PCI-E 2.1	PCI-E 2.0	PCI-E 2.0
Declared maximum power consumption, W	(250) 190	(200) 140	294	188	244	250

Test bench

The test bench configuration is as follows:

• processor: Core i7-965 ([email protected] GHz, BCLK 172 MHz);
• cooler: Thermalright Venomous X;
• motherboard: Gigabyte GA-X58A-UD3R (Intel X58 Express);
• memory: G.SKILL F3-12800CL8T-6GBRM (3x2GB, DDR3-1600@1720 MHz, 8-8-8-24-1T);
• hard drive: Hitachi HDS721010CLA332 (1 TB, SATA2, 7200 rpm);
• power supply: Seasonic SS-850HT (850 W);

• operating system: Windows 7 Ultimate x64;
• video card drivers: NVIDIA GeForce 260.99, ATI Catalyst 10.11.

User Account Control, Superfetch, Windows Defender, swap file and interface visual effects.

Testing methodology

The games in our testing are divided into three groups, according to the supported DirectX version. Applications for the ninth version of this API come first, and latest games under DirectX 11. Within each group, applications are located in accordance with the release date, with the exception of Metro 2033, this one demanding game, and we traditionally place the results in it at the very end of the article. All tests were performed at 1920x1200 resolution on maximum settings quality, anisotropic filtering AF16x is always enabled. Anti-aliasing was activated from the game menu; forcing through drivers was not used. The list of games and testing methodology are described in more detail in the GeForce GTX 580 review.

Test results in applications running DirectX 9

Borderlands

This game based on Unreal Engine 3 favors NVIDIA video adapters. The Radeon HD 6950 is slightly inferior to the Radeon HD 5870, but the HD 6970 model even outperforms the dual-chip card in terms of minimum fps.

Mafia 2

No more than 12% performance difference between Cayman cards. The older Radeon HD 6970 is 8% faster than the GeForce GTX 480.

The very low result of the Radeon HD 5970 immediately catches your eye. Multi-chip configurations are not supported by the game. The Radeon HD 6950 is a couple of percent behind the HD 5870, but its older brother is hot on the heels of the GeForce GTX 580.

Test results in applications running DirectX 10

Nothing unexpected. The Radeon HD 6950 is a couple of percent behind the Radeon HD 5870, and the Radeon HD 6970 is 9% ahead of the GeForce GTX 480.

Far Cry 2

The complete superiority of the Fermi architecture. You can also see a good advantage of the Cayman over the Cypress, when the overclocked Radeon HD 6970 almost overtakes the Radeon HD 5970.

Tom Clancy's H.A.W.X. 2

Common situation. The active use of tessellation baffles AMD cards, especially older Cypress-based solutions; even the Radeon HD 6950 turns out to be 27% faster than the Radeon HD 5870.

Sid Meier's Civilization 5

Unusual results for both GeForces - when overclocked, there was not an increase in the final result, but a tiny decrease to about 3000 points (these data are roughly rounded). In face value there is a negligible difference between them. But even with such a miniscule drop in results, NVIDIA cards left behind all AMD's single-chip competitors.

According to the results of our testing, AMD's new single-chip flagships demonstrated superiority over the old Radeon HD 5870, but did not create any sensation. Still, something more was expected from the new series. Polishing the architecture and fixing bugs made it possible to slightly increase performance; the rest was completed by higher frequencies.

The performance of the Radeon HD 6950 is close to the Radeon HD 5870, in some places the newcomer is inferior by a couple of percent, and in some places it comes out ahead by the same tiny margin. The main advantage is noticeable in games that actively use tessellation (Lost Planet 2, H.A.W.X. 2). The good lead over its predecessor in Metro 2033 is partly due to the increased memory capacity - in other cases the second gigabyte is not in demand.

The Radeon HD 6970 is already a more interesting product; it definitely outperforms the Radeon HD 5870. Although in some games for older versions of DirectX the difference between them is no more than 10%. But in DirectX 11 we are seeing a thirty percent increase. This is still not enough to compete strongly with the dual-chip Radeon HD 5970 and NVIDIA's flagship. The older Cayman successfully competes only with the GeForce GTX 480, although in DirectX 11 the first place almost always goes to the old NVIDIA. However, no one was aiming to be a leader - the new graphics cards are designed to create competition for the newly minted GeForce GTX 570, and as a rival for the GeForce GTX 580, the company still has the same Radeon HD 5970 in its arsenal.

Rather, the Cayman is nothing more than a preparation for the transition to a thinner process technology, where new graphics processors will be able to acquire a large number computing units with an acceptable level of energy consumption. Now, of course, this is impossible to achieve. This is confirmed by almost the same number of stream processors and TDP levels as the Cypress. Quite the contrary, the Radeon HD 6970 is reminiscent of the GeForce GTX 480 in its hot temper, although, fortunately, its noise characteristics are no different from other top-end video adapters. The main advantage of AMD's new products is the price. Their cost is quite attractive compared to the same GeForce GTX 580. To compete in the top segment, the company will have to release a solution based on two Caymans. But that is another story...

Testing equipment was provided by the following companies:

• 1-Inkom - memory G.SKILL F3-12800CL8T-6GBRM;
• AMD - Radeon HD 6970, Radeon HD 6950 and Radeon HD 5970 video cards;
• Gigabyte - motherboard Gigabyte board GA-X58A-UD3R;
• Inno3D - Inno3D GTX 480 video card;
• PCShop - Force3D HD5870-1GB video card (F5870P-G5-390-FS);
• Syntex - Seasonic SS-850HT power supply (S12D-850);
• — hard drive WD3200AAKS;
• Zotac - GeForce GTX 580 (ZT-50101-10P) video card.

Specifications

Core: Cayman* Number of transistors: 2.64 billion * Technical process: 40nm* Number of stream processors: 1408 pcs. * Graphics core frequency: 800 MHz * Frequency stream processors: 800 MHz * Type, Memory: GDDR5, 2 GB * Memory frequency: 5000 MHz * Data bus: 256 bit* Number of texture blocks: 88 pcs. * Number of rasterization blocks: 32 pcs. * Energy consumption: 200 W * Interface: PCIe 2.0 x16* Price as of March 2011: 10,000 rubles

In the February issue we thoroughly tested the new video card from AMD, Radeon HD 6970, and came to the disappointing conclusion that compared to the much cheaper Radeon HD 5870 it's a little faster in DX11 applications, but nothing more. Another creation of the company, released on the same day, has only now reached our laboratory. What will please you Radeon HD 6950, the little sister of the company's single-chip flagship?

Minimal changes

AMD engineers decided to contain the potential of the Radeon HD 6950 in a proven way: by disabling some of the functional blocks of the graphics core and reducing operating frequencies. Of the 24 SIMD cores, two were cut off, so the chip lost 128 stream processors and 8 texture units. However, the remaining arsenal is still impressive: the video card has 1408 stream processors and 88 ROPs - this should be enough for any modern game. The memory subsystem was not cut: the device’s bus is still 256-bit, and the volume of soldered GDDR5 is 2 GB. The chips themselves, however, are of lower quality, so their operating frequency is 5000 MHz versus 5500 MHz for the older model. Note that many manufacturers have already announced cheaper Radeon versions HD 6950 with 1 GB of memory.

AMD usually gives video cards based on the same chip different designs, but this is not the case with the HD 6950/6970. Externally, the adapters are absolutely identical - the same length, the same design of the cooling system casing. The only difference lies in the power connectors: the youngest board is content with two 6-pin PCIe, since its maximum power consumption does not exceed 200 W, but the HD 6970 with its TPD of 250 W is equipped with 6-pin and 8-pin connectors.

As practice shows, it was not necessary to make this distinction: HD 6950 boards converted into HD 6970 by flashing the BIOS have enough power. Under the CO casing you can see the already familiar cooler based on an evaporation chamber. The PCB design of both video cards is also identical.

In your own niche

Slides from the presentation of the Radeon HD 6950 indicate that this adapter should create its own niche in the market and is not directly responsible for competing with any video card. In essence, this is true: the recommended price of the new product is $280, and there are no other relevant solutions in this segment today. For comparison with ours ASUS EAH 6950 Radeon HD 6970, we took four video cards -, HD 6870 And HD 5870 GeForce GTX 470

What's the result? The newcomer's lag behind the Radeon HD 6970 ranges from 5 to 12 percent, depending on the test. Given the large difference in price, this result from the HD 6950 can be considered excellent. The gap between the Radeon HD 6950 and the recent HD 6870 is still the same 5-12 percent in DX10 benchmarks, but the new Radeon HD 6950 copes much better with rendering under DirectX 11 - thanks to the new tessellation engines. But the Radeon HD 5870 is already outdated and is inferior to the Radeon HD 6950 in tests. A representative of the green camp, GeForce GTX 470, lost to the HD 6950 in all tests, which was quite expected.

* * *

It can be stated that AMD It turned out to be a really good video card with support for many new technologies, including the long-awaited tessellation. Radeon HD 6950 st O It’s worth your money, and if you are willing to spend about 10,000 rubles on a graphics card, then the choice is obvious.

PROS:

• significantly cheaper, but only slightly slower than the Radeon HD 6970
• + by flashing the BIOS, you can turn the board into a Radeon HD 6970
• There are cheaper versions with 1 GB of memory
• NVIDIA does not have boards in this price range

MINUSES:

• the price is too high due to lack of competitors

Table 1

Specifications Characteristic AMD Radeon HD 695 0 AMD Radeon HD 6970 AMD Radeon HD 6870 AMD Radeon HD 5870 NVIDIA GeForce GTX 470 Core Cayman Cayman Barts XT Cypress GF100 Number of transistors 2.64 billion 2.64 billion 1.7 billion 2.15 billion 3 billion Technical process 40nm 40nm 40nm 40nm 40nm Number of stream processors 1408 pcs. 1536 pcs. 1120 pcs. 1600 pcs. 448 pcs. Graphics core frequency 800 MHz 880 MHz 900 MHz 850 MHz 607 MHz Stream processor frequency 800 MHz 880 MHz 900 MHz 850 MHz 1215 MHz Type, memory capacity GDDR5, 2 GB GDDR5, 2 GB GDDR5, 1 GB GDDR5, 1 GB GDDR5, 1.28 GB Memory frequency 5000 MHz 5500 MHz 4200 MHz 4800 MHz 3348 MHz Data bus 256 bit 256 bit 256 bit 256 bit 320 bit Number of texture blocks 88 pcs. 96 pcs. 56 pcs. 80 pcs. 56 pcs. Number of rasterization blocks 32 pcs. 32 pcs. 32 pcs. 32 pcs. 40 pcs. Energy consumption 200 W 250 W 151 W 188 W 215 W Interface PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 Price as of March 2011 10,000 rubles 12,500 rubles 7600 rubles 9600 rubles 7800 rubles

table 2

Synthetic tests 3DMark Vantage Video card model GPU CPU Overall AMD Radeon HD 6950 17 572 17 303 17 504 100% AMD Radeon HD 6970 19 098 17 175 18 578 106% AMD Radeon HD 6870 15 131 16 493 15 450 88% AMD Radeon HD 5870 17 402 17 326 17 383 99% NVIDIA GeForce GTX 470 13 656 45 937 16 566 95% Unigine Heaven Benchmark 2.0 Video card model FPS Overall Performance ratio AMD Radeon HD 6950 23,7 596 100% AMD Radeon HD 6970 24,8 626 105% AMD Radeon HD 6870 13,8 348 58% AMD Radeon HD 5870 11,4 288 48% NVIDIA GeForce GTX 470 19,9 502 84%

Table 3

Gaming tests (frames per second) Settings AMD Radeon HD 6950 AMD Radeon HD 6970 AMD Radeon HD 6870 AMD Radeon HD 5870 NVIDIA GeForce GTX 470 Resident Evil 5 (DX10) High, 1680x1050, AF 16x, AA 8x 101,5 110,1 96,1 105,8 97 High, 1920x1080, AF 16x, AA 8x 94,5 108,5 89,6 99,4 88,2 Performance ratio 100% 112% 95% 105% 94% Devil May Cry 4 (SC2, DX10) SuperHigh, 1680x1050, AF 16x, AA 8x - - - - 122 SuperHigh, 1920x1080, AF 16x, AA 8x 143,9 157,9 126,3 131,5 110,4 Performance ratio 100% 110% 88% 91% 77% Aliens vs. Predator (Demo, DX11) VeryHigh, 1680x1050, AF 16x, AA 2x 49,6 53,9 39,6 48,7 47,2 VeryHigh, 1920x1080, AF 16x, AA 2x 44,4 48,3 35,4 43,5 42,3 Performance ratio 100% 109% 80% 98% 95% Value for money 100% 125% 76% 96% 78% Performance ratio 100% 110% 88% 98% 89%

AMD Radeon HD 6970M(also known as ATI Mobility Radeon HD 6970) is a high-performance graphics card for large, powerful laptops. It is based on the HD 6850 desktop graphics card and hence has UVD3 and Eyefinity+ GPU features. Blackcomb XT is the code name for the GPU. The power consumption of the HD 6970M is exactly the same as that of the GeForce GTX 480M.

The 960 stream processors are based on a "legacy" 5D architecture similar to the one used in the 5000 series. That is, the number of shaders is not comparable to modern single-core graphics NVIDIA processors. WITH AMD Mobility Radeon HD 6950M Tessellation performance will be improved (as is the case with the 6800-series desktop). Stream processors can be used with DirectX 11, OpenGL 4.1, DirectCompute 11 and OpenCL. Can also be used to perform such complex tasks like video transcoding.

In terms of 3D performance, the adapter falls between the high-end GeForce GTX 470 and 485M. Judging by the tests, the AMD Radeon HD 6970M is faster than the GeForce GTX 480M by an average of 33%. However, the new GTX 485M adapter is still more powerful than the HD 6970M, by about 12%. Compared to the ATI Mobility Radeon HD 5870 (the previous generation video card), the Radeon 6970M is 50% more powerful than its predecessor. However, this card also consumes more power.

WITH Radeon HD 6970M gamers can safely play the latest games on high settings. Only Crysis and Metro 2033 will slow down at a resolution of 1920x1080 pixels.

The new UVD3 supports MPEG-4 AVC/H.264, VC-1, MPEG-2, Flash, as well as Multi-View Codec (MVC) and MPEG-4 Part 2 (DivX, XVid) decoding.

In addition, the 6900 series integrates an HD audio controller for transmitting HD Audio (TrueHD or DTS Master Audio) via HDMI and DisplayPort (for example, for Blu-Ray video).

AMD HD3D technology supports Blu-Ray 3D and 3D displays (integrated and external). However, this technology is clearly inferior to NVIDIA 3D Vision.

Thanks to Eyefinity support, the graphics chip allows you to connect up to 6 monitors simultaneously. However, this is only possible if the laptop has enough DisplayPort. Chip power consumption is identical to cards Nvidia GeForce GTX 480M or 485M, that is, about 100 W with the MXM module and 2 GB of GDDR5 video memory. The TDP of the card is approximately 75-100 W. Therefore, the card can only be installed in large laptops with a good cooling system.

Manufacturer:	AMD
Series:	Radeon HD 6970M 960@680MHz
Code:	Blackcomb XT
Streams:	960 - unified
Clock frequency:	680* MHz
Shader frequency:	680* MHz
Memory frequency:	900* MHz
Memory bus width:	256 Bit
Memory type:	GDDR5
Maximum memory:	2048 MB
Common memory:	No
DirectX:	DirectX 11, Shader 5.0
Transistors:	1700 million
Technology:	40 nm
Laptop size:	big
Release date:	04.01.2011

*Clock speeds indicated may be subject to change by the manufacturer.