In February 2010, AMD introduced to the video card market a new graphics accelerator Radeon HD 5570 512 MB, based on the Redwood core and belonging to the lower price segment. This product has quite attractive technical specifications and supports the DirectX 11 API. Today we will test this video card and see what it can do.

advertising

Test configuration

The tests were carried out on the following stand:

• CPU: Intel Core i7 920 (Bloomfield, D0, L3 8 MB), 1.18 V, Turbo Boost - on, Hyper Threading - off - 2660 @ 4000 MHz
• Motherboard: GigaByte GA-EX58-UD5, BIOS F5
• CPU cooling system: Cooler Master V8 (~1100 rpm)
• RAM: 2 x 2048MB DDR3 Corsair TR3X6G1600C7
(Spec: 1528MHz / 8-8-8-20-1t / 1.5V) , X.M.P. - off
• Disk subsystem: SATA-II 500 GB, WD 5000KS, 7200 rpm, 16 MB
• Power Supply: Thermaltake Toughpower 1200W (stock fan: 140mm blown)
• Frame: outdoor test stand
• Monitor: 24" BenQ V2400W (Wide LCD, 1920x1200 / 60Hz)

Video cards:

• Radeon HD 4730 512 MB - 625/625/3600 @ 820/820/4600 MHz (Sapphire)
• Radeon HD 5570 512 MB - 650/650/1800 @ 780/780/2200 MHz (PowerColor)
• Radeon HD 4670 512 MB - 750/750/2000 @ 850/850/2300 MHz (HIS)
• GeForce 9600 GT 512 MB - 650/1625/1800 @ 720/1900/2200 MHz (Zotac)
• GeForce 9600 GSO 512 MB - 650/1625/1800 @ 700/1800/2100 MHz (Zotac)
• GeForce GT 240 512 MB - 550/1340/3400 @ 600/1412/4400 MHz (Inno3D)
• GeForce GT 240 512 MB - 550/1340/2000 @ 600/1412/2200 MHz (Gainward)

Software:

• Operating system: Windows 7 build 7600 RTM x86
• Video card drivers: ATI Catalyst 10.3 and NVIDIA GeForce 197.45 WHQL
• MSI AFTERBURNER 1.5.1

advertising

Testing tools and methodology

Looking closely at the technical specifications of the Radeon HD 5570 512 MB, we have the impression that this video card should replace a worthy veteran - the Radeon HD 4670 512 MB. Indeed, in terms of balancing functional blocks (hereinafter referred to as FB), video cards are similar - the Radeon HD 5570 512 MB has more FB, but they operate at a lower frequency than the Radeon HD 4670 512 MB FB. Let's see if the Radeon HD 5570 512MB can replace the Radeon HD 4670 512MB.

For a more visual comparison of processors, all games used as test applications were launched at 1280x1024 and 1680x1050 resolutions.

The following games used performance measurement tools (benchmark):

• Colin McRae: DIRT 2
• Crysis Warhead
• Just Cause 2
• Resident Evil 5 (scene 1)
• Warhammer 40000 Dawn of War 2: Chaos Rising

A game where performance was measured by loading demo scenes:

• Left 4 Dead 2

In these games, performance was measured using the FRAPS v3.0.3 build 10809 utility:

• Aliens vs Predators (2010)
• Battlefield: Bad Company 2
• Bioshock 2
• Borderlands
• Call of Duty: Modern Warfare 2
• Dragon Age: Origins
• Grand Theft Auto 4
• Mass Effect 2
• Metro 2033
• Napoleon: Total War
• Need for Speed: SHIFT
• prototype
• Risen
• S.T.A.L.K.E.R.: Call of Pripyat

In all games measured minimal and medium FPS values.

In tests in which there was no possibility of measuring minimum FPS, this value was measured by the FRAPS utility.

vsync disabled during testing.

To avoid errors and minimize measurement errors, all tests were performed three times. When calculating the average FPS, the arithmetic mean of the results of all runs was taken as the final result. As the minimum FPS, the minimum value of the indicator based on the results of three runs was chosen.

Let's go directly to the tests.

Introduction

In September 2008 (almost a year and a half ago), AMD surprised gamers on a budget with the release of its Radeon HD 4670 graphics card. The listed price was $80, which meant that the card was aimed at direct competition with the entry-level GeForce 9500 GT model, and the specifications of the 4670 were comparable with the flagship of the previous generation Radeon HD 3870.

The performance of the Radeon HD 4670 surpassed the competition. With 320 streaming cores, the Radeon HD 4670 has changed the balance of power in this price segment. The release of the 4670 meant that Nvidia had to create the GeForce 9600 GSO based on high-end GPUs, which are more expensive to manufacture. Not to mention that the prices for the GeForce 9600 GT had to be lowered.

Since its announcement, the Radeon HD 4670 has remained one of the best "budget" graphics cards on the market (and we regularly include it in our recommendations). In addition, she for a long time remained the fastest reference card that did not require additional PCIe power until Nvidia introduced its GeForce GT 240, which then gave way ATI Radeon HD 5670 .

AMD really raised the bar for $80 graphics performance with the release of the Radeon HD 4670. As it happens, today AMD is introducing the successor to this acclaimed graphics card in the form of , which should also be priced at $80.

video card Radeon HD 5450 too slow to provide decent gaming performance at an affordable price, and Radeon HD 5670 costs more than $100 (), so let's hope that the new HD 5570 will be the "golden mean" for many gamers on a budget. Perhaps we will get the game on three Eyefinity monitors even in the entry-level market?

Radeon HD 5570 architecture

At the end of the article on the exit Radeon HD 5670, we have expressed our hope that the Radeon 5500 line will receive a version of the Radeon HD 5670 with DDR3 memory. And so it happened!

		Radeon HD 5670
Stream Cores	400	400
texture blocks	20	20
	8	8
Core frequency	650 MHz	775 MHz
Memory frequency	900 MHz (DDR3)	1000 MHz (GDDR5)
Memory bus	128 bit	128 bit
Effective memory frequency	1.8GHz	4 GHz
Number of transistors (millions).	627	627

Yes, the new one is Radeon HD 5670 with DDR3 memory, as well as with a GPU slowed down by 125 MHz. If you look at the effective memory frequency, you will find that the Radeon HD 5570 provides less than twice the bandwidth compared to the HD 5670. This is due to the fact that DDR3 memory theoretically provides half the bandwidth of GDDR5 at equal clock speeds and memory bus width. As a result, we can expect a significant performance difference between the two graphics cards.

Let's take a look at the GPU schematic.

GPU like Radeon HD 5670, contains five SIMD engines, each with four texture units and 16 stream processors. Of course, each stream processor supports five ALUs (AMD calls them stream cores). As a result, this GPU contains a total of 400 stream cores and 20 texture units. Note that there are two 64-bit memory controllers on the back of the render pipeline. There you can also see a couple of sections of four ROP blocks, which gives a total of eight raster operation blocks (ROP) and a 128-bit memory interface.

Let's compare the specifications of the video card with the Radeon HD 4670 model, which the 5570 is intended to replace.

		Radeon HD 4670
Stream Cores	400	320
texture blocks	20	32
Raster Operations Units (ROPs)	8	8
Process technology	40 nm	55 nm
Core frequency	650 MHz	750 MHz
Memory frequency	900 MHz	1000 MHz
Memory bus	128 bit	128 bit
Effective memory frequency	1.8GHz	2 GHz
Number of transistors (millions).	627	514

At first glance, the 5570 looks impressive due to the increase in the number of streaming cores. But if you dig a little deeper, you can see some holes in the armor of the novelty. The old Radeon HD 4670 has a 100MHz GPU clock advantage over the new one HD 5570, which almost negates the higher shader count of the HD 5570. The older HD 4670 also has more texture units and higher memory clock speeds. However, in our experience, many Radeon HD 4670 cards shipped with 800 MHz memory (200 MHz below the reference). But judging by the reference specs, the Radeon HD 5570 can be called a sort of parallel port of the HD 4670.

So far, nothing surprises us. Sub-$100 Radeon HD 5000 graphics cards aren't designed to raise the bar for gaming performance. They should simply give the same 3D performance for this segment, but with some bonuses related to the new line: support for DirectX 11, Eyefinity, Dolby TrueHD and DTS-HD Master Audio transfer to an external receiver in encoded form (bitstream). Good news for gamers looking to upgrade from integrated graphics to a discrete graphics card. And not so good if you have already purchased a graphics card from the Radeon HD 4000 line.

Radeon HD 5570 features

It's hard not to repeat when talking about the functionality of the 5000 series video cards: this is already the eighth new-generation AMD video card reviewed on our pages over the past six months. And the functions of all video cards are the same. For an in-depth look at the new features in the Radeon HD 5000 line, we recommend returning to our article on release of the Radeon HD 5870, because here we will only briefly consider them.

DirectX 11 support

Click on the picture to enlarge.

Before video cards GeForce GTX 470 and 480, which will be based on GF100, the ATI Radeon HD 5000 line remains the only one to support DirectX 11.

So far, the list of DirectX 11 games hasn't been rich. But with the gradual spread of this API, more and more developers will begin to submit games based on it. So far, we cannot say that DirectX 11 games have impressed us that much (most recently we tested DiRT 2). But we are looking at the upcoming Aliens Vs. Predator as a likely candidate to show gamers the beauty of DirectX 11-class hardware. Our expectations are largely based on the tessellation examples we've seen so far.

Playing on three Eyefinity monitors

Having played on three monitors with Eyefinity support, we can confidently say that the sensations turned out to be much stronger than we expected. Peripheral vision has a very tangible effect. However, it should be recognized that the Radeon HD 5000 line still has problems with support for games on multiple monitors: video cards for the mass market do not handle high resolutions very well, the game must support an unusual frame format, and one monitor will have to be connected via an interface DisplayPort (or via an active DisplayPort adapter for older displays). These shortcomings make it difficult for Eyefinity to spread, but we expect most of the problems to be resolved over time. As usual, the first "testers" of the new technology will have a hard time while AMD will optimize their drivers.

Introduction

In September 2008 (almost a year and a half ago), AMD surprised gamers on a budget with the release of its Radeon HD 4670 graphics card. The listed price was $80, which meant that the card was aimed at direct competition with the entry-level GeForce 9500 GT model, and the specifications of the 4670 were comparable with the flagship of the previous generation Radeon HD 3870.

The performance of the Radeon HD 4670 surpassed the competition. With 320 streaming cores, the Radeon HD 4670 has changed the balance of power in this price segment. The release of the 4670 meant that Nvidia had to create the GeForce 9600 GSO based on high-end GPUs, which are more expensive to manufacture. Not to mention that the prices for the GeForce 9600 GT had to be lowered.

Since its announcement, the Radeon HD 4670 has remained one of the best "budget" graphics cards on the market (and we regularly include it in our recommendations). In addition, it was the fastest reference card that did not require additional PCIe power for a long time until Nvidia introduced its GeForce GT 240, which then gave way ATI Radeon HD 5670 .

AMD really raised the bar for $80 graphics performance with the release of the Radeon HD 4670. As it happens, today AMD is introducing the successor to this acclaimed graphics card in the form of , which should also be priced at $80.

That said, we can't deny the appeal of the Nvidia GeForce 9600 GT to gamers looking for optimal performance on a single monitor at $80 ().

We'd also like to point out that in this price range, where there are a lot of different graphics cards, it's impressive to see the performance difference that a relatively small amount makes. The Radeon HD 5670 video card can already be found on the international market from $95 (). If the Radeon HD 5570 is sold at the recommended price or cheaper than the GeForce 9600 GT by $5 or $10, then the purchase will be quite attractive. But AMD now has a monopoly on DirectX 11 hardware (and will continue to do so for the foreseeable future, given the lack of details on nVidia's new mainstream products), so it's unlikely that prices will drop until the competitor introduces new models.

Today, the range of AMD graphics cards with DirectX 11 support is complete. The only thing missing in the 5000 line is perhaps an analogue of the Radeon HD 4650 model, but so far we have not seen a mention of such a video card. Below we have provided links to our reviews, which will allow you to get acquainted with a particular model in more detail.

• "ATI Radeon HD 5850: great performance at an affordable price ";
• "ATI Radeon HD 5770 and HD 5750: New DirectX 11 Graphics Cards for the Mass Market ";
• "ATI Radeon HD 5670: DirectX 11 Graphics Card for Thrifty Gamers ";
• "ATI Radeon HD 5450: DirectX 11 and Eyefinity graphics card at a low price ".

An impressive list, considering he's been up since September 2009.

We continue a series of articles about representatives of the budget class. AMD has always paid considerable attention to this market segment and the choice of its products is now richer than ever. In this article, we will look at the entire line of Radeon HD 5500 and Radeon HD 5600 using the example of several graphics video adapters with different type memory, and soon we will pay attention to fresh models of the six thousandth series.

All cards are based on the same GPU codenamed Redwood. In structure, this GPU resembles a quarter from Cypress. It has five SIMD cores with 16 universal superscalar stream processors, each of which executes five instructions per clock, i.e. in total we have 400 execution units.

Older HD 4600 series cards had a GPU with 320 compute units. The number of texture units in the new Redwood is slightly less - 20 versus 32 for the RV730. But this should be compensated by the higher frequencies of new video cards.

We also note that a full-fledged GPU with 400 stream processors and 20 TMUs is used in the Radeon HD 5670 and Radeon HD 5570, while the Radeon HD 5550 is content with 320 stream processors and 16 texture units. Compared to its relative from the old generation, the Radeon HD 4550, the new solution has noticeably gained power and is not that far removed from the HD 4650/4670. The number of rasterization units for all video cards is identical - eight.

The core clock speed of the older Radeon HD 5670 is 775 MHz (for comparison, the Radeon HD 4670 is 750 MHz), the Radeon HD 5570 is 650 MHz, and the younger Radeon HD 5570 is only 550 MHz. The memory bus is 128 bits, GDDR3 and GDDR5 are supported, although there are many versions on sale with slow DDR2.

You can compare the characteristics of the Radeon 5000th and 4000th series using the table below. We have given the official specifications in it, but the memory frequencies of manufactured products do not always correspond to them. So, initially AMD did not provide for a combination of the Radeon HD 5570 and Radeon HD 5550 with DDR2 memory, although such instances are produced by some of the company's partners. Yes, and the Radeon HD 5670 with GDDR3 instead of GDDR5 on board is also not uncommon in our stores. In this review, we will cover the issue of the impact of memory bandwidth on the example of the Radeon HD 5570 with GDDR3 and GDDR5, and make sure that the gain from faster memory is very significant.

Video adapter	Radeon HD 5670	Radeon HD 5570	Radeon HD 5550	Radeon HD 4670	Radeon HD 4650	Radeon HD 4550
Nucleus				RV730XT	RV730PRO	RV710
	627	627	627	514	514	242
Process technology, nm	40	40	40	55	55	55
Core area, sq. mm	104	104	104	145	145	73
	400	400	320	320	320	80
Number of texture blocks	20	20	16	32	32	8
Number of render units	8	8	8	8	8	4
Core frequency, MHz	775	650	550	750	600	600
Memory bus, bit	128	128	128	128	128	64
Memory type	GDDR5	DDR3 GDDR5	DDR3 GDDR5	GDDR3 GDDR4	DDR2 GDDR3	DDR2 GDDR3
Memory frequency, MHz	4000	1800 DDR3 3600/4000 GDDR5	1600/1800 DDR3 3600/4000 GDDR5	2000GDDR3, 2200GDDR4	800/1000 DDR2 1400GDDR3	800 DDR2 1600GDDR3
Memory size, MB	512/1024	512/1024	512/1024	512/1024	512/1024	256/512/1024
	11	11	11	10.1	10.1	10.1
Interface	PCI Express 2.1	PCI Express 2.1	PCI Express 2.1	PCI Express 2.0	PCI Express 2.0	PCI Express 2.0
	64	43	39	75	60	25

As for the size of the video buffer, Redwood-based cards are equipped with 512 or 1024 MB. Some manufacturers have also released 2GB versions. But such a large amount is always achieved through the use of slow memory, so such adapters are of minimal interest.

The new 40nm process technology has achieved very low power consumption. At maximum load, the Radeon HD 5670 consumes up to 64 watts (some sources give a figure of 61 watts), the younger versions are limited to 43 and 39 watts. At idle, the cards are super-economical solutions at all - the appetite of the HD 5670 is no more than 14 W, while the HD 5570 and HD 5555 are content with 10 W.

Supports HD-Video hardware acceleration, ATI Stream and ATI Eyefinity technologies. The latter allows you to connect up to three monitors to one card. Digital signal output is possible via HDMI 1.3a with support for Dolby TrueHD and DTS-HD Master Audio. Reference models are also equipped with DVI and D-Sub interfaces.

The Radeon HD 5670 looks like its predecessor, the Radeon HD 4670. It is cooled by a small turbine cooler.

The reference Radeon HD 5570 and Radeon HD 5550 are available in a low profile variant.

In this material, the following video cards will be considered:

The junior Radeon will be represented by only one model. This is a card manufactured by HIS, one of AMD's main partners. Moreover, it will be the most productive version with GDDR5 memory.

Delivery set includes:

• 10% off coupon for Battlefield: Bad Company 2 ;
• 50% discount coupon for ArcSoft TotalMedia Theater with ATI Stream support;
• disk with drivers;
• instructions.

The video card has a full-format performance. The prefix Silence in the name of the model indicates the use of passive cooling. The curved, bristling heatsink resembles a hedgehog and occupies two slots.

External interfaces are represented by a standard set of D-Sub, DVI and HDMI.

The aluminum radiator has a corrugated surface, which additionally allows to increase the dissipation area.

In general, for a 39-watt card, such a cooling system looks quite impressive.

The board has a simple design with single-phase core and memory power systems.

The Redwood GPU has become noticeably smaller than the RV730 - the area has decreased from 145 to 104 mm². The crystal is surrounded by a protective frame. There are also rubber washers under the mounting screws, which also prevent the heatsink base from tilting and damaging the GPU.

The card is equipped with 512 MB of video memory - four Hynix H5GQ1H24MFR T0C chips, which are designed for a frequency of 4 GHz.

The operating frequencies of the HIS HD 5550 Silence are 550/4000 MHz for the core and GDDR5 memory. But already when preparing this material, we found out that 550/3800 MHz are indicated on the official website of HIS. It may well be that our test sample, provided by the manufacturer, is slightly different, and the company released cards with a lower memory frequency at retail. In any case, 4000 MHz is the limit for the HD 5550 and this video card is one of the fastest among analogues.

Despite our expectations, the core temperature under gaming load was quite high. A 12-minute test of Crysis: Warhead at maximum settings at a resolution of 1680x1050 warmed up the GPU to 91 ° C (in an open case at 25 degrees indoors). And after a two-hour load, the temperature rose to 93 degrees Celsius. High, but it is worth remembering that such indicators were obtained in a completely passive mode without any airflow. Properly thought-out ventilation in the case will achieve a more acceptable temperature regime.

Taking into account the high frequencies of the older Radeon HD 5600, we can hope for a good overclocking of Redwood in the younger model as well. And in this regard, HIS really did not let us down - the GPU was stable at 740 MHz, and the memory frequency was raised to 4300 MHz.

Of course, with passive cooling it is better to forget about overclocking. We used additional airflow for higher frequencies: a low-speed 140 mm fan at 950 rpm was installed on the side of the case, 12 centimeters from the video card. Thanks to this, the temperature during acceleration did not rise above 69 ° C.

The ASUS graphics card comes in a small compact box.

Delivery set includes:

• optional low profile plug;
• disk with drivers;
• instructions.

This model is low profile. The D-Sub connector is connected with a detachable cable. The cooling system is quite modest.

A set of standard connectors: DVI, D-Sub and HDMI.

An aluminum radiator with curved "petals" is attached with simple plastic latches.

The tiny fan has a diameter of 40 mm (blade diameter 35 mm).

Let's take a look at the board design now. Although it is compact, it has a more complex circuitry compared to the HIS HD 5550. High-quality components are used, which are also used in more expensive ASUS products.

The GPU power system includes two phases under the control of the uP6205 controller.

The HD 5570 uses a full Redwood processor with all 400 active stream processors and 20 texture units.

The card is equipped with one gigabyte of video memory. Eight chips are soldered on both sides of the board. According to the marking, Samsung K4W1G1646E-HC12 chips are designed for an operating frequency of 1600 MHz. We saw exactly the same in Inno3D GeForce GT 430 and ASUS ENGT240/DI/1GD3/A .

The operating frequencies of ASUS EAH5570/DI/1GD3(LP) are 650/1600 MHz for core and memory. In idle, they drop to 157/400 MHz.

By the way, for some reason, this card displayed the memory frequency incorrectly in MSI Afterburner, although the program had no problems with other Radeons. In gaming mode, the core temperature did not exceed 62 ° C according to MSI Afterburner, GPU-Z showed 52 ° C. The first value is closer to the truth. The noise level was surprisingly low, although we expected the tiny fan to run at very high speeds.

The card showed good overclocking potential.

We were able to achieve stability at frequencies of 840/2000 MHz with the maximum speed of the blowing fan. In total, an increase in frequency of 29% for the core and 25% for memory was obtained.

Next up is the Radeon HD 5570 with GDDR5 memory.

The video card comes in a traditional rectangular box with a sword. Comes with the following:

• voucher with a 10% discount on Battlefield: Bad Company 2;
• disk with drivers;
• instructions.

A feature of this product is a massive cooling system from Arctic Cooling.

The rear panel has Display Port, HDMI and DVI connectors.

The cooler itself is well known to us. We saw a radiator similar in design to the GeForce GT 240 and a variety of budget Radeons. But in this case it is equipped with a large 90mm fan.

The board is a complete copy of the HIS HD 5550 Silence with exactly the same single-phase GPU power system.

Redwood GPU in person:

On board, HIS has only 512 MB of video memory - these are four Hynix H5GQ1H24MFR T0C chips.

Operating frequencies are 650/4000 MHz, although the manufacturer's website again indicates a lower memory frequency of 3800 MHz.

The temperature performance is just great - in the game mode, the core did not warm up above 44 ° C at 25 degrees in the room. And this is at a minimum noise level due to the large size of the blowing fan and its low speed.

But overclocking HIS HD 5570 iCooler IV did not please - only 720/4112 MHz for the core and memory.

The last participant in our testing is the older Radeon HD 5670 by ASUS.

The video card is made on red textolite. The usual cooler from Arctic Cooling is responsible for cooling. Some complaints are caused by a large plastic bar, which interferes with the flow of air. The cooling system ASUS ENGT240/DI/1GD3/A had the same design.

The design of the radiator is standard, it is only slightly smaller in width than that installed on the HIS HD 5570 iCooler IV. Fanned by 80mm fan. According to the manufacturer, the fan has a special dust-proof design, which extends its service life by 25%.

The core power supply system is built according to a two-phase scheme, the memory power supply system includes one phase.

Redwood GPU:

Eight memory chips with a total capacity of 1024 MB are soldered on both sides of the board. Elpida W1032BABG-50-F chips are used, which are rated at 5 GHz.

The frequency characteristics correspond to the standard specifications: the core runs at 775 MHz, the memory at 4 GHz.

Under gaming load, the graphics card did not warm up above 61 ° C in a room with a temperature of 25 degrees Celsius.

When overclocked, we managed to achieve stability at a core frequency of 930 MHz, and the memory was able to run at 4600 MHz.

A very good increase relative to the nominal value and excellent performance against the background of both Radeon HD 5570s. True, I had to turn the fan speed up to 100%, and its noise was already quite loud. Characteristics of the tested video cards

We will compare the reviewed Radeon with video cards from a recent article about the GeForce GT 430. Since the honor of NVIDIA will be defended by only two representatives, and one of them has lower frequencies (ASUS ENGT240/DI/1GD3/A), we decided to add to the test the results of the GeForce GT 240 at the frequencies recommended for the GDDR3 version of 550/1340 /2000 MHz. But the memory frequency of ASUS changed with a large step and it turned out to be impossible to set exactly 2000 MHz, so we stopped at frequencies of 550/1340/1974 MHz.

Video adapter	ASUS EAH5670/DI/ 1GD5	HIS HD 5570 iCooler IV	ASUS EAH5570/DI/ 1GD3(LP)	HIS HD 5550 Silence	Sapphire HD 4670 1G GDDR3 New Edition	Inno3D GeForce GT 430	ASUS ENGT240/DI/ 1GD3/A
Nucleus					RV730XT	GF108	GT215
Number of transistors, million pieces	627	627	627	627	514	585	727
Process technology, nm	40	40	40	40	55	40	40
Core area, sq. mm	104	104	104	104	145	116	139
Number of stream processors	400	400	400	320	320	96	96
Number of texture blocks	20	20	20	16	32	16	32
Number of render units	8	8	8	8	8	4	8
Core frequency, MHz	775	650	650	550	750	700	550
Shader domain frequency, MHz	128	128	128	128	750	1400	1340
Memory bus, bit	128	128	128	128	128	128	128
Memory type	GDDR5	GDDR5	GDDR3	GDDR5	DDR3	DDR3	GDDR3
Memory frequency, MHz	4000	4000	1600	4000	1594	1800	1580
Memory size, MB	1024	512	1024	512	1024	1024	1024
Supported version of DirectX	11	11	11	11	10.1	11	10.1
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	PCI-E 2.0
Declared maximum power consumption, W	64	43	43	39	<75	49	<69

test stand

The test bench configuration is as follows:

• processor: Core i7-965 EE (3, [email protected].91 GHz, BCLK 170 MHz);
• cooler: Thermalright Venomous X;
• motherboard: Gigabyte GA-X58A-UD3R (Intel X58 Express);
• memory: G.Skill F3-12800CL8T-6GBRM (3x2GB, [email protected] 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;
• Radeon driver: ATI Catalyst 11.6.
• GeForce driver: NVIDIA GeForce 275.33.

User Account Control, Superfetch, Windows Defender and interface visual effects were disabled in the operating system. The swap file is fixed at 1.5 GB, which was a prerequisite for running one of the applications.

Test Methodology

Testing was carried out at 1280x1024 resolution, in some applications and at 1680x1050. Tests in the new DirectX 11 are combined with tests in DirectX 9, which makes it possible to compare performance in simpler modes and evaluate the advantage of new products over older video cards. Games are listed in alphabetical order, divided into groups. Within each of them, the first are games in which testing was carried out only under the 9th version of Microsoft DirectX, and then applications with the results of using the capabilities of DirectX 10 or DirectX 11.

Borderlands

Game version 1.01. Graphics settings are maximum, all visual effects are enabled. The standard game test timedemo1_p was run seven times. Directly to start the test itself, Borderlands.exe is launched with the timedemo1_p -benchmark -seconds=60 -novsync parameter. In this test, there is a decent spread of results, so the test was run 5-6 times. Three testing modes are selected:

• resolution 1280x1024, high quality textures, vegetation and detail, dynamic shadows, SSAO, Bloom, Depth of Field and Flare Outs disabled (6 runs);
• resolution 1280x1024 at maximum image quality settings (5 passes);
• resolution 1680x1060 at maximum settings (5 runs).

Maximum quality of anisotropic filtering in all three modes. The game does not support anti-aliasing and it was not forcibly forced.

The minimum performance occurs at the start of the 3D application, when data is being reloaded, so the built-in test data on the minimum fps is inconsistent, greatly underestimated, and generally does not correspond to the real situation. To measure the minimum fps, we used the Fraps program. The average frame rate results are still based on the built-in gaming benchmark. Note that the real minimum of performance in the test falls on the moment when the camera shows a general plan of the gorge with a stele in the center, around which people and various monsters are fighting.

Bullet Storm

All settings in the High position. The SmoothFrameRate parameter, which limits the maximum fps to 62 frames, is deactivated in Engine.ini. For testing, a game screensaver was selected, which is activated after the user does nothing in the main menu. The test time is 207 seconds, it was run twice, the data was taken using Fraps.

Battlefield: Bad Company 2

Testing was carried out at the beginning of the "Heart of Darkness" mission in a real game, because in scripted cutscenes (of which there are plenty in Bad Company 2), the frame rate is noticeably higher and does not correspond to the real performance in the application. A jog was made along a strictly defined route - from the landing zone in the jungle to the car at the end of the village, nearby houses and objects were shot from a grenade launcher along the way. For more accurate results, the test was repeated four times in each mode. The image quality settings are maximum, the visibility range in the configurator is set to the highest value, the filtering is AF16x. DirectX 9 was tested at 1280x1024 and 1680x1050, DirectX 10 only uses 1280x1024.

Tests with the activation of DirectX 11 were not carried out for the reason that cards for this mode are already rather weak. And the difference in picture quality between DirectX 10 and DirectX 11 is minimal, but between DirectX 9 and DirectX 10 it is significant.

Crysis: Warhead

Game version 1.1. Testing was carried out using the Crysis Warhead Benchmarking Tool. The time of day in the test is set to 10:00, the Mainstream and Gamer settings profiles (low and medium) were used. For each mode, two cycles of the Ambush demo were run five times.

Crysis 2

Game performance (version 1.8) was measured using Fraps in the first location. The same short episode was replayed three times, including a small scripted video in which, through the sight, we observe a couple of C.E.L.L. soldiers shooting a prisoner. The settings were set to High and Maximum, which actually corresponds to low and medium quality, because image quality is not set below High.

The built-in performance test (Aspen track) was run four times. Since the game turned out to be not very demanding for our video cards in DirectX 9, and the activation of the simplest anti-aliasing mode had a minimal effect on performance, the tests at low resolutions were carried out at MSAA 2X. In total, we used three modes:

• resolution 1280x1024, maximum quality under DirectX 9, MSAA2x;
• resolution 1680x1060, maximum quality under DirectX 9;
• resolution 1280x1024, SSAO=High, for other settings the maximum quality is selected.

Just Cause 2

The built-in game performance test Concrete jungle ("Concrete jungle") was used. Repeated four times for each mode and video card. The tests were carried out in one resolution of 1280x1024 in two modes: with minimum quality settings and all additional visual effects disabled; with medium quality settings, high textures, low quality SSAO and additional effects enabled (high-quality shadows, soft particles, spotlight highlights, etc.).

mafia 2

The built-in performance test was used, which was run four times without taking into account the result at the first start of the test, because due to the reloading of data on all cards, “lags” are noticeable in the first seconds of the test, which greatly affects the final result. The data on the minimum fps was obtained using the Fraps utility. Resolution 1280x1024, two test modes: medium quality of details and shadows, environment blur disabled; high quality without activating the "anti-aliasing" parameter and APEX PhysX.

Metro 2033

The built-in performance test was used. Again, an ambiguous situation with a minimum fps. On all video cards, it reaches 8-10 frames, regardless of settings and resolution. Of course, the Frontline test demo is modeled to put the maximum load on the entire system, but in the real game it's not that bad and there are no such drawdowns. So we abandoned the minimum fps and show only the average result on the graph, which, however, is also not constant. To minimize the error, we had to increase the number of runs to 15 (three five-fold cycles). The tests were carried out at 1280x1024 under DirectX 9 using the Low, Medium, High settings profiles (low, medium and high quality).

A fresh game with DirectX 11 support. The game's "destruction mode" was launched, a run was made through the first level without massive destruction and shooting, only at the end two container barrels exploded. The data was taken using Fraps. For each mode and map, the test was repeated four times to reduce the error. The test modes are as follows:

• resolution 1600x900, DirectX 9, MSAA4x;
• resolution 1360x768, DirectX 11;
• resolution 1600x900, DirectX 11.

The game defaults to 16:9 widescreen resolutions, which is why we used them. In the last two cases, the "background shading" parameter was switched to a special mode "DX10 and higher". Anisotropic filtering 16x is enabled in all cases. The rest of the settings (textures, detail) were set to Low. Why did we use low settings? Yes, because the harsh domestic localizers distinguished themselves so much that they translated High as “low”, and, accordingly, Low as “High”. We were sure that we were using high quality settings, and already at the stage of preparing this article for publication, such circumstances became clear.

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

The game has modest system requirements when rendering in DirectX 9, so at 1280x1024 the tests were performed with MSAA 4x anti-aliasing. There are four modes in total:

• resolution 1280x1024, maximum quality (High) under DirectX 9, MSAA4x;
• resolution 1680x1050, High settings, DirectX 9;
• resolution 1280x1024, High, DirectX 11, tessellation enabled;
• resolution 1680x1050, High, DirectX 11, tessellation enabled.

Another new game on our list. It has a built-in performance test, which we used. It was run four times in two modes - the High settings profile with additionally disabled anti-aliasing and the Ultra settings at a resolution of 1280x1024.

One of the best RPGs of recent times and without a doubt the most beautiful game of the year so far. Excellent detailing and textures, gorgeous lighting and visual effects - and all this when using DirectX 9. For our video cards, we naturally did not use the most beautiful “outrageous quality” mode, but used the following settings:

• resolution 1280x720, settings profile Low, "glow" is additionally disabled;
• 1280x720 resolution, Medium settings profile, optionally disabled SSAO;
• resolution 1280x720, High settings profile, additionally disabled motion blur effect (Motion Blur).

Also, in all cases, vertical synchronization was disabled, which is activated by default at medium and high quality. In the settings, the resolution was set to 1280x1024, but in any case, the game produces a widescreen type image with a stripe at the bottom and top, so the actual resolution of the working area is 1280x720.

Latest 3DMark with DirectX 11 support. Tests performed with the Performance settings profile (1280x720 resolution).
Test results

Borderlands

At medium quality settings, the Radeon HD 5570 GDDR5 is slightly inferior to the GeForce GT 430, but with additional visual effects and shadows turned on, it outperforms the rival by 1.7% in average fps. The Radeon HD 5570 GDDR3 is not only inferior to the NVIDIA card, but also turns out to be slower than the younger Radeon HD 5550 with fast memory and even the old Radeon HD 4670. When overclocked, the ASUS card is not able to reach the performance level of the HD 5570 GDDR5 and GeForce GT 430. The older Radeon HD 5670 is inferior GeForce GT 240 with a memory frequency of 1974 MHz, which is close to the standard specifications. But with a lower memory frequency, this NVIDIA card is slightly inferior to the older AMD representative in terms of average fps, but retains a tiny advantage in the minimum. The Radeon HD 5550 is not inferior to the Radeon HD 4670 in terms of average, but loses 9-12% in terms of the minimum.

High resolution at maximum quality settings is beyond the power of all video cards. But when overclocked, the GeForce GT 240 and Radeon HD 5670 show more or less acceptable results. In nominal terms, there is also a minimal difference in performance between them. The GT 430 is outperformed by the Radeon HD 5570 GDDR5, but confidently holds the lead over the GDDR3 card.

Bullet Storm

Like Borderlands, this game is based on Unreal Engine 3. But the results are significantly different from what we saw above. The Radeon HD 5670 and Radeon HD 5570 with GDDR5 memory leave all other participants behind by a significant margin. The Radeon HD 5550 GDDR5 outperforms the Radeon HD 5550 GDDR3, GeForce GT 430 and is almost as good as the GeForce GT 240 at 550/1340/1580 MHz in the simpler mode, bypassing all NVIDIA cards when anti-aliasing is enabled. However, for the last mode, the performance of the card is still not enough. On the Radeon HD 5670 and older Radeon HD 5570, it is quite possible to play at AA4x. The Radeon HD 5570 with GDDR3 memory has a very low result, it is 40-45% slower than its counterpart.

Battlefield: Bad Company 2

First, let's look at the results in DirectX 9.

At low resolutions, all cards showed good performance. But in fact, according to subjective feelings, on some, even at an average frame rate of 40 fps, there was no complete comfort - there was no smoothness during movements, there were “jerks”. In a single player campaign, this will not create any special problems, but you will not play much in multiplayer. This effect was not noticeable on the two older Radeons and on all GeForces. So you should pay attention to cards that give out 40 frames and above, and the minimum fps is close to 30 frames.

But let's move on to the comparison. The leader is the Radeon HD 5670. The closest competitor is only one percent behind in terms of average fps, but loses almost 12% in terms of the minimum. The Radeon HD 5570 with GDDR5 memory is about on par with the ASUS GeForce GT 240 with slower memory. This balance of forces is not preserved during overclocking, and in this mode the top version of the Radeon HD 5570 only slightly outperforms the overclocked GeForce GT 430. By the way, the last card is nominally slightly more productive than the Radeon HD 5550 and Radeon HD 4670, but with the Radeon HD 5570 GDDR3 they have an ambiguous situation - the average fps is higher for NVIDIA, the minimum for AMD (only the GT 430 does not have jerks).

The older Radeon and GeForce GT 240 at 550/1340/1974 MHz handle higher resolutions. When overclocked, acceptable results are demonstrated by the Radeon HD 5570 GDDR3 and the GeForce GT 430.

And finally, let's look at the results when rendering in DirectX 10. As we already wrote in the methodology, when switching from DirectX 9 to DirectX 10, the difference in image quality is striking, but between DirectX 10 and DirectX 11, you already need to look for differences with a magnifying glass. It is curious that in this mode the picture is smoother even at 30-35 fps, without jerks, which we talked about above. Only the Radeon HD 5670 can adequately handle this mode. The GeForce GT 240 with 2 GHz memory is 10% behind the leader, probably, the version with GDDR5 would be no worse than the top Redwood. However, ASUS ENGT240/DI/1GD3/A can also be played in this mode during overclocking. But the power of the other cards, even with an increase in frequencies, is not enough. Overclocked Radeon HD 5570s show 33-34 fps, but their minimum figure is only 25 frames. Not a lot, but for an undemanding player it will also work, although it will be far from comfortable in multiplayer. It is noteworthy that in DirectX 10 the difference between these two cards with different memory is no longer so critical, ASUS is only 12-13% behind HIS in terms of value.

Crysis: Warhead

The undisputed leader is the Radeon HD 5670. It is 15-18% behind the Radeon HD 5570 GDDR5, which is 3.5% faster than the GeForce GT 240 (we are talking about a card with 2 GHz memory). The rest of the AMD cards are inferior to the even "slower" version of the GeForce GT 240 with 1580 MHz memory. The Radeon HD 5550 outperforms the GeForce GT 430 by 0.7% in average fps, but is almost 4% behind in the minimum. The Radeon HD 5570 GDDR3 is inferior to its counterpart with GDDR5 memory by about 21%. The overclocked younger Redwoods outperform the GeForce GT 240 without any problems. Although the latter also shows a good increase during overclocking, it still falls short of the level of the Radeon HD 5670.

In a heavier mode, only one Radeon HD 5670 showed a more or less acceptable result, although its minimum fps of 23 frames is too small for a comfortable game. But this is easily corrected by overclocking. But the rest of the cards, even at higher frequencies, cannot demonstrate a sufficient level of performance. In such a difficult mode for test participants, the gap between the two versions of the Radeon HD 5570 widens - now the card with GDDR5 has a 30% advantage. The Radeon HD 5570 GDDR3 is even slightly behind the Radeon HD 4670, and shows a minimal advantage over the fastest Radeon HD 5550 with GDDR5 memory.

Crysis 2

It comes as a surprise that the last place in the first test mode is occupied by the Radeon HD 5570 GDDR3. The brother with 4 GHz memory is 30-34% more productive, slightly faster than the GeForce GT 240 and 15-17% slower than the leader Radeon HD 5670 GDDR5. In a simpler mode, two older Radeons and a GeForce GT 240 demonstrate a comfortable fps level. With overclocking, all cards already show acceptable results. At higher image quality settings, no one except the main trinity can exceed the 30 frame mark even when overclocked.

Let's consider the test results at the maximum possible settings under DirectX 9.

The Radeon HD 5670 GDDR5 and Radeon HD 5570 GDDR5 are in the lead, and the rest of the cards show similar results. The older GeForce GT 240 actually ranks third, but the Radeon HD 5570 GDDR3 is only 6% behind. When overclocked, all Redwood bypass the overclocked GeForce GT 430 and GeForce GT 240.

When moving to high resolution, the balance of forces does not change much. Only the Radeon HD 5550 is losing ground and moving to the place of an outsider, this card is worse at high resolution than at lower resolution. For other cards, on the contrary, fps increases slightly. The difference between Radeon HD 5570 GDDR5 and HD 5570 GDDR3 is 13-16% and about 20% with anti-aliasing enabled at 1280x1024.

In general, the performance of all video cards is sufficient for this game. Only at a resolution of 1680x1050, some cannot provide a comfortable minimum fps, but overclocking saves them.

In DirectX 11, only one video card can provide comfortable fps - this is the older Radeon HD 5670. Although not very demanding users will be satisfied with the overclocked Radeon HD 5570 with the GeForce GT 430. By the way, finally, the Radeon HD 5570 GDDR3 overtakes its brother with increased frequencies thanks to the work of the GPU at 840 MHz. The overclocked GeForce GT 430 is more productive than the Radeon HD 5570 GDDR5 running at nominal frequencies.

Just Cause 2

Excellent results for all AMD video cards and complete defeat of NVIDIA solutions. The three older Radeons demonstrate acceptable results both in terms of average and minimum fps. Although if we remember that the Concrete Jungle test is very difficult, then we can say that even a Radeon HD 5550 or GeForce GT 430 is enough for an undemanding player.

At medium quality settings, everything looks different. The Radeon HD 5550 is already inferior to the GeForce GT 430, and the Radeon HD 5550 GDDR3 is weaker than the GeForce GT 240. However, the results are still low and only the Radeon HD 5670 can play normally at these settings. mafia 2

In this game, NVIDIA representatives again look very weak. At low and medium settings, they can only compete with the old Radeon HD 4670. And only at high quality does the GeForce GT 240 show identical results to the Radeon HD 5550 GDDR5 and Radeon HD 5570 GDDR3. Memory bandwidth has a significant impact on in-game performance. In the first two modes, the Radeon HD 5570 GDDR3 lags significantly behind the Radeon HD 5550 GDDR5 with fewer compute units. The older Radeon HD 5570 outperforms its counterpart by 30% in High mode and by as much as 44% at minimum quality. The advantage of the Radeon HD 5550 over the ASUS ENGT240/DI/1GD3/A remains even when overclocked, despite the advantage of 100 MHz in GPU frequency.

Testing in DirectX 9 was carried out in one mode.

Again, AMD's budget cards are in the lead. The GeForce GT 430 and GeForce GT 240 compete only with the Radeon HD 5550. The difference between the two versions of the Radeon HD 5570 with different memory is 18% in average fps and 23% in minimum. Between the older Radeon HD 5570 and Radeon HD 5670, there is a tiny difference in the minimum indicator and more than 15% in the average - it seems that even fast GDDR5 memory limits Redwood's potential in this game a little.

In DirectX 11, the GeForce GT 430 cannot compete with the Radeon HD 5550, losing 6-7% of performance. When overclocked, this representative of NVIDIA almost catches up with the nominal Radeon HD 5570 GDDR3. For a resolution of 1360x768, any video card is sufficient, and the GeForce GT 430 and Radeon HD 5550 cannot cope with 1600x900 at nominal frequencies.

But let us remind you that the tests were carried out at low settings, for which many thanks to the localizers from Buka, who guessed to translate High into the menu as “Low”. So the practicality of the results obtained is highly controversial. It remains to be seen how the situation would have developed with high textures and detail, and at least one video card could have demonstrated acceptable fps in DirectX 11. The ratio between AMD and NVIDIA cards could also be different.

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

A small rehabilitation of NVIDIA video cards. They look especially good in high resolution - here the older GeForce GT 240 even bypasses the Radeon HD 5670. In both modes, the GeForce GT 430 turns out to be more productive than the Radeon HD 5550 and Radeon HD 5570 GDDR3, and at a resolution of 1680x1050 it and the Radeon HD 5570 GDDR5 come on the heels. With overclocking, this NVIDIA newcomer is second only to the overclocked Radeon HD 5670. The latter retains its leadership at higher frequencies at 1280x1024 with AA4x, but it fails to snatch victory from the GeForce GT 240 in high resolution. The Radeon HD 5570 GDDR3 demonstrates a modest advantage over the Radeon HD 5550 GDDR5 and is inferior to the HIS HD 5570 iCooler IV by 17 to 23% (maximum with anti-aliasing). When overclocked, these comrades are about on a par, although HIS retains an advantage of one frame.

The Radeon HD 5550 is inferior to the GeForce GT 430, and even the Radeon HD 5570 GDDR3 has a slight 4% advantage over this competitor. The overclocked GeForce encroaches on the position of the Radeon HD 5570 GDDR5, which works at par. None of the participants can come close to the results of the leader Radeon HD 5670.

Another new game in our testing. The latest expansion for Warhammer 40,000: Dawn of War 2.

A very low minimum fps immediately catches your eye. This dramatic drop in performance occurs in the most intense battle scenes with lots of explosions. Such drawdowns will create noticeable discomfort when manipulating the map during moments of particularly violent skirmishes. Although chasing 30 frames is not necessary here, 10-15 frames is too little. For an undemanding player, a minimum value of 20 fps and above is quite enough. From this point of view, it will still be possible to play normally only on the Radeon HD 5670 and GeForce GT 240, and not at the maximum quality settings. In Ultra mode, the leaders are the Radeon HD 5670 and Radeon HD 5570 with GDDR5 memory, but the minimum fps barely exceeds 11 fps. The outsider is the GeForce GT 430. The Radeon HD 5570 GDDR3 in High mode is closer in performance to its brother than the Radeon HD 5550, but when the settings are increased, it outperforms the younger Redwood by only 3%.

Witcher 2: The Assassins of Kings

It's time to evaluate the performance of video cards in the hottest game of the season - The Witcher 2.

As you can see, playing at minimum and medium settings does not require an outrageous system with super-powerful video cards. And the difference between the modes turned out to be small - the performance drop for video cards is no more than 10%. For such settings, the Radeon HD 5670, Radeon HD 5570 and GeForce GT 240 with normal memory will be enough. Yes, and the rest of the cards with an increase in frequencies give an acceptable "frame rate". Only one old Radeon HD 4670 lags behind, demonstrating consistently low results. The Radeon HD 5550 GDDR5 and Radeon HD 5570 GDDR3 look like cards of the same level, only when overclocked, the second one slightly outperforms the younger Redwood. Both of these Radeons are more productive than the GeForce GT 430, and they are inferior to the GeForce GT 240 by 4-10%.

High image quality settings confuse participants. And only the Radeon HD 5670 can withstand this test, demonstrating a more or less acceptable result. With overclocking, the card will suit even the most demanding players, because even the minimum fps exceeds 30 frames. The ratio between other cards does not change, but none of them can match the leader even when overclocked.

Last final test. The most productive version of the Radeon HD 5550 with 4000 MHz GDDR5 memory is 4.5% behind the regular GeForce GT 430. The Radeon HD 5570 GDDR3 is 15.6% faster than the younger Redwood and 15% slower than the older HD 5570 with fast memory. The Radeon HD 5670 is 16.5% more productive than the Radeon HD 5570. The difference between junior and senior Redwood reaches 55%.

conclusions

It's time to take stock. The 5,000th series of budget Radeon video cards has become a qualitative evolutionary leap in its segment. The older Radeon HD 5670 is more productive than its predecessor Radeon HD 4670 by an average of 30-50%, and sometimes the gap reaches a huge 75-90% (Bulletstorm, Witcher 2, Metro 2033). However, we must make allowances for the fact that not quite a full-fledged old man with an underestimated memory frequency took part in our testing. But still, the advantage of the older Redwood is striking. The 2GHz GeForce GT 240 is inferior to it in all applications except Borderlands. Between these cards, the difference in performance is already from 10 to 40%, and it is clear that an NVIDIA card cannot compensate for such a lag with overclocking. Price positioning matches this level of performance, and for the Radeon HD 5670 GDDR5 you will have to shell out an average of $10 more than for the GeForce GT 240 GDDR3.

But these $10 are more than justified, because you get a more modern, more productive solution. And in some games you can even use DirectX 11. In Metro 2033 or Crysis 2 you will not play at maximum settings, but in DIRT 3 or H.A.W.X. 2 card can pull out not very high resolutions in DirectX 11. The overclocking capabilities of this adapter are also good, although we cannot speak for all the Radeon HD 5670. But the ASUS EAH5670/DI/1GD5 model we reviewed really did not disappoint in this regard - a 20% increase in GPU frequency, and the memory overclocking is the best among all other participants with GDDR5. So if you're looking to buy a Radeon HD 5670, then the EAH5670/DI/1GD5 might be a good choice.

There are Radeon HD 5670 models with GDDR3 memory on the market, but we would not recommend saving a couple of dollars on such a purchase. After all, on the example of ASUS EAH5570/DI/1GD3(LP) and HIS HD 5570 iCooler IV cards, we clearly saw how slow memory limits Redwood's computing potential. The Radeon HD 5570 GDDR5 is 12-40% more productive than the Radeon HD 5570 with GDDR3, while the difference between the Radeon HD 5670 and the Radeon HD 5570 GDDR5 is on average about 15%, sometimes less. Slow memory immediately takes the Radeon HD 5570 down a notch and makes the card a competitor for the Radeon HD 5550 GDDR5 and GeForce GT 430. But some manufacturers still release versions with DDR2, and one can only guess how much the performance of such hybrids will decrease. But judging by the results of our Radeon HD 5550 GDDR3, a card with DDR2 would certainly be an outsider in the test.

As for the specific model we reviewed, the HIS HD 5570 iCooler IV was quiet and very cold at just 44°C under load! But overclocking the card upset a little. The Radeon HD 5570 differs from the older Redwood only in operating frequencies, so there is always hope of getting an analogue of the older card for less money. But with HIS this will not work - a weak single-phase power supply system and possibly very low supply voltages did not allow the core to be overclocked above 740 MHz. The memory turned out to be completely unresponsive to overclocking. In this regard, the ASUS EAH5570/DI/1GD3(LP) turned out to be better. This small low-profile card already has a two-phase power system, but a weak cooling system (although it is quite enough for the nominal mode). The GPU overclocking result from ASUS is 100 MHz higher, and this is pretty good for such a "baby". However, this advantage does not make the weather, HIS is still faster thanks to GDDR5 memory.

The HIS HD 5550 Silence graphics card demonstrated a good level of performance. Despite using a truncated GPU with 320 execution units, the card is almost as good as the Radeon HD 5570 GDDR3 and Radeon HD 4670. Such results are largely due to the high bandwidth of GDDR5 memory running at 4 GHz. On the example of the same Radeon HD 5570, it can be seen that in the case of using a slower GDDR3 with frequencies below 2 GHz, the Radeon HD 5550 could well lose a third of the performance. In fact, thanks to the memory alone, the card delivers near-HD 4670 performance levels with higher core clocks and more texture units. Among NVIDIA's competitors, the closest rival is the GeForce GT 430. According to the test results, they have an approximate parity - somewhere a little more productive than GeForce, and somewhere a couple of percent pulls out Radeon. The competitor is faster in DirectX 10 (Battlefield: Bad Company 2, Just Cause 2 at medium settings) and DirectX 11 (DiRT 3, H.A.W.X. 2), so perhaps the GeForce GT 430 will be a little more interesting as a purchase. Although the main criterion for choosing will be the cost of the final products. Well, the overclocking potential should not be discounted. The HIS HD 5550 Silence has an excellent one and allows you to achieve a performance increase more than overclocking the GeForce GT 430, which makes buying the latter already of little interest. If you choose between the Radeon HD 5550 GDDR3 and the full-fledged GeForce GT 430, you won't have to think long, and the second one will already be better in all respects.

A few words should also be said about the passive cooling system of the HIS HD 5550 Silence. Without additional airflow, operating temperatures are very high, so you should take care of the organization of proper ventilation in the case. The ideal option is a fan on the side wall of the case, then you can forget about overheating and overclock the card to the limit without any problems.

To complete the picture, our extensive testing lacks the new Radeon HD 6670 and HD 6570. After all, they are AMD's budget leaders in the sub-$100 price range right now. We have not forgotten about them, just a separate article will be devoted to them, which will be released in the near future.

Testing equipment was provided by the following companies:

• 1-Incom - G.Skill F3-12800CL8T-6GBRM memory;
• DCLink - video cards ASUS ENGT240/DI/1GD3/A and EAH5670/DI/1GD5, Sapphire HD 4670 1G New Edition;
• Gigabyte - Gigabyte GA-X58A-UD3R motherboard;
• HIS - HIS HD 5550 Silence and HD 5570 iCooler IV video cards;
• Inno3D - Inno3D GeForce GT 430 video card;
• Intel - Intel Core i7-965 EE processor;
• PCShop - ASUS EAH5570/DI/1GD3(LP) video card;
• Syntex - Seasonic SS-850HT (S12D-850) power supply;
• Thermalright - Thermalright Venomous X.

It's no secret that when it comes to gaming-class 3D accelerators, many readers immediately associate with top products that can give playable results at the highest graphics settings in games, and even with anti-aliasing and anisotropic filtering.

And the average level of such accelerators is now showing excellent results for its price niche. But what to do with those that are very cheap, and therefore not so strong in three-dimensional graphics?

If adult uncles buy expensive and large cars, then small children also want to, and they have substitutes - toys. That's just about such "toy" video cards, we'll talk. They can do everything the same as the "real, adult" counterparts, but only very slowly. Therefore, high resolutions are no longer for them. Yes, and the settings in games will have to be reduced from maximum to medium, and sometimes even lower.

Quite recently, we studied a new low-end product from AMD (ATI) - Radeon HD 5670, and today we will study an even weaker product in terms of three-dimensional performance - Radeon HD 5570, which differs from 5670 in lower frequencies and a reduced number of texture modules and ROPs. Well, it comes with cheaper GDDR3 memory.

Of course, the lower the price, the lower the performance in 3D, and hence the lower demand for such accelerators specifically for 3D games. Usually, cards of the level of 60-70 US dollars are bought just to have a picture on the monitor, movies to play well, as well as various flash products for modern Internet content. And if they play three-dimensional games, they are very weak and primitive. Of course, you can play more modern complex games, but there you will have to drastically reduce the quality settings and reduce the screen resolution.

Unfortunately, we don't have a full-fledged method for testing low-end accelerators yet, so today we'll use a method familiar to all readers, based on tests at maximum quality in games (benchmarks) and at resolutions starting from 1280x1024.

Of course, we should only be interested in this resolution, because acceptable playability can only be obtained in it. The rest are just for academic interest. And even sometimes at 1280x1024 we got extremely poor results. But first things first.

HIS Radeon HD 5570 1024MB PCI-E

GPU: Radeon HD 5570 (Juniper) Interface: PCI Express x16 GPU frequencies (ROPs/Shaders): 650/650 MHz (nominal - 650/650 MHz) Memory frequencies (physical (effective)): 830 (1660) MHz (nominal - 830 (1660) MHz) Memory exchange bus width: 128bit Number of vertex processors: - Number of pixel processors: - Number of universal processors: 400 Number of texture processors: 20 (BLF/TLF) Number of ROPs: 8 Dimensions: 170x75x30 mm (the last value is the maximum thickness of the video card) Textolite color: blue RAMDACs/TMDS: integrated into the GPU Output jacks: 1xDVI (Dual-Link/HDMI), 1xVGA, 1xHDMI VIVO: No TV out: not bred Support for multiprocessing: CrossFire (Software).

We conducted a study of the temperature regime using the RivaTuner utility (author A.Nikolaichuk AKA Unwinder) and obtained the following results:

HIS Radeon HD 5570 1024MB PCI-E

As we can see, the heating is very low, so even such a simple cooler produces high efficiency. And once again you will think why there is no passive CO.

Now about the configuration.

The basic delivery set should include: a user manual, a disk with drivers and utilities. Since TV-outs are no longer installed, therefore, the corresponding adapters are not needed, and also these cards have all three output sockets, therefore, adapters are also not required. Below we will show what is offered to the card in addition.

Package.

Installation and drivers

Test bench configuration:

• Computer based on Intel Core I7 CPU 975 (Socket 1366)
  • processor Intel Core I7 CPU 975 (3340 MHz);
  • Asus P6T Deluxe motherboard based on Intel X58 chipset;
  • RAM 6 GB DDR3 SDRAM Corsair 1600MHz;
  • hard drive WD Caviar SE WD1600JD 160GB SATA;
  • Tagan TG900-BZ 900W power supply.
• operating room Windows system 7 32bit; DirectX11;
• monitor Dell 3007WFP (30");
• ATI drivers version Catalyst 10.4; Nvidia version 197.41/197.45.

vsync is disabled.

Test Results: Performance Comparison

As a tool we used:

• Far Cry 2 (Ubisoft) - DirectX 10.0, shaders 4.0 (HDR), the utility from the game bundle was used for testing (Middle level). All settings are set to maximum quality. We thank Nvidia for providing a licensed product.
• Unigine Tropics Benchmark 1.3 (Unigine) - DirectX 10.0, . Test settings - High.

  Unigine and personally Alexandra Zapryagaeva

• 3DMark Vantage 1.02 (FutureMark) - DirectX 10.0, shaders 4.0, multitexturing, test settings - Extreme
• CRYSIS 1.2 (Crytek/EA), DirectX 10.0, shaders 4.0, (battery and demo for launch, test settings - Very High, RESCUE level is used). We thank Nvidia for providing a licensed product.
• CRYSIS Warhead (Crytek/EA), DirectX 10.0, shaders 4.0, (battery and demo for launch, test settings - Very High, CARGO level is used). We thank Nvidia for providing a licensed product.
• Unigine Heaven Benchmark 2.0 (Unigine) - DirectX 10.0, . Test settings - High.

  We would like to thank the team Unigine and personally Alexandra Zapryagaeva for help in setting up the benchmark

• Unigine Heaven Benchmark 2.0 (Unigine) - DirectX 11.0, . Test settings - High.

  We would like to thank the team Unigine and personally Alexandra Zapryagaeva for help in setting up the benchmark

• Colin McRae: DiRT2 (Codemasters) - DirectX 11.0, testing settings - Ultra High (run benchmark dirt2.exe -benchmark example_benchmark.xml). We thank AMD for providing the licensed product.
• Warhammer 40.000: Dawn Of War 2 (Relic Entertainment/THQ) - DirectX 10.0, test settings - Super High (launch the benchmark in the game itself in the settings). We thank Nvidia for providing a licensed product.
• Just Cause 2 (Avalanche Studios/Eidos Interactive) - DirectX 10.0, test settings - Super High (launch the benchmark in the game itself in the settings). We thank Nvidia for providing a licensed product.

Attention! About a set of test tools!

In the updated and supplemented material on the FRAPS testing tool, we clearly showed how inaccurate and unreliable the tests obtained using this utility are, and testers have no other tool than benchmarks built into games.

Therefore, we believe that even if the set of test games is not so large, all tests will be transparent, accurate, and, most importantly, perfectly reflect the picture of the ratio of accelerators.

Video card performance

Important! Having come to the decision to independently choose a three-dimensional accelerator, that is, a video card in his computer, the user must be aware that he is changing one of the main components of the operation of his system unit, which may require additional settings for better performance or the inclusion of a number of quality functions. It is not a final consumer product, but only one link among all computer components. And so the user must understand that in order to get the most out of the new video card, he will have to learn some basics of three-dimensional graphics. And graphics in general. If he does not want to do this, then you should not start an independent upgrade in this regard. It is better to purchase ready-made system units with already configured software (and it will also be provided with technical support from the assembly company of such a system unit) or game consoles where you do not need to configure anything - everything you need is already included in the game itself.

Readers who are well versed in 3D graphics will be able to figure it out by looking at the diagrams below and draw their own conclusions. And for beginners and those who have just taken up the issue of choosing a video card, we will make some explanations.

First, it makes sense to get acquainted with our brief guides to the families of modern video cards and processors, on the basis of which they are produced. It should be noted the frequency of work, support for modern technologies (shaders), as well as pipeline architecture.

ATI Radeon Reference

Nvidia Geforce Reference

Secondly, in the section, our reader, who has just encountered the problem of choosing a video card and is confused, can familiarize himself with the basics of three-dimensional graphics (they will still be needed, because when starting the game and entering its settings, the user will encounter such concepts, like textures, lighting, etc.) and base materials for new products. There are only two companies that produce the now popular graphics processors: AMD (the ATI division is engaged in graphics) and Nvidia (there are also Matrox, S3, but their share in discrete graphics today is less than 1%, and therefore they can be ignored). Therefore, the bulk of the information is divided into two parts. It comes out monthly, where, as it were, all comparisons of various cards for different price sectors are reduced into one.

Thirdly, let's look at the tests of the cards we're considering today.

• 1. Unigine Tropics Benchmark

  

• 2. Far Cry 2
  • All permissions on one page, No AA, No AF
  • All resolutions on one page, AA 4x + AF 16x

  

• 3. Unigine Heaven Benchmark DirectX 11.0
  • All permissions on one page, No AA, No AF
  • All resolutions on one page, AA 4x + AF 16x

  

• 4. CRYSIS, RESCUE
  • All permissions on one page, No AA, No AF
  • All resolutions on one page, AA 4x + AF 16x

  

• 5. CRYSIS WARHEAD
  • All permissions on one page, No AA, No AF
  • All resolutions on one page, AA 4x + AF 16x

  

• 6. Unigine Heaven Benchmark DirectX 10.0
  • All permissions on one page, No AA, No AF
  • All resolutions on one page, AA 4x + AF 16x

  

• 7. 3DMark Vantage Graphics MARKS