Radeon R9 290 Review | Revealing Hawaii Behavior

The opinion about the new product, in the past month, was, to put it mildly, ambiguous. On the one hand, we have a new GPU based on an already familiar architecture, but equipped with new technologies and generally outperforming GeForce GTX 780 and GeForce GTX Titan... On the other hand, this GPU has a high power consumption, it is designed to operate at 95 ° C and is cooled by a very noisy fan, if, of course, it is allowed to be noisy.

So, two groups stood out: those who liked the very fast video card at an attractive price, and those who were not impressed with the new GPU from AMD, which overtakes the flagship Nvidia chip (which is already eight months old).

But no matter which side you choose, you have to agree that better performance for a lower price is a good thing. Just look at the implications: Nvidia cut the price GeForce GTX 770 up to attractive $ 330, and GeForce GTX 780- up to $ 500. We also know that the GeForce GTX 780 Ti will sell for $ 700.

Today, more and more lucrative offers appear. AMD recently introduced Radeon R9 290 based on the same Hawaii GPU.

Hawaii "cut back"

Radeon R9 290 Is a derivative product, that is, its characteristics will not differ much from. As we already know, the Hawaii processor features 6.2 billion transistors built in a 28nm manufacturing process. But instead of 44 computational units, AMD leaves 40, lowering the number of shader cores to 2560 (from 2816). As a side effect, the number of texture units is reduced from 176 to 160. Although AMD does not give specific information on the four disabled compute units, company representatives say that they are disabled in order to maintain a stable transition from one card to another.

Something from our previous review needs to be clarified: Hawaii does not offer double precision calculations at one-fourth the frequency that Tahiti does. Instead, AMD is dropping DP performance by one-eighth of the FP32 bandwidth while maintaining the computing power for professional FirePro cards and following in Nvidia's footsteps. Thus, the peak speed of floating point computation y Radeon R9 290 reaches 4.84 TFLOPS, and DP frequency reaches 606 GFLOPS.

	Radeon R9 290X	Radeon R9 290	Radeon R9 280X	GeForce GTX Titan	GeForce GTX 780
Process technology, nm	28	28	28	28	28
Number of transistors, bln.	6,2	6,2	4,3	7,1	7,1
GPU frequency	1 GHz	up to 947 MHz	1 GHz	836 MHz	863 MHz
Number of shader cores	2816	2560	2048	2688	2304
Performance FP32, TFlops	5,6	4,8	4,1	4,5	4
Number of texture units	176	160	128	224	192
Texture filling rate, Gtex / s	176	152	128	188	166
Number of ROP units	64	64	32	48	48
Pixel fill rate, Gpix / s	64	61	32	40	41
Memory bus, bit	512	512	384	384	384
Memory size, GB	4 GDDR5	4 GDDR5	3 GDDR5	6 GDDR5	3 GDDR5
Memory data transfer rate, Gbps	5	5	6	6	6
Memory bandwidth, GB / s	320	320	288	288	288
Power consumption of the video card, W	250 (declared)	250 (declared)	250	250	250

The rest of Hawaii's vital characteristics, oddly enough, remained intact. The geometry block in each of the four shader clusters produces the same number of primitive elements per clock cycle. In addition, each shader cluster is equipped with four rendering pipelines, providing up to 64 pixels per clock to the GPU. The cumulative 512-bit memory bus has been preserved. The video memory configuration has remained unchanged - 4 GB GDDR5 RAM with a frequency of 1250 MHz.

With so many similarities between and Radeon R9 290, in the younger model of the AMD card, in addition to the number of shader cores, it also lowers the maximum frequency so that there is a difference in speed between the cards. reaches a frequency of 1 GHz, and Radeon R9 290- 947 MHz.

Clock Fanning: Marketing Goes (Too) Wild

However, let's talk about clock speeds in more detail, as they caused a lot of controversy in this review. In particular, it turns out that the base frequency is 727 MHz, although the Hawaii GPU wants to work as close to 1000 MHz as possible. By the time the chip reaches the 95 degrees Celsius limit, the fan will most likely be spinning at 40% duty cycle in Quiet mode. Then the GPU frequency starts to drop. Depending on the quality of the chip and the task being processed, Hawaii can drop the frequency down to 727 MHz and remain at that level if the fan fails to provide adequate cooling.

We got a sample from AMD. With 40% of the fan's duty cycle, the frequency was kept at an average of 874 MHz. During a real game on a "hot" card (and not during testing a "cold" chip) a two-minute test of Metro: Last Light showed that the frequency was in the 700 MHz range most of the time. In fact, in some games, the frequency can drop below 1000 MHz even before you exit the menu and enter the battlefield (Arma and BioShock).

This can be called dubious marketing. Indeed, in order to really get stable 1000 MHz frequencies, you have to accept the fact that the video card fan howls like a March cat, or play games taking into account the platform's limitations. But on the other hand, if it's still faster than the competition, then what's the difference how Hawaii does it, right?

How is it positioned Radeon R9 290 relative to competitors?

We left the scaling and fan speed profile at 40% at the same settings so you can see the fluctuations between sags and spikes in performance. AMD states that Radeon R9 290 operates at frequencies up to 947 MHz, but in our seven games the average frequency was 832 MHz. In the most difficult scenarios, the lower threshold of the clock frequency (or base frequency) dropped to 662 MHz. If the GPU does not receive the necessary cooling even at such a low base frequency, the fan speed will exceed 40% (in three runs of the Metro: Last Light stress test, the duty cycle reached 44%).

This can be said to be the opposite of Nvidia's GPU Boost technology. AMD sells cards, indicating the highest frequency, which in fact turns out to be lower. Nvidia cites a base clock, but it goes up when GPU thermal headroom is present. The company correctly specifies the base and typical frequencies in Boost mode. AMD's spec sheet clearly lacks transparency, and we hope, after the praise in our review, that Nvidia will not follow AMD's lead in marketing.

Ultimately, performance metrics are important. But do not rush to draw final conclusions. The longer we run the tests on stock settings, the further the average frequency moves away from the declared 947 MHz. Our sample lacks Quiet and Uber modes. Both positions of the firmware switch leave the maximum fan duty cycle at 40%. And to complicate matters, before the premiere, AMD released an updated driver that overwrites the BIOS setting, and now, by default, the fan can run at 47% of the duty cycle. What is the impact of this modification?

As we know, to overcome AMD's throttling mechanism, you need to manually increase the duty cycle in the OverDrive applet. By raising the fan speed from 40% to 47%, AMD allows the reference cooler to "blow" harder, maintaining higher frequencies for longer, but at the cost of noticeably higher noise levels. We again pin our hopes on AMD partners, maybe they can release a more stable and quiet version Radeon R9 290.

But now we have only a reference sample in our hands. With this in mind, let's consider Radeon R9 290 in details.

Radeon R9 290 Review | By the way: volatility is an unenviable characteristic for a video card.

Before considering in detail Radeon R9 290 that AMD sent to our lab, we need to talk about the phenomenon of mutability of two identical GPUs.

Hawaii has the potential of a very, very fast GPU. If you provide good cooling and keep the frequency ceiling, it will be able to compete with the GPU GeForce GTX Titan... We've already seen this in our flagship card review and awarded it our Elite Award for that. But if the card is not pampered, the chip will quickly show that it is working close to the limit of its capabilities. Unfortunately, the AMD reference cooler, spinning at normal acoustical speeds, cannot provide Hawaii with adequate cooling to maintain a stable frequency in various applications. At first it will be 1000 MHz, but after a few minutes the frequency will drop. Depending on the specific GPU, it can fluctuate between 900, 800 or 700 MHz. That is, the test results on different cards may differ noticeably.

The card that AMD sent in is a "purebred horse". Even if it is thoroughly warmed up before the tests, dropping the frequency from supposedly 1000 MHz, it still overtakes GeForce GTX 780 and sometimes GeForce GTX Titan... But, which we ourselves bought in the online store, turned out to be much weaker. It "slowed down" at the frequency of 727 MHz and so it remained on it, and the reference cooler could not cool it down properly. As a result, the fan speed limit of 40% was exceeded. But also the comic of the situation is that the press sample Radeon R9 290 in most tests it outperforms the flagship we bought in the store. You will see this in our benchmarks.

In line with Tom's Hardware policy, we shared the issues we found with AMD prior to posting, and the company insisted there was something wrong with the retail graphics card we tested. We will continue to investigate and update the article in case of any changes.

Does this mean that our confession is unfounded? We can say that the price cuts on Nvidia cards are putting pressure on AMD's flagship that was not felt a few weeks ago. And in light of the fluctuation between the minimum and maximum observed frequency levels of about 30%, it is difficult to hope for a representative sample taken for the press.

Not sure if it's a super fast graphics card or a better sample

Of course, in this way, we find ourselves between a rock and a hard place. For a correct comparison, we can go and buy another card. But we still cannot know how it will work. Radeon R9 290 that you bought in a store: at the maximum (947 MHz) or at the minimum (662 MHz).

The good news in this whole situation, perhaps, is that the discussed cards and Radeon R9 290 use AMD's reference cooler design. This is a weak link affecting all Hawaii GPU-based products that we have tested up to this point (and we have been running tests for three weeks non-stop). The situation will be changed by OEM partners who will be able to offer coolers of their own design.

However, rumor has it that AMD is holding back partners until the GeForce GTX 780 Ti hits the market, allowing the company to re-analyze the ultra-high-end graphics card market and chart a target for another victory. We have a Hawaii chip in our lab that is stable at 1.158 GHz, and we know that a card with two 8-pin auxiliary power connectors is a beast. However, we do not think that AMD or its partners will sell a competitor for the 780 Ti for the same $ 550.

So, given: a stripped-down GPU and a reference cooler. Let's figure it out.

Radeon R9 290 Review | We meet a new video card

It will hardly come as a surprise to anyone that the second graphics card using GPU Hawaii GPU, with the exception of minor changes, looks identical to the model? Considering the lack of development of the situation with the cooler, we expected that Radeon R9 290 will be indistinguishable from the older model in a number of characteristics. Therefore, today's description was not difficult to do.

In a nutshell, this is a familiar graphics card in a 280mm long dual-slot format with a 75mm centrifugal fan.

Today AMD finally officially presented its flagship - Radeon R9 290X, the most powerful of the new generation video cards. Now let's figure out what is interesting about it and how much faster it is than the leader of the GeForce 700 series.

Introduction

Innovations

Like the entire Radeon R series, the R9 290X is based on the Graphics Core Next architecture. There are not so many differences from last year's flagship in the face of the Radeon HD 7970. First, the GCN blocks are now not thirty-two, but forty-four, which together gives 2816 instead of 2048 Compute Units. Second, the number of geometric blocks responsible for tessellation was doubled in the crystal. Third, the RoP set has doubled, swelling to 64 processors per graphics card. And fourthly, the memory bus expanded from 384 to 512 bits and received 4 GB of GDDR5.

Naturally, it was not without new technologies announced at the launch of the line. The Radeon R9 290X supports the AMD TrueAudio audio engine, DirectX 11.2 and will soon learn to work with its own API called Mantle, which developers want to replace DirectX. Another interesting thing related to the Radeon R9 290X is that you no longer need connecting bridges to organize the CrossFireX mode.

Testing

The board came to us in a reference version from the AMD office. Apart from the changed casing with swirling red lines, the cooling system migrated here directly from the Radeon HD 7970. The crystal is covered with a large radiator with a base in the form of an evaporation chamber and is blown by a cylindrical turntable. AMD was not surprised by the set of ports - a modern standard consisting of a pair of DVI, HDMI 1.4a and DisplayPort 1.2. But what struck me was that power supply requires not two, but only one 8-pin socket from the power supply unit, the second - 6-pin, as on cards of a lower class. Unfortunately, it is not known exactly how much the new product consumes; AMD did not provide official data on this point.

For testing, we have assembled a test bench based on the motherboard MSI Eclipse SLI, equipped it with a processor Intel Core i7-920 and three strips of RAM Kingston HyperX DDR3-2000 MHz... Operating system Windows 7 Ultimate 64-bit and driver Catalyst 13.11 BetaV5 put on a solid state drive Kingston SSDNow 64 GB, and the entire software package rolled onto a classic hard drive Seagate 7200.10.

He got hit 3DMark 11, Unigine Heaven Benchmark 4.0, Unigine Valley 1.0, Aliens vs. Predator, Tomb raider, Batman: arkham city, Bioshock Infinite, Metro: Last Light and Crysis 3... As for the competitors, we included only one video card among them - NVIDIA GeForce GTX 780, which is the target of the R9 290X. The result is interesting.

Synthetics aside, the R9 290X only managed to win in Aliens vs. Predator. In all other games it has parity with the GTX 780. Of course, not 100%, in some modes either NVIDIA or AMD showed a gap of two or three frames per second, but overall the performance of both motherboards is very close.

During the tests, we did not have any questions about the operation of the R9 290X. And although we spent only a day with the board - not a single driver failure or glitch. Perhaps the only thing that can be found fault with is the floating frequency of the chip, which from time to time went up to 1000 MHz, then dropped to 920 MHz. Here we will also say about the high temperature. While the GTX 780 stuck around 70 ° C, AMD's flagship reached 90 ° C. However, it is quite possible to cope with the last snag. The R9 290X, like the HD 7970, has two BIOS versions, Uber and Quiet. Our tests took place in the first one, but if we switch to the second, the card will calm down a little, although it will lose 2-3 fps.

Eventually

Does parity with NVIDIA mean that AMD failed to cope with the task? There is no answer to this question, since the “red” are still silent about prices. If the video card costs equal to the GTX 780 (22,000 rubles), then, of course, AMD will not be in a very advantageous position. But if they ask 15-16 thousand rubles for the R9 290X, the flagship will be more than a reasonable choice.

The R9 290X should also be an interesting proposition for those who are going to buy an Ultra HD monitor; the doubled ROP unit, in theory, will show a noticeable advantage over the GTX 780. However, we were unable to verify this, there were no such matrices in the edition at the time of the tests.

P.S. Due to the fact that the "Iron Shop" is being transferred to a new test bench, at the moment the second stage of testing these video cards on another hardware has been launched in "Igromania".

We will definitely add the results to this article and expand the general conclusion a little.

Specifications
Characteristic	AMD Radeon R9 280X	NVIDIA GeFroce GTX 780
Core	-	GK110
Number of transistors	6.2 billion	7.1 billion
Technical process	28 nm	28 nm
Number of stream processors	2816 pcs.	2304 pcs.
Graphics core frequency	up to 1000 MHz (in fact - 920-1000 MHz)	863-900 MHz
Type, memory size	GDDR5 4GB	GDDR5, 3 GB
Memory frequency	-	6008 MHz
Data bus	512 bit	384 bit
Number of texture units	176 pcs.	192 pcs.
Number of ROP units	64 pcs.	48 pcs.
Energy consumption	-	250 watts
Price for October 2013	-	22,000 rubles

Synthetic Benchmarks (Radeon R9)
3DMark 11 (Performance)
Graphics card model	Graphics	Physics	Score	%
AMD Radeon R9 290X	15 144	6110	10 904	100%
NVIDIA GeForce GTX 780	12 767	6014	9873	91%
Unigine Heaven Benchmark 3.0
Graphics card model	FPS	Scores		%
AMD Radeon R9 290X	50,2	1263		100%
NVIDIA GeForce GTX 780	49,3	1241		98%
Unigine Valley 1.0
Graphics card model	FPS	Scores		%
AMD Radeon R9 290X	52,7	2204		100%
NVIDIA GeForce GTX 780	54,9	2298		104%

Game tests (frames per second)
Aliens vs. Predator (DX11)
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
VeryHigh, 1680x1050, AF 16x, AA 2x	127,2	108,8
VeryHigh, 1680x1050, AF 16x, AA 4x	107,8	92,6
VeryHigh, 1920x1080, AF 16x, AA 2x	106,5	95,9
VeryHigh, 1920x1080, AF 16x, AA 4x	94	83,7
%	100%	87%
Tomb raider
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
High, 1680x1050, AF 16x, AA 4x	43,4	43,7
High, 1680x1050, AF 16x, AA 8x	43,2	43,8
High, 1920x1080, AF 16x, AA 4x	37,7	38,4
High, 1920x1080, AF 16x, AA 8x	37,1	38,3
%	100%	102%
Batman: arkham city
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
	67	101
	44	100
Ultra, 1680x1050, AF 16x, PhysX	29	52
	60	99
	39	93
Ultra, 1920x1080, AF 16x, PhysX	28	52
%	3%	186%
Bioshock Infinite
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
Ultra, 1680x1050, AF 16x, AA 4x	91,9	93,3
Ultra, 1680x1050, AF 16x, AA 8x	91,8	92,9
Ultra, 1920x1080, AF 16x, AA 4x	81,6	85,5
Ultra, 1920x1080, AF 16x, AA 8x	81,5	84,4
%	100%	103%
Metro: Last Light
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
VeryHigh, 1680x1050, AF 16x	48	43
VeryHigh, 1680x1050, AF 16x, Adv. PhysX	29	37
VeryHigh, 1920x1080, AF 16x	42	39
VeryHigh, 1920x1080, AF 16x, Adv. PhysX	27	33
%	100%	104%
Crysis 3
Settings	AMD Radeon R9 290X	NVIDIA GeForce GTX 780
Max, 1680x1050, AF 16x, AA 4x	41,6	41,6
Max, 1680x1050, AF 16x, AA 8x	36	38,4
Max, 1920x1080, AF 16x, AA 4x	39,1	39,8
Max, 1920x1080, AF 16x, AA 8x	32,9	34,1
%	100%	103%

Hawaii has even smaller islands: a little less speed, but the price is much lower!

• Part 2 - Practical introduction
• Part 3 - Game Test Results (Performance)

We continue the saga about "crabs". The beginning can be read. So, the "reddish crab" made a "baaaaal ball" and successfully rolled it into the "hole", which caused some nervousness in the "greenish crab", which became even greener with anger and went to conjure something with his "balls".

However, the "reddish crab" decided that such a huge "ball" was not enough, and, having trimmed the existing "balls", slightly reducing them, began to roll them into the "hole". Smart - yes, he understands, smart - yes, he knows. The rest - read below :)

Part 1: Theory and Architecture

Following the release dedicated to the current line of AMD graphics solutions, consisting of the Radeon R7 and R9 families, as well as a full review of the senior solution in the line - Radeon R9 290X, which became the first video card based on the new graphics chip codenamed Hawaii, there was an article about the younger model of the top line based on the same chip, which AMD decided to release a little later.

The release of such a video card after the older version is not typical for AMD. At first, the release of the video card one step below the top-end was scheduled for a week later, and then it was postponed again, already due to military cunning. AMD still could not finally determine the strategy and tactics of the struggle of the Radeon R9 290 family with its competitors from the Nvidia camp, for which this company reduced prices after the release of the top model Radeon R9 290X, and also announced the release of a new top - Geforce GTX 780 Ti. In this regard, AMD decided to further refine the drivers, squeezing out a few more percent of the performance and at the same time bring the date of the Radeon R9 290 announcement closer to the date of the actual appearance of the announced video cards on sale.

Let us recall again which video cards are included in the new families of AMD Radeon video cards. The current lineup of the company contains several series: R9 and R7. The new line of the company contains the following models, covering most of the market segments:

The R7 250 and R7 260X graphics cards target the $ 90- $ 140 price range (US market prices), the R9 270X retails for $ 200 and the R9 280X for $ 300. The recommended price for the flagship of the line - model R9 290X - in the US market is $ 549, and for the R9 290 they decided to ask for another hundred and a half less, and it is located between the R9 280X and R9 290X (it is not on the diagram).

The Radeon R9 290 model is not much different from the top-end R9 290X, since it is based on the same Hawaii GPU of the Graphics Core Next architecture. In this GPU, the number of executive devices was increased while keeping the power consumption of solutions within acceptable limits. The main task in the development of the new GPU for AMD was to improve its energy efficiency, which was achieved, as we saw in the corresponding review.

It is clear that the new solution supports all the company's technologies, including ZeroCore, updated PowerTune, Eyefinity, etc., etc. The R9 290 simply costs less than the top-end R9 290X and provides an even better performance-to-price ratio, since it is not much inferior in terms of speed, but has a noticeably lower price. So let's take a look at the AMD Radeon R9 290 graphics card.

Let's see how AMD itself is positioning its new solution on the market using the traditional Fire Strike tool in two different versions (Performance and Extreme) from the latest version of the Futuremark 3DMark benchmark. This benchmark is great for modern AMD graphics cards, and the points scored in it by the Radeon R9 290 video card slightly exceed the performance of its main competitor - the Geforce GTX 780.

Since the new model of the video card Radeon R9 290 is almost a twin of the video card R9 290X, and the Hawaii graphics processor largely repeats the features of the GPU of the previous Radeon HD 7000 series, it will be useful to familiarize yourself with detailed information about the early decisions of AMD before reading this material:

• AMD Radeon R9 290X: Reach Hawaii! You will get new heights of speed and functionality
• AMD Radeon R7 and R9 - updated line of video cards: new families so far without their flagship
• AMD Radeon HD 7790: a senior representative of the middle class of accelerators
• AMD Radeon HD 7970: the new single-processor 3D graphics leader

Graphics Accelerator Radeon R9 290

• Chip codename: "Hawaii"
• Manufacturing technology: 28 nm
• 6.2 billion transistors
• Unified architecture with an array of common processors for streaming processing of multiple types of data: vertices, pixels, etc.
• Hardware support for DirectX 11.2, including Shader Model 5.0
• 4 geometry processors
• 512-bit memory bus: eight 64-bit controllers with support for GDDR5 memory
• Core frequency up to 947 MHz (dynamic)
• 40 (out of 44 physically present in the chip) GCN computing units, including 160 (out of 176) SIMD cores, consisting of a total of 2560 (out of 2816) ALUs for floating point calculations (integer and floating point formats are supported, with FP32 precision and FP64)
• 160 (out of 176) texture units, with support for trilinear and anisotropic filtering for all texture formats
• 64 ROPs with support for full-screen anti-aliasing modes with the possibility of programmable sampling of more than 16 samples per pixel, including FP16 or FP32 framebuffer format. Peak performance up to 64 samples per cycle, and in Z only mode - 256 samples per cycle
• Integrated support for up to six monitors connected via DVI, HDMI and DisplayPort

Radeon R9 290 Graphics Card Specifications

• Core frequency: up to 947 MHz
• Number of universal processors: 2560
• Number of texture units: 160, blending units: 64
• Effective memory frequency: 5000 MHz (4 × 1250 MHz)
• Memory type: GDDR5
• Memory capacity: 4 gigabytes
• Memory bandwidth: 320 gigabytes per second.
• Computing Performance (FP32) 4.9 teraflops
• Maximum theoretical fill rate: Up to 60.6 gigapixels per second.
• Theoretical texture sampling rate: up to 152 gigatexels per second.
• PCI Express 3.0 bus
• Two Dual Link DVI, one HDMI, one DisplayPort
• Power consumption up to 275 W
• One 8-pin and one 6-pin power connector;
• Two-slot design
• The recommended price for the US market is $ 399 (for Russia - 13,990 rubles).

We continue to get used to the new AMD video card naming system, but once again we doubt the correctness of the naming of the Radeon R9 290 and 290X models, which differ only in an additional symbol in the name of the older model. Both video cards belong to the top-end R9 family and have the maximum serial number in the series - 290, but the older one has an additional index "X" in the name, although in the previous family they got by with numbers. In our opinion, names like R9 2950 and R9 2970 would be more logical and understandable, as before.

As usual, the recommended price for the Radeon R9 290 is lower than the price of the corresponding solution from the competitor from the same price segment, even taking into account the price cuts announced by Nvidia recently. The Radeon R9 290 is clearly aimed at fighting the Nvidia Geforce GTX 780 based on the GK110 chip, which is the youngest top-end motherboard of the competitor (after all, there is a Geforce GTX Titan for a long time, and the GTX 780 Ti has already been announced and will be released soon). The Nvidia model has a higher MSRP ($ 499 versus $ 399), but it can provide better performance in games - this isn't the AMD-friendly Fire Strike from 3DMark. However, we will explore all such questions in the third part of the material.

The second top model of the AMD video card, which we are considering today, as well as the older board, has GDDR5 memory of four gigabytes. Since the Hawaii graphics chip has a 512-bit memory bus, which is not cut down in the Radeon R9 290, 2 GB could have been supplied, but for a top solution this amount is too small, because modern projects are already starting to use large amounts of video memory, which will only worsen in the coming months, when multi-platform games will be released for the next generation of consoles: PS4 and Xbox One.

Unlike the older model, which has a special switch on the card that allows you to select one of two BIOS firmwares: (quiet mode and supermode), which differ in power consumption and performance, the younger model Radeon R9 290 does not have this option, only one mode is available for it. It is clear that the power consumption has decreased compared to the R9 290X, especially when compared to the latter's super mode, but the official figures are not known.

Architectural features

We are already talking about the new Hawaii graphics chip, on which the AMD Radeon R9 290X and 290 graphics cards are based. This GPU is based on the well-known Graphics Core Next (GCN) architecture, which has been modified in terms of computational capabilities and support for all DirectX 11.2 capabilities, similar to the Bonaire chip (Radeon HD 7790 and R7 260X). The architectural changes also include support for more concurrently executing instruction streams and a new version of AMD PowerTune technology.

We already know that the Hawaii graphics chip differs from the Tahiti in a large number of executive devices. But what is the difference between the Hawaii variant used in the Radeon R9 290 and the full-fledged one? Let's see exactly how AMD engineers "slowed down" the junior top model so that it does not become too strong a competition for the R9 290X:

The Hawaii GPU consists of four Shader Engines, each of which physically contains 11 large Compute Units, including texture units, a geometry processor and a rasterizer, and ROPs. In total, the Hawaii graphics chip includes 44 Compute Units containing 2816 stream processors, 64 ROPs and 176 TMUs.

But this applies to a full-fledged chip with all active blocks, which is used in the Radeon R9 290X. The younger R9 290 received a chip with 40 active Compute Units containing 2560 stream processors and 160 texture units. But the number of ROP units was not cut, there are 64 of them left. The same applies to the memory bus, it remains 512-bit, consisting of eight 64-bit controllers.

In our Radeon R9 290X review, we took a close look at the internals of the GPU, the shader engines that make up it, and the Compute Units (CUs). The only difference between R9 290 and R9 290X is that the shader engine in this case does not include 11 CUs, but fewer (for example, 10 pieces, if one CU is disabled in each of the engines). And already each computational unit CU contains functional blocks: 16 texture sampling units, 4 texture filtering units, a branch prediction unit, a scheduler, four vector and one scalar computational units, memory for vector and scalar registers, 16 KB cache memory of the first level and 64 KB of shared memory.

Note that the Hawaii chip has four geometry processing units and rasterization engines and is able to process up to four geometric primitives (triangles) per clock cycle. Also in Hawaii, the buffering of geometric data was improved and the caches for the parameters of geometric primitives were increased, which provides a significant increase in performance for large calculations in geometry shaders and the active use of tessellation.

The Hawaii chip contains 16 Render Back End blocks (double that of the Tahiti) containing 64 ROPs capable of handling up to 64 pixels per clock cycle. This GPU has 1 megabyte of L2 cache, which is divided into 16 sections of 64 KB each, and their internal bandwidth has also been increased by a third. Memory access is carried out using eight 64-bit controllers, and since the memory chips in the Radeon R9 290 also operate at 5.0 GHz, as in the older model, the total memory bandwidth has not changed and is equal to 320 GB / s.

Functional features

It is quite logical that the new video card Radeon R9 290 supports exactly the same technologies as the older model. We have already covered all the new technologies supported by the Hawaii graphics chip in the corresponding Radeon R9 290X review, and will repeat this information here in a very short form.

The solution presented today has support for a new graphical API Mantle which is able to take advantage of all the hardware capabilities of AMD GPUs without being limited by the software limits of available graphics APIs like OpenGL and DirectX, and using a thinner software interface between the game engine and the GPU hardware resources, similar to the way it is done on game consoles. Mantle was developed at AMD with significant contributions from leading game developers from DICE, and Battlefield 4 is the first project to use Mantle.

Theoretically, using Mantle can provide up to a ninefold advantage in the execution time of drawing function calls compared to other graphics APIs, but such an advantage is possible only in artificial conditions, in real games it is good if ten percent is typed (which is also not bad). Battlefield 4 Mantle API support is expected in December, when a free update optimized for AMD Radeon graphics is released. On systems with GCN graphics cards, the Frostbite 3 engine will use this API, which will reduce the load on the CPU by parallelizing work across eight cores, and will also introduce special low-level performance optimizations for AMD video chips.

Another important technology supported by the Radeon R9 290 is "sound" technology AMD TrueAudio, which we have already talked about. With the release of the Radeon R7 and R9 series, AMD introduced a programmable audio engine, which is only supported by AMD Radeon R7 260X, R9 290 and R9 290X graphics cards, since Bonaire and Hawaii chips are the only ones so far from the company that have received the most modern technologies. ...

TrueAudio is an embedded programmable audio engine in AMD's GPU that provides real-time processing of audio tasks, regardless of the installed CPU. Several Tensilica HiFi EP Audio DSP DSP cores are integrated into the Hawaii and Bonaire chips, and their capabilities are accessed using popular sound processing libraries, whose developers use the resources of the built-in audio engine using a special AMD TrueAudio API.

The new top-end video cards have been improved and PowerTune power management technology... We wrote about these improvements back in the Radeon HD 7790 review, for more efficient power management, most modern AMD graphics chips have multiple states with different frequencies and voltages, which allows for higher clock speeds. The Hawaii video chip integrates a second generation serial voltage regulator interface and the GPU always operates at the optimal voltage and frequency for the current GPU load and power consumption, on which the switching between states is based.

In addition, the transitions between the states themselves became very fast, the GPU switches PowerTune states 100 times per second, and the Radeon R9 290 series solutions simply cannot have a single operating frequency, there is only an average over a certain period of time. Based on such global changes, modified options for overclocking and tuning PowerTune appeared in the settings of the Catalyst Control Center driver in the OverDrive tab:

Power Limit and GPU clock are linked in the power consumption versus heat dissipation diagram, which makes overclocking more logical and understandable. In addition, the diagram reflects the fully dynamic control of the GPU frequency that has appeared in the series solutions. The increase in GPU frequency is now indicated by increasing the corresponding value (GPU Clock) by a certain percentage, and there is no longer the possibility of specifying a specific frequency.

Also in the new OverDrive interface, the fan speed control has been seriously changed. Unlike the previous option to set a fixed fan speed, the new interface has a “Maximum Fan Speed” setting that only sets the upper limit for the fan speed. In this case, the fan speed will change based on the load on the GPU and its temperature, and will not remain fixed, as it was before.

The most interesting innovation in the AMD Radeon R9 290 series graphics cards is the support of the technology AMD CrossFire without the need to connect video cards to each other using special bridges. Instead of dedicated communication lines, the new GPUs communicate with each other over the PCI Express bus using the built-in hardware DMA engine. The performance and image quality is exactly the same as with the connecting bridges.

In the case of a pair of Radeon R9 290 graphics cards, CrossFire provides excellent scaling on multi-chip systems when considering average frame rates. The following diagram shows the comparative performance of a single AMD Radeon R9 290 versus two such cards running on AMD CrossFire rendering technology:

In the games shown in the diagram, a very good increase in the average frame rate is provided when the second video card is connected - up to almost two times. In the worst case, these applications show 80% CrossFire efficiency, and on average it turns out to be more than 85%.

About changes in the output of images to displays, technology AMD Eyefinity and support for Ultra HD resolution, we also wrote. The new graphics card allows connection of two possible variants of Ultra HD displays: TVs that support only 30 Hz and below at 3840 × 2160 and connected via HDMI or DisplayPort, as well as monitors, the image of which is divided into two halves with a resolution of 1920 × 2160 at 60 Hz ...

To support "split" monitors, the new VESA Display ID 1.3 standard has been introduced, which describes additional display capabilities. The standard allows you to automatically "glue" the image for such monitors, if it is supported by both the monitor and the driver. And in the future, AMD Radeon graphics cards will also support a third type of UltraHD display, which only needs one stream to run in ultra-high resolution at a refresh rate of 60 Hz.

The Radeon R9 290 delivers reasonably high 3D performance for multi-monitor configurations, which is essential at maximum gaming settings and the highest rendering resolutions in such systems. The new product has a clear advantage over the Nvidia Geforce GTX 780, expressed in the larger amount of video memory, which is important at ultra-high resolutions, as seen in comparison with the performance in Ultra HD resolution:

As you can see, the performance of the new solution makes it possible to play at maximum settings in this resolution, getting an acceptable average frame rate above 30 FPS (and more will not be seen in the case of some displays) even in such demanding games as Crysis 3 and Battlefield 3.

The Radeon R9 series graphics cards have also been updated with technology AMD Eyefinity... The new top-end video card Radeon R9 290 supports configurations of up to six displays using all available connectors, and up to three HDMI / DVI displays. For the latter case, without using DisplayPort connectors with hubs, you will need three displays that support identical timings. The output setting is only available at system startup and does not support hot-plugging the display for a third HDMI / DVI connection.

Theoretical performance assessment and conclusions

Let's consider the theoretical performance indicators of the new product, estimating how much the new Radeon R9 290 video card should be slower than the top model Radeon R9 290X in reality and how comparable it is with the R9 280X, previously known as the Radeon HD 7970 GHz. We compare all the most important theoretical indicators for GPUs:

The peak geometry processing speed of solutions based on Hawaii (R9 290 and R9 290X) is almost twice as high as that of solutions based on Tahiti (R9 280X and HD 7970 GHz), and in terms of computational and texture performance, the new product is located exactly in the middle between R9 290X and R9 280X ... The video memory bandwidth of both top-end motherboards is the same and it is 20% higher than that of the R9 280X, and the fill rate (fill rate) of the R9 290 is almost the same as that of the R9 290X (on the plate it is equal, but this is an error - in the case of R9 290 this figure is 60.6 gigapixels per second).

To confirm the capabilities of the new product, especially in memory bandwidth and fill rate, AMD cites the average frame rate achieved in the latest game Battlefield 4 - the sequel to the most popular series developed by DICE. The game is part of the AMD Gaming Evolved Partner Program and therefore has no problem optimizing Battlefield 4 for GCN GPUs. Let's see what happens in the version of the game without Mantle support:

In this game, the fresh Radeon R9 290 outperforms the more expensive competing Nvidia Geforce GTX 780 in both modes with different (but high in both cases) resolutions. Perhaps, the Nvidia video card in such high resolutions is hampered by the lack of video memory, which it has only 3 GB, as opposed to 4 GB in the Radeon R9 290, so a larger amount of video memory is another indisputable advantage of the new product from AMD.

We will test the performance of the new solution in the next parts of the article. Summing up the theoretical part, we note that the AMD Radeon R9 290 video card model announced today should become one of the most profitable offers among single-chip 3D accelerators, and one of the most successful acquisitions in the top price segment intended for enthusiasts. It is also very important that the new model should not lag far behind in speed the Radeon R9 290X, which is the most expensive and powerful single-chip solution in the company's lineup.

AMD offered the market a graphics card with a very competitive price and features. Already in terms of theoretical indicators, it can be argued that this video card model has an excellent price-performance ratio. And the functionality of the new item is all right, it has a built-in sound DSP engine, supports TrueAudio technology and the new graphics API Mantle, as well as all the other technologies that are already familiar to us, which we are already tired of listing.

After we got acquainted with the characteristics and theoretical capabilities of the Radeon R9 290 video card, it's time to move on to the next part of our material, which is traditionally devoted to the study of the rendering speed of a new AMD video card in a set of synthetic tests. It will be interesting to compare the performance of the new product from the upper price segment with the top-end Radeon R9 290X, as well as with the speed of competing Nvidia video cards.

AMD Radeon R9 290 - Part 2: Video Cards and Synthetic Benchmarks →

And he followed. The Radeon R9 290 video card, along with the X index, has lost some of the computing units. Of the 44 Compute Units, four are disabled. As a result, 2560 ALUs, 160 texture units and 64 ROPs are active. Additionally, the operating frequency of the graphics processor has been reduced to a value of 947 MHz. However, it must be remembered that in the case of the Hawaii GPU, this is the maximum frequency that is maintained until the temperature and power limits are reached. If the card goes beyond the specified limits, then the frequency begins to decrease. The effective memory frequency is the same with the older model - 5000 MHz. Data exchange with the GPU is performed via a 512-bit bus.

Video adapter	Radeon R9 290X	Radeon R9 290	Radeon R9 280X	Radeon HD 7970 GHz Edition	Radeon HD 7970
Core	Hawaii	Hawaii	Tahiti	Tahiti	Tahiti
	6020	6020	4312	4312	4312
Process technology, nm	28	28	28	28	28
Core area, sq. mm	438	438	365	365	365
	2816	2560	2048	2048	2048
Number of texture units	176	160	128	128	128
Number of rendering units	64	64	32	32	32
Core frequency, MHz	up to 1000	up to 947	1000	1000-1050	925
Memory bus, bit	512	512	384	384	384
Memory type	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Memory frequency, MHz	5000	5000	6000	6000	5500
Memory size, MB	4096	4096	3072	3072	3072
	11.2	11.2	11.2	11.2	11.2
Interface	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0
Power, W	290	275	250	250	250

AMD's flagship graphics accelerator came out very hot, so the company did not particularly advertise data on power consumption or TDP. According to unofficial data, the TDP level for the Radeon R9 290X reaches 290 W (in other sources - 300 W). In the case of the Radeon R9 290, a value of 275 watts is announced. Such a decrease in the "appetite" of the junior flagship should have a positive effect on the final temperatures and noise. This is confirmed by the 47% fan speed limit, relative to the maximum for Uber's high-performance mode. In the Radeon R9 290X in Uber Mode, the fan speed was limited to 55% of the maximum possible. Although in this mode the flagship dropped the core frequency a little in the heaviest applications and after long loads. In the quieter Quiet Mode, a drop in frequency is completely normal, which led to a decrease in performance by up to 15% in the most severe cases. The main limitation of performance was heating due to rapid peak values. Similar behavior can be expected from the reference Radeon R9 290.

Good cooling can make a difference. Provide lower operating temperatures and you will get stable 1000 MHz for the older video card and 947 MHz for the lower version. In the case of reference cards, this is only possible with a complete replacement of the cooler. But many manufacturers have already taken care of the Radeon R9 overheating problem by releasing models with non-standard cooling systems. It is such a product that will become the hero of our review. ASUS R9290-DC2OC-4GD5 is cooled by DirectCU II and even operates at higher frequencies.

The adapter comes in a large box with a standard ASUS design.

The scope of delivery is as follows:

• power adapter from two 6-pin PCI-E to one 8-pin PCI-E;
• CD with software;
• instruction.

The massive graphics card is larger than the reference version, but still takes up two expansion slots. It looks like a complete copy of ASUS R9280X-DC2T-3GD5.

The streamlined casing carries red decorative accents on the surface, which have become, for example, a distinctive design feature of all Republic of Gamers products. Thick heat pipes peek out from under the casing.

The reverse side of the board is covered with a perforated metal plate. There are no connectors for connecting video cards with a CrossFireX bridge - new Radeon cards work perfectly in pairs without additional hardware interfaces.

The power supply connectors are unfolded with a latch upwards with special cutouts in the PCB. This simplifies the connection of appropriate cables and does not require free space under the board, eliminating the need for cutouts in the heatsink.

The rear panel contains all modern interface connectors: HDMI, DisplayPort and two DVI.

The cooler design does not differ from the cooling of the aforementioned ASUS R9280X-DC2T-3GD5 card. There are five heat pipes at the base - two with a diameter of 6 mm, two with 8 mm and one with 10 mm. The side tubes, the thinnest, do not directly contact the graphic crystal, which is clearly visible from the thermal paste imprint.

Direct contact technology has been implemented. The tubes are tightly fitted to each other at the base, touching the surface of the GPU crystal without additional metal plates.

The massive radiator is evenly pierced with tubes along its entire length. The fans are mounted on narrow mounting strips that are screwed onto the radiator. There are no massive elements that interfere with the blowing of the structure.

Both fans are 100 mm in size, but with different impeller designs. Closer to the rear panel is the most unusual one - with combined blades of two types. It simultaneously acts as an axial and radial fan, due to which part of the heated air is blown out of the case through the rear grill.

The second fan is of a more familiar design, but with more blades.

The power elements on the board are covered with a separate finned heatsink. It does not differ in large size, but it is perfectly blown.

The design of the board resembles ASUS R9280X-DC2T-3GD5, the power phases are even slightly less, but DrMOS power modules are used. There is a small BIOS switch on the left side of the board.

The core is directly powered from a six-phase converter controlled by the DIGI + ASP1300 controller. The memory is powered by two phases. The older Radeon R9 290X of the reference design has a 5 + 1 power supply.

And here is GPU Hawaii in person:

One of the important advantages of the top-end Radeon relative to competitors from NVIDIA is the large memory capacity of four gigabytes. In this case, they are recruited with 16 Elpida W2032BBBG-6A-F microcircuits.

The ASUS frequency formula has been increased from the recommended values ​​of 947/5000 MHz to 1000/5040 MHz.

ASIC Quality of this specimen is 69%.

ASUS R9290-DC2OC-4GD5 carries two firmwares on board. By default, the graphics card runs in Uber mode. In this configuration, the Unigine Valley benchmark warmed up the GPU to 78 ° C at 24-25 ° C indoors. At the same time, the frequency line on the MSI Afterburner graph is almost flat. The reference version of the Radeon R9 290X at an operating core temperature of 94 ° C "pleases" with drawdowns up to 950 MHz in this test application.

Note that, for some unknown reason, the MSI Afterburner and ASUS GPU Tweak utilities did not display the real fan speed, although there shouldn't be any problems with that. If you look at other ASUS cards with a similar cooling system, then the fan speed at 55% of the maximum should correspond to the level of 2000 rpm. Not to say that it is quiet, but there is no real discomfort either. Better than a standard turbine anyway.

In Quiet Mode, the core temperature gradually rises to 94 ° C, after which the usual Hawaii frequency reduction begins. But this process is gradual, and not as fast as the standard version of the Radeon R9 290X. This inertia once again speaks of the advantage of the cooling system over a standard turbine. And ASUS turns out to be very quiet in this mode, although the use of Quiet Mode is rather controversial.

To correctly assess the performance of the Radeon R9 290, we should have measured in both modes. However, under conditions of long core warm-up, in most cases it is problematic to get lower frequencies in short tests. And the dynamics of changing the GPU frequency will differ from that with reference cooling. As a result, we abandoned this idea and conducted tests only in Uber Mode at a stable frequency of 1000 MHz.

By the way, the manufacturer estimates the advantage of ASUS over the reference at 19% based on the results of its measurements in the game Battlefield 4. An impressive gain, especially considering the increase in the core frequency by only 6%. The standard card "loses" additional percentages due to frequency reduction due to overheating. And if marketers have decided to compare ASUS in Uber Mode with the reference in Quiet Mode, then such a gap does not seem extraordinary at all.

Overclocking such a hot video card even with DirectCU II cooling is associated with high noise levels. Both GPU overclocking and an increase in the frequency of the GDDR5 memory lead to a noticeable increase in heating. After long tests, we settled on the 1120/6040 MHz configuration, which is higher than the overclocking results for the Radeon R9 290X. Such frequencies required an increase in the supply voltage of the core by 38 mV with an increase in the power limit to the limit (+ 50%). The fans had to be unscrewed by 75%, and after a long warm-up, set even higher values.

In all respects, the video card worked at its maximum in noisy mode. But if we restrict ourselves to overclocking without a softvolt mod and a significant increase in the memory frequency, then it is quite possible to get half of our total increase while maintaining an acceptable noise level of blowing fans. Characteristics of the tested video cards

We will compare the reviewed video adapter with the Radeon R9 290X made by ASUS and with a similar cooling system DirectCU II. The older friend was also tested only in the maximum performance mode Uber Mode with a constant core frequency. From NVIDIA's side - absolutely standard versions of the GeForce GTX 780 and GeForce GTX 780 Ti. All their characteristics are shown below.

The table for GeForce shows the official data on the average Boost Clock. The graphs show the full range of core frequencies, including barely reached peak values.

Video adapter	Radeon R9 290X		Radeon R9 290	GeForce GTX 780 Ti	GeForce GTX 780
Core	Hawaii	Hawaii	Hawaii	GK110	GK110
Number of transistors, mln.pcs	6020	6020	6020	7100	7100
Process technology, nm	28	28	28	28	28
Core area, sq. mm	438	438	438	561	561
Number of stream processors	2816	2560	2560	2880	2304
Number of texture units	176	160	160	240	192
Number of rendering units	64	64	64	48	48
Core frequency, MHz	up to 1000	up to 1000	up to 947	875-926	863-900
Memory bus, bit	512	512	512	384	384
Memory type	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Memory frequency, MHz	5000	5000	5000	7000	6008
Memory size, MB	4096	4096	4096	3072	3072
Supported DirectX Version	11.2	11.2	11.2	11.1	11.1
Interface	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0	PCI-E 3.0
Power, W	290	n / a	275	250	250

Test stand

The test bench configuration is as follows:

• processor: Intel Core i7-3930K (3, [email protected], 4 GHz, 12 MB);
• cooler: Thermalright Venomous X;
• motherboard: ASUS Rampage IV Formula / Battlefield 3 (Intel X79 Express);
• memory: Kingston KHX2133C11D3K4 / 16GX (4x4 GB, [email protected] MHz, 10-11-10-28-1T);
• system drive: Intel SSD 520 Series 240GB (240 GB, SATA 6Gb / s);
• additional disk: Hitachi HDS721010CLA332 (1 TB, SATA 3Gb / s, 7200 rpm);
• power supply unit: Seasonic SS-750KM (750 W);
• monitor: ASUS PB278Q (2560x1440, 27 ″);

• operating system: Windows 7 Ultimate SP1 x64;
• GeForce driver: NVIDIA GeForce 335.23;
• Radeon R9 280X driver: ATI Catalyst 14.3.

User Account Control, Superfetch and visual effects of the interface are disabled in the operating system. Driver settings are standard.

Testing technique

All applications were tested at 2560x1440 with maximum graphics quality. Whenever possible, heavy anti-aliasing modes were also activated. For GeForce, in many cases an additional run was carried out in each test due to a gradual decrease in frequency during core warm-up (peculiarities of the Turbo mode).

A short run was made on the very first island after the first checkpoint during the story prologue. Four repetitions, the procedure was repeated. The game route runs through an area with abundant vegetation, which is most critical for video cards. Highest quality details, highlights and shadows with 2x MSAA anti-aliasing.

Testing was carried out in the first mission after blowing up the wall. Jogging over a small area with dense vegetation was repeated before descending to a large construction site. Five repetitions. Frame rates were measured using Fraps.

All graphics options are in Ultra position. The "resolution" setting is 140%. Anti-aliasing is completely disabled.

Highest quality graphics with 4x MSSA anti-aliasing. Testing was carried out on the first mission while jogging along a city street. The corresponding scene is shown below. Five repetitions. Frame rates were measured using Fraps.

The test was carried out at the beginning of the "Just a Human" level. A short run through the limited visibility zone was combined with an inspection through the sight of distant objects (that's why such a large-scale level was chosen). Ultra graphics quality, 2x MSAA anti-aliasing. Four repetitions of the test. Measurements were taken using Fraps.

Five repetitions of the built-in gaming test. Graphics settings are maximum with 8x MSAA anti-aliasing.

Performance was measured during a four-time run of the built-in benchmark and during "manual" testing at the "Ashes" level (video of the test segment below).

Graphics quality is maximum (Very High) with SSAA anti-aliasing. In a real game, this resource intensive anti-aliasing was set to 2x mode. In the benchmark, SSAA quality is not separately regulated. Since even top-end cards with such a mode cope with great difficulty, additional testing was carried out without anti-aliasing, but only using the built-in benchmark.

Thief

Five runs of the built-in performance test. Graphics quality is maximum under DirectX 11, high quality SSAA anti-aliasing is active.

Mission "Shelter" begins in Benghazi. The frame rate was measured during the opening scene and during a short walk down the street. Maximum graphics quality with 4x MSSA anti-aliasing. Four repetitions of the test.

Maximum graphics quality. All additional parameters are active. Anti-aliasing FXAA. TressFX technology is active. Five repetitions of the introductory script scene. We do not use the built-in benchmark for the reason that it simulates a too simple game scene.

Four runs of the built-in performance test. All quality parameters are at an extreme level, all additional effects are included.

Ultra quality with 4x MSAA anti-aliasing. Three repetitions.

Testing on the Extreme profile.

Tested in Extreme mode, which assumes a resolution of 2560x1440

Energy consumption

For a comprehensive assessment of energy consumption, measurements were made in several games:

• Thief;
• Metro Last Light (SSAA);

The measurements were carried out with a Cost Control 3000 device. The maximum readings of the device were taken into account. Based on the results of all test runs, the average maximum for each application was calculated. And already from these values, the final average was calculated, which is displayed on the comparative chart. Test results

In the last part of Assassin's Creed, NVIDIA solutions have a significant advantage. Even the Radeon R9 290X loses slightly to the GeForce GTX 780. The difference between AMD representatives is small. When overclocked, ASUS easily bypasses its older brother.

В // www .. And with the same core frequency of 1000 MHz, the difference between the cards is even smaller. The youngest Radeon is 9-11% faster than the GeForce GTX 780, and when overclocked to 1120/6040 MHz it slightly bypasses the leader in the face of the GeForce GTX 780 Ti.

Slight lag of Radeon R9 290 over GeForce GTX 780 in Ghosts. ASUS is slightly better than this competitor in terms of minimum fps. The difference between ASUS and Radeon R9 290X is only 4-5%, and the gain from overclocking is 13%.

Radeon R9 290 is 5-7% faster than the younger GeForce and 6-8% weaker than its older brother AMD. Overclocking allows not only bypassing it, but also reaching the performance level of the GeForce GTX 780 Ti.

AMD video adapters take the lead in Absolution. For the first time, the Radeon R9 290 manages to slightly outperform the GeForce GTX 780 Ti even at factory frequencies. Overclocking to 1120/6040 MHz gives a performance increase of 17% over the standard frequency configuration.

Without anti-aliasing, the new Radeon R9 290 occupies an average position between NVIDIA's competitors, losing 8% to its older brother. Overclocking allows you to come close to the leader GeForce GTX 780 Ti.

When you activate anti-aliasing, the lag of the GeForce GTX 780 slightly decreases, but there are no fundamental changes in the alignment of forces.

During manual testing, we see that the difference between the junior participants is even smaller, and they are equal in terms of the minimum fps. And when overclocked, the GeForce GTX 780 confidently bypasses the Radeon R9 290. The difference between the junior and senior representatives of AMD reaches 10%. ASUS outperforms the reference by 2-3%, accelerating by another 10-12% when overclocked.

Thief

The Radeon R9 290 loses no more than 8% to its older brother in this test, outperforming the younger GeForce by 9-19%. When overclocked, the difference between the Radeon R9 290 and the GeForce GTX 780 is reduced to 3-11%. The hero of the review in this mode bypasses the GeForce GTX 780 Ti in minimum fps.

Support for the Mantle API was recently added to the game. Therefore, it was decided to conduct additional testing in this mode in order to compare the results with rendering under DirectX 11.

In our configuration, there was no significant gain when switching to Mantle, which, however, does not mean that such a mode is not useful. Mantle should provide the most noticeable gain on weak CPUs. And the Core i7 at 4.4 GHz seems to be so powerful that it negates the benefits of the new API.

In the Blacklist, the new Radeon outperforms the GeForce GTX 780 by no more than 8%. With the same core frequency of 1000 MHz, the difference between AMD representatives is 5%. ASUS overclocking allows you to reach the performance level of the top-end GeForce.

known for optimizing for AMD solutions, especially when using TressFX technology. But even in this case, the GeForce GTX 780 Ti is in the lead. But the Radeon R9 290 is 12-15% faster than the GeForce GTX 780. The most significant difference between the younger and older Radeon is in this game - from 12% in average frame rate to 15% in the minimum indicator. Overclocking allows you to bypass the older brother. It is worth noting a noticeable decrease in the lag of the GeForce GTX 780 with a corresponding increase in operating frequencies.

Radeon R9 290 again confidently outperforms GeForce GTX 780 in par. The lag behind the flagship Radeon reaches 10% in average fps. According to the minimum parameter, the difference is twice as large, but here the error due to rounding of the final values ​​can play a certain role. In any case, the drawdowns for all participants are so significant that it is impossible to talk about a comfortable gaming process with such indicators.

The younger members change places. Now the Radeon R9 290 is 5% faster than the GeForce GTX 780. When overclocked, the difference between them is only a couple of percent. Increasing the frequencies to 1120/6040 MHz allows you to bypass the Radeon R9 290X and GeForce GTX 780 Ti.

Energy consumption

In terms of power consumption, ASUS is almost the same as the reference GeForce GTX 780 Ti. The most voracious is the Radeon R9 290X. The GeForce GTX 780 is a pleasant surprise, but this is due to the rapid decrease in frequencies as the core warms up. At peak Boost values, significant jumps in power consumption are possible, and after normalizing the core frequency to the base level or slightly higher, the GeForce GTX 780 really turns out to be a very economical solution by the standards of other top-end cards.

conclusions

Based on the results of testing, first of all, I would like to note a slight difference in performance between the Radeon R9 290 and Radeon R9 290X - on average about 10%. The explanations for this result are simple - only 10% of stream processors and texture units are disabled, and the frequency is reduced by a modest 5%. At the same time, neither the memory bus nor the GDDR5 frequency was cut. In the vast majority of applications (9 out of 13), the Radeon R9 290 is faster than the GeForce GTX 780, although the difference between them decreases with overclocking. When comparing overclocked cards, the GeForce GTX 780 manages to snatch victory in seven out of 13 applications, in half of the remaining cases the Radeon advantage is minimal. And if you remember that the top-end modification of the Radeon R9 290 was compared with the reference GeForce GTX 780, it becomes clear that the ratio of forces between ordinary video cards in overclocking may turn out to be slightly better for an NVIDIA representative. This also suggests a minimal difference between the factory overclocked versions, although in many respects everything will depend on the final characteristics of each model. On the side of the Radeon R9 290, there is more video memory. Let's not forget about the presence of the Quiet mode, which will automatically reduce performance on the reference card.

A significant advantage of the Radeon R9 290 is its low price. Initially, the cost of the video card was announced at $ 400, which is $ 150 less than the cost of the Radeon R9 290X. Such a price tag already makes the Radeon R9 290 the best buy among AMD's older graphics solutions and the most profitable offer in the top segment in general. The only problem is that due to the shortage of such video cards, the real price may differ significantly from the stated one. Well, I would not recommend the reference cards for purchase either. Better to overpay a little, but take the version with normal cooling.

These include ASUS R9290-DC2OC-4GD5. This video adapter will not only offer lower operating temperatures and lower noise compared to the reference version, but also initially higher frequencies, providing a modest 5% lag relative to the most powerful Radeon. Judging by the overclocking results of the reference Radeon R9 290X, the frequency potential of the reviewed ASUS can be confidently called good. It will be difficult to fully implement it while maintaining an acceptable noise level, but it will not be difficult to compensate for the very 5% lag behind the flagship. An excellent video card in all respects in the series of the Radeon R9 290 series.

The AMD Radeon R9 290 is the second graphics card to be powered by the new "Hawaii XT" GPU, following the flagship Radeon R9 290X. Inheriting the same Graphics Core Next (GCN) architecture, the next Hawaiian has a 5% lower maximum frequency of the GPU, a 9% reduction in the number of unified shader processors and texture units, as well as an immediate 27% lower recommended price. There are no more differences in the characteristics of these two video cards. It remains to be seen how much the Radeon R9 290 will lag behind the R9 290X in terms of performance, and how this video card looks against the background of its direct price competitor, which is the GeForce GTX 780 on the market at the moment. today's material.

1. Review of the reference video card AMD Radeon R9 290 4 GB

technical characteristics and recommended cost

Technical characteristics of the AMD Radeon R9 290 video card are shown in the table in comparison with the reference versions AMD Radeon R9 290X, AMD Radeon R9 280X and NVIDIA GeForce GTX 780:


PCB design and features

The reference AMD Radeon R9 290 looks exactly the same as the older Radeon R9 290X, you can distinguish these video cards only by the stickers on the back of the PCB:

Its dimensions are also 276x99x39 mm, and among the outputs are the same two DVI (Dual-Link), one HDMI version 1.4a, and one DisplayPort version 1.2:

You will have to forget about CrossFireX connectors on the new Radeon R9, and additional power is organized here through six- and eight-pin connectors, that is, as on the older Radeon R9 290X:

The maximum power consumption of the Radeon R9 290 should also not exceed 275 watts.

Next to the now empty pads of the CrossFireX connectors, a BIOS selection switch is installed, through which you can select the operating mode of the video card cooling system ("Uber mode" or "Quiet mode"):

The PCB of the reference Radeon R9 290 has no differences from the R9 290X:

Here, to power the GPU, five phases are also used with power transistors in a copper DirectFET case, and one phase each for the video memory chips and the power circuit:

Naturally, the IR3567B controller has not undergone any changes either:

Externally, the "Hawaii XT" GPU on the 438 sq. Mm Radeon R9 290 differs only in the week of release - week 20 of 2013 (mid-May):

If you missed the introduction and the spec sheet, remember that on the Radeon R9 290, the GPU contains 2560 universal shader processors, 160 texture units, and 64 raster units. In 3D mode, it should operate at frequencies up to 947 MHz, depending on the degree of load and temperature. When switching to 2D mode, the frequency drops to 300 MHz, and the voltage drops from 1.148 V to 0.977 V. The latter is true for the tested sample and may differ on other samples.

The ASIC quality of the graphics processor of our AMD Radeon R9 290 was even higher than that of the reference Radeon R9 290X (76.2%) and amounts to 83.2%:

After the appearance of AMD Radeon R9 290, all overclockers were interested in activating some of the shaders disabled in the GPU, and some even managed to do this by flashing the BIOS. In order to check the possibility of unlocking the video card to the level of Radeon R9 290X, you need to download and run the Hawaii info utility:

If all the lines “Rx1” contain the value “F8000005”, then the video card can be almost guaranteed to be unlocked by flashing the BIOS from the Radeon R9 290X into it. The methodology is described in detail and step by step here. As you can see, we were out of luck. However, this is a very small loss, as we will very soon be convinced by the test results.

The Radeon R9 290 has four gigabytes of GDDR5 video memory in FCBGA chips, soldered from the front side of the printed circuit board, manufactured by SK Hynix (marking H5GQ2H24AFR R0C):

Despite the fact that the theoretical effective operating frequency of such microcircuits is 6000 MHz, on the AMD Radeon R9 290 the memory operates at 5000 MHz with a peak bandwidth of 320 GB / s. That is, in this regard, the Radeon R9 290 is completely identical to the older R9 290X.

The final specifications of the AMD Radeon R9 290 are as follows:

Let's supplement the review of the video card with a link to its BIOS, and pass on to assessing the efficiency of the cooling system.

cooling system - efficiency

The reference AMD Radeon R9 290 features the exact same cooling system as the Radeon R9 290X with a plastic shroud, aluminum heatsink with copper vapor chamber, steel heat spreader plate and 70mm radial fan:

Thermal pads of two types and gray plastic thermal paste are used as thermal interfaces:

As in the previous week, we used five test cycles of the highly resource-intensive game Aliens vs. Predator (2010) at maximum graphics quality at 2560x1440 pixels with 16x anisotropic filtering and 4x MSAA anti-aliasing:

To monitor temperatures and all other parameters, we used MSI Afterburner version 3.0.0 beta 17 and GPU-Z utility versions 0.7.4-0.7.5. All tests were carried out in a closed case of the system unit, the configuration of which you can see in the next section of the article, at a room temperature of about 25 degrees Celsius.

After running the test in the automatic mode of the cooling fan, we got almost the same picture as on the older Radeon R9 290X, with the only difference that the decrease in the frequency of the GPU and its voltage turned out to be less significant here:

Auto mode

As you can see, the temperature of the GPU already in the third cycle of the test reached 94 degrees Celsius with a fan speed of 2680 rpm (49% power). After that, at first insignificant and short, and then more and more noticeable and prolonged drops in the GPU frequency began immediately. However, the fan speed and noise level of the video card did not increase anymore.

By increasing the fan speed step by step and repeated testing, we have determined that a fan speed of 3070 rpm or 55% of the power is sufficient in our test conditions to avoid a decrease in the GPU frequency. True, in this case, the maximum GPU temperature reached 92 degrees Celsius, that is, it was close to critical:

55% power

Replacing the standard thermal interface of the graphics processor with the high-performance thermal paste Arctic MX-4 made it possible to slightly delay the response of the GPU overheating protection mechanism, which in our test did not have time to occur with automatic fan speed control:

Auto + Arctic MX-4

But, most likely, with a longer load, a simple replacement of thermal paste will not save the reference Radeon R9 290 from overheating with automatic speed control. At the same time, at 55% power, this allowed us to play another 2-3 degrees Celsius at the peak of the load:

55% power + Arctic MX-4

In general, the temperature regime of the reference Radeon R9 290 is practically the same as that of the older Radeon R9 290X, which, however, is logical, given the insignificant differences in GPUs and a completely identical cooler. And yes - the graphics card is just as noisy as the R9 290X.

overclocking potential

Since, as in the case of AMD Radeon R9 290X, the efficiency of the cooling system of the Radeon R9 290 leaves much to be desired, to test for overclocking we were forced to fix the rotation speed of the fan of its cooling system at 60% power or 3380 rpm, and still abandoned the use of increasing the voltage on the core, leaving such experiments for the original video cards. Nevertheless, the overclocking potential of the video card did not disappoint and amounted to +168 MHz or 17.7% for the GPU, and 960 effective megahertz or 19.2% for the video memory:

The final frequencies of the video card in 3D mode were 1115/5960 MHz:

This is hardly less in GPU than overclocking the Radeon R9 290X (1130 MHz), and 200 MHz more than in video memory (5760 MHz). In our opinion, the result is quite decent for the first examples of the reference design.

After overclocking with 60% fan power and a very high noise level, the maximum GPU temperature reached 87 degrees Celsius, and we managed to completely avoid reducing the GPU frequency:

60% power

In general, we were satisfied with the overclocking potential of the Radeon R9 290X, especially since the performance of the video card in this operating mode turned out to be at a very high level. But, more about this in the third section of today's article.

2. Test configuration, tools and testing methodology

Performance testing of video cards was carried out on a system with the following configuration:

Motherboard: Intel Siler DX79SR (Intel X79 Express, LGA 2011, BIOS 0590 dated 07.17.2013);
CPU: Intel Core i7-3970X Extreme Edition 3.5 / 4.0 GHz(Sandy Bridge-E, C2, 1.1 V, 6x256 KB L2, 15 MB L3);
CPU cooling system: Phanteks PH-TC14PЕ (2xCorsair AF140, 900 rpm);
Thermal interface: ARCTIC MX-4;
Video Cards:

EVGA GeForce GTX 780 Superclocked ACX 3GB 863/916/6008 MHz and overclocked to 1032/1085/7348 MHz;
AMD Radeon R9 290X 4 GB 1000/5000 MHz;
AMD Radeon R9 290 4 GB 947/5000 MHz and overclocked to 1115/5960 MHz;
MSI Radeon R9 280X Gaming 3 GB 1050/6000 MHz;

RAM: DDR3 4x8 GB G.SKILL TridentX F3-2133C9Q-32GTX(2133 MHz, 9-11-11-31, 1.6 V);
System disk: SSD 256 GB Crucial m4 (SATA-III, CT256M4SSD2, BIOS v0009);
Disk for programs and games: Western Digital VelociRaptor (SATA-II, 300 GB, 10,000 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 1020 rpm; back - two Noiseblocker NB-BlackSilentPRO PL-1 at 1020 rpm; top - standard 200 mm fan at 400 rpm);
Control and monitoring panel: Zalman ZM-MFC3;
PSU: Corsair AX1200i (1200W), 120mm fan
Monitor: 27 "Samsung S27A850D (DVI-I, 2560x1440, 60 Hz).

From the participants in the previous article, only NVIDIA GeForce GTX 780 Ti was excluded, since it is not at all a competitor to the Radeon R9 290 in terms of cost. All other video cards are again included in the test, but in overclocking, except for the Radeon R9 290 itself, we tested only the EVGA GeForce GTX 780 Superclocked ACX, which, by the way, costs the same today as the Radeon R9 290 based on the reference design.

As before, to reduce the dependence of the performance of video cards on the platform speed, a 32-nm six-core processor with a multiplier of 48, a reference frequency of 100 MHz and an activated Load-Line Calibration function was overclocked to 4.8 GHz when the voltage in the motherboard BIOS was increased to 1. 38 V:

Hyper-Threading Technology is activated. At the same time, 32 GB of RAM operated at a frequency of 2.133 GHz with timings 9-11-11-20_CR1 at a voltage of 1.6125 V.

Testing, which began on December 20, 2013, was carried out under the operating system Microsoft Windows 7 Ultimate x64 SP1 with all critical updates as of the specified date and with the following drivers installed:

motherboard chipset Intel Chipset Drivers - 9.4.4.1006 WHQL from 09.21.2013;
DirectX End-User Runtimes libraries, released on November 30, 2010;
video card drivers on AMD GPUs - AMD Catalyst 13.12 WHQL (13.251.0.0) dated 12/18/2013;
drivers for video cards on graphics processors NVIDIA - GeForce 331.93 Beta from 11/27/2013.

The performance of the video cards was tested in two resolutions: 1920x1080 and 2560x1440 pixels. Two graphics quality modes were used for the tests: "Quality + AF16x" - the quality of textures in the drivers by default with anisotropic filtering at 16x level, and "Quality + AF16x + MSAA 4x (8x)" with anisotropic filtering at 16x level and full-screen anti-aliasing. 4x or 8x, in cases where the average frames per second remained high enough for comfortable gaming. Anisotropic filtering and full-screen anti-aliasing was enabled directly in the game settings. If these settings were absent in games, then the parameters were changed in the control panel of Catalyst or GeForce drivers. Vertical sync was also forcibly disabled there. Apart from the above, no additional changes were made to the driver settings.

The video cards have been tested in two graphics tests and fifteen games, updated to the latest versions as of the date of the start of preparation of this material. Batman: Arkham Origins returned to test list and Battlefield 4 included:

3DMark (2013)(DirectX 9/11) - version 1.2.250.0, testing in the scenes "Cloud Gate", "Fire Strike" and "Fire Strike Extreme";
Unigine Valley Bench(DirectX 11) - version 1.0, maximum quality settings, AF16x and (or) MSAA 4x, resolution 1920x1080;
(DirectX 11) - version 1.1.0, built-in test (battle at Sekigahara) at maximum graphics quality settings and using one of the MSAA 8x modes;
Battlefield 3(DirectX 11) - version 1.4, all graphics quality settings to "Ultra", double sequential pass of the scripted scene from the beginning of the "On the Hunt" mission lasting 110 seconds;
Sniper elite v2 benchmark(DirectX 11) - version 1.05, used Adrenaline Sniper Elite V2 Benchmark Tool v1.0.0.2 BETA maximum graphics quality settings ("Ultra"), Advanced Shadows: HIGH, Ambient Occlusion: ON, Stereo 3D: OFF, Supersampling: OFF, double sequential test run;
Sleeping dogs(DirectX 11) - version 1.5, used Adrenaline Action Benchmark Tool v1.0.2.1, maximum graphics quality settings for all items, Hi-Res Textures pack installed, FPS Limiter and V-Sync disabled, double sequential test run with total anti-aliasing at "Normal" and "Extreme" levels;
Hitman: Absolution(DirectX 11) - version 1.0.447.0, built-in test with graphics quality settings at "Ultra" level, tessellation, FXAA and global illumination are enabled.
Crysis 3(DirectX 11) - version 1.2.0.1000, all graphics quality settings to maximum, blur degree - medium, glare enabled, modes with FXAA and MSAA4x anti-aliasing, double sequential pass of the scripted scene from the beginning of the "Swamp" mission lasting 110 seconds;
Tomb Raider (2013)(DirectX 11) - version 1.1.748.0, Adrenaline Action Benchmark Tool used, quality settings at Ultra level, V-Sync disabled, modes with FXAA and 2xSSAA anti-aliasing, TressFX technology enabled, double sequential pass of the test built into the game;
BioShock Infinite(DirectX 11) - version 1.1.24.21018, Adrenaline Action Benchmark Tool with "Ultra" and "Ultra + DOF" quality settings, double run of the test built into the game;
Metro: Last Light(DirectX 11) - version 1.0.0.15, the built-in game test was used, graphics quality and tessellation settings were set to "Very High", Advanced PhysX technology was enabled, tests with and without SSAA anti-aliasing, double sequential scene pass "D6".
GRID 2(DirectX 11) - version 1.0.85.8679, the built-in game test was used, graphics quality settings to the maximum level for all positions, tests with and without MSAA8x anti-aliasing, eight cars on the Chicago track;
Company of Heroes 2(DirectX 11) - version 3.0.0.11811, double sequential run of the test built into the game with the maximum settings for graphics quality and physical effects;
Total War: Rome II(DirectX 11) - version 1.8.0 build 8891.481024, quality settings at “Extreme” level, V-Sync disabled, SSAA anti-aliasing enabled, double sequential pass of the test built into the game;
ArmA III(DirectX 11) - version 1.08.0.113494, ArmA3Mark test was used, quality settings were at “Ultra” level, V-Sync disabled, double sequential test pass;
Batman: arkham origins(DirectX 11) - version 1.0 update 8, quality settings at "Ultra" level, V-Sync disabled, all effects enabled, all "DX11 Enhanced" functions enabled, Hardware Accelerated PhysX = Normal, double sequential pass of the test built into the game;
Battlefield 4(DirectX 11) - version 1.4, all graphics quality settings to "Ultra", double sequential pass of the scripted scene from the beginning of the "TASHGAR" mission lasting 110 seconds;

If the games implemented the ability to fix the minimum number of frames per second, then it was also reflected in the diagrams. Each test was carried out twice, the best of the two obtained values ​​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 one more time to get a reliable result.

3. Results of performance tests and their analysis

On the diagrams, the results of AMD Radeon R9 290 and R9 290X at nominal frequencies and during overclocking are highlighted in red, the EVGA GeForce GTX 780 Superclocked ACX performance in both operating modes is shown in light green, and the FPS of the MSI Radeon R9 280X Gaming video card is in purple.

3DMark (2013)

The second test scene of the 3DMark package - Cloud Gate - cannot squeeze all the juices out of Hi-End video cards, so the test results in it are not indicative. But Fire Strike in its two modes of operation is perfect for this. AMD Radeon R9 290, as expected, lags behind its older sister, whose name differs only by one character, but easily outstrips its direct competitor in the face of GeForce GTX 780. Having overclocked the Radeon R9 290, we got a video card that is not only ahead of AMD Radeon R9 290X, but also leaving behind the EVGA GeForce GTX 780 with the original cooler and 19.6 ~ 22.3% overclocking. Not a bad start.

Unigine Valley Bench

In Unigine Valley, NVIDIA looks more convincing, but this did not prevent the AMD Radeon R9 290 at higher frequencies from outperforming the more expensive Radeon R9 290X and GeForce GTX 780 in the EVGA performance. Only the overclocking of the latter brings it to the final first place.

The analysis of the results in the games was carried out in the order in which these games were released.

Total War: SHOGUN 2 - Fall of the Samurai

The confrontation between Radeon R9 290 and GeForce GTX 780 in Total War: SHOGUN 2 - Fall of the Samurai ends in a draw both in the nominal operating mode of these video cards and during overclocking. The difference in performance between these video cards is 1-3 frames per second. As for the lag of the new product from the Radeon R9 290X, it lies within 5-9%, and is easily overclocked.

Battlefield 3

We can observe almost the same alignment of forces in the game Battlefield 3:

Sniper elite v2 benchmark

Unlike the previous two games, in Sniper Elite V2, NVIDIA has an advantage:

So AMD Radeon R9 290 is inferior to GeForce GTX 780 from 15 to 17% in each of the test modes. The overclocked Radeon R9 290 is capable of reaching the level of the nominal GeForce GTX 780, while leaving behind the AMD Radeon R9 290X.

Sleeping dogs

On the other hand, Sleeping Dogs is dominated by video cards based on AMD GPUs. The Radeon R9 290 is up to 15% faster than the GeForce GTX 780, 3-7% slower than the X version and immediately 2-9% faster when overclocked.

Hitman: Absolution

Hitman: Absolution is better suited to video cards based on AMD GPUs, as we have seen again today.

Crysis 3

Choosing a graphics card for Crysis 3 from among the Radeon R9 290 and GeForce GTX 780 is very difficult based on performance alone, since the results of these 3D cards are almost equal here. As for the lag of the heroine of today's article from the Radeon R9 290X, it is only 4-7%. That is, the difference, one might say, is negligible. In overclocking, the Radeon R9 290 easily comes out ahead.

Tomb Raider (2013)

Contrary to our expectations, in Tomb Raider AMD did not manage to single-handedly take the lead:

As you can see, the results of the Radeon R9 290 and GeForce GTX 780 are very close to each other. As for the comparison of R9 290 and R9 290X, we will not add anything new to the above.

BioShock Infinite

Another game in which the results of the Radeon R9 290X and GeForce GTX 780 differ slightly is BioShock Infinite. However, we are not talking about the minimum number of frames per second, according to which AMD has an obvious advantage here. However, as we have already noted more than once, this is most likely a feature of the test scene itself.

Metro: Last Light

As usual, let's start with the test results for Metro: Last Light with Advanced PhysX enabled:

Here is a mirrored picture in comparison with BioShock Infinite - the minimum FPS is higher on NVIDIA GeForce, and the average too, and there is only an internal struggle between video cards on AMD GPUs.

When you turn off the "Advanced PhysX" technology, the picture changes:

Now AMD Radeon R9 290 is at least no slower than GeForce GTX 780, and in modes without anti-aliasing it is ahead of its competitor. When pitting the R9 290 against the flagship R9 290X, the balance of power is exactly the same as in most of the previous games we tested.

GRID 2

GRID 2 did not help to determine the leader among the GeForce GTX 780 and Radeon R9 290. Perhaps, when overclocked, AMD is a little faster than the competitor, and also again outperforms the Radeon R9 290X.

Company of Heroes 2

But in Company of Heroes 2, it is difficult to oppose AMD with video cards based on NVIDIA GPUs:

Here the advantage of the Radeon R9 290 over the GeForce GTX 780 is the maximum in the entire test today and can reach 27%. Lagging behind the Radeon R9 290X by 3-10% in the nominal operating mode, the new product outperforms it during overclocking, which makes this video card quite a profitable purchase for overclockers.

Total War: Rome II

In Total War: Rome II, we state the equality of the performance of the Radeon R9 290 and the GeForce GTX 780, as well as the absence of differences in the balance of power between other video cards, and proceed to the ArmA III test.

ArmA III

Here, only the mode with the maximum quality in the resolution of 2560x1440 pixels is relevant for us, ...

... in which the performance of the Radeon R9 290 and the GeForce GTX 780 is almost the same.

Batman: arkham origins

Without MSAA in Batman: Arkham Origins, it is more comfortable to play on GeForce GTX video cards, but when MSAA8x is turned on, AMD video cards look more confident with an increase in the load on the video memory. So, if the Radeon R9 290 was inferior to the GeForce GTX 780, then in the maximum quality mode at a resolution of 2560x1440 pixels it is ahead by as much as 19%.

Battlefield 4

Finally, Battlefield 4 results:

Both without overclocking, and with it, the Radeon R9 290 has a minimal advantage over the GeForce GTX 780 in all test modes. In addition, the new product lags behind the Radeon R9 290X by 5-7% at nominal frequencies, and is ahead of it by 8% after overclocking.

Let's supplement the diagrams with a summary table with test results ...

... and move on to the summary charts, of which there will be six pairs today.

4. Pivot charts

Since we had the Radeon R9 290X benchmark just a week ago, you probably remember that the R9 290X was about 30% faster than the R9 280X. How much ahead of its Radeon R9 290 (without "X"), the first pair of pivot diagrams will tell us:

On average, according to all tests, AMD Radeon R9 290 in the nominal operating mode outperforms the MSI Radeon R9 280X (1050/6000 MHz) by 15.1-19.9% ​​at a resolution of 1920 × 1080 pixels and by 17.9-21.7% at a resolution of 2560 × 1440 pixels. Considering that the difference in the recommended cost between these video cards should be no more than $ 100 (or 25% of the price of the Radeon R9 290), the advantage of the new product is quite natural. However, we recall that the current market situation is very far from what the GPU manufacturers planned to see it.

Next, let's evaluate the lag of the Radeon R9 290 over the older Radeon R9 290X:

In general, the lag of the junior video card on the Hawaii XT turned out to be unexpectedly small: only about 7% in the resolution of 1920 × 1080 pixels and about the same in the resolution of 2560 × 1440 pixels. The maximum difference was noted in one resolution of the game Batman: Arkham Origins, but this is more likely due to the features of the built-in benchmark than the real difference in performance between the Radeon R9 290X and Radeon R9 290.

Now let's see how much the Radeon R9 290 can add when overclocked. As a reminder, we managed to increase the frequency of the GPU by 17.7% and the video memory by 19.2%. The scalability of the Radeon R9 290 is excellent:

At a resolution of 1920 x 1080 pixels, the average performance gain for all games was from 14 to 16%, and at a resolution of 2560 x 1440 pixels - 15-16%. Of course, with such an increase in performance of the overclocked Radeon R9 290, it was not difficult to bypass the nominal Radeon R9 290X:

At frequencies of 1115/5960 MHz, the Radeon R9 290 is ahead of the overclocked Radeon R9 290X by 6-8% at a resolution of 1920 x 1080 pixels and by 7-8% at a resolution of 2560 x 1440 pixels. That is, the overclocking obtained even with the reference version of the video card was enough for us to cover the defective Hawaii XT core.

Now two, perhaps the most interesting comparisons: Radeon R9 290 versus GeForce GTX 780 at the nominal frequencies of both video cards:

So, the Radeon R9 290 has the upper hand in games like Sleeping Dogs, Hitman: Absolution, Company of Heroes 2, and Total War: Rome II. In turn, the GeForce GTX 780 outperforms its rival in Sniper Elite V2, Metro: Last Light with Advanced PhysX, and in three of the four Batman: Arkham Origins modes. In other games, the performance of these two video cards is either the same, or differs quite insignificantly, and if you deduce the geometric mean for all today's gaming tests, then we can say that the speed of the Radeon R9 290 and GeForce GTX 780 is the same.

Thanks to the higher overclocking potential of the original EVGA graphics card, after overclocking, the GeForce GTX 780 slightly narrows the gap with the Radeon R9 290 in the games it lost and slightly increases its superiority where it was. However, the overall picture of the rivalry between these two video cards does not change:

5. Power consumption

The power consumption of the system with various video cards was measured using the Zalman ZM-MFC3 multifunctional panel, which shows the power consumption of the system "from the socket" as a whole (excluding the monitor). The measurement was carried out in 2D mode, during normal work in Microsoft Word or Internet "surfing", as well as in 3D mode. In the latter case, the load was created using four consecutive cycles of the opening scene of the "Swamp" level from the Crysis 3 game at 2560x1440 pixels at maximum graphics quality settings, but without using MSAA anti-aliasing.

Let's compare the power consumption of systems with the video cards tested today:

A system with an AMD Radeon R9 290 at the nominal frequencies of the video card consumes only 16 watts less than a system with a Radeon R9 290X. In overclocking the younger "Hawaii XT" power consumption increases by another 33 watts and reaches 569 watts at peak load. The configuration with GeForce GTX 780 is somewhat more economical both at nominal frequencies and during overclocking. But, frankly speaking, the difference is insignificant for this class of video cards. Overall, a 600 watt PSU should be sufficient for any of the configurations tested today. In 2D mode, the power consumption of the systems is minimal.

Conclusion

Based on the results of the performance tests obtained today and based on the recommended price, AMD Radeon R9 290, in our opinion, looks like a more attractive video card than the older Radeon R9 290X. Lagging behind it by only 7%, the Radeon R9 290 should immediately cost 27% cheaper than the Radeon R9 290X and at the same time easily outperform its sister in overclocking even with the use of an insufficiently efficient and noisy standard cooler. In addition, potential buyers of reference variants of the Radeon R9 290 still have hope of unlocking the blocks disabled in the GPU and thus getting a full-fledged R9 290X at a much more attractive price.

As for the comparison of AMD Radeon R9 290 with GeForce GTX 780, everything is not so simple here. In terms of performance, these video cards are equal. At a cost, nowadays, too. Although, judging by the MSRP, the Radeon R9 290 should be $ 100 cheaper. But this is still not visible either in the Russian market or in foreign online stores. In addition, the GeForce GTX 780 has a significant head start in terms of release time, which led to the fact that now it is very difficult to buy such a model in the reference version (and is it necessary?), And the original ones differ not only in highly efficient coolers, but also in significant increases in frequencies (+100 MHz for GPU and above), as well as impressive overclocking potential of video memory, in contrast to the first copies of the Radeon R9 290. Of course, the original Radeon R9 290 is also on the way, but judging by the announcements of the models of the main companies, the factory overclocking is quite modest, so in a comparison of non-standard GeForce GTX 780 and Radeon R9 290, the former is likely to prevail.

Be that as it may, we should be grateful to AMD for the fact that with the release of its Radeon R9 290 and R9 290X, it has already forced the competitor to reduce prices, as well as launch a more powerful model (we are talking about the GeForce GTX 780 Ti). Increased competition has always played into the hands of end users, and today's situation is no exception. Well, we will look forward to the appearance of the original models of the Radeon R9 290 and R9 290X to see if the manufacturers can eliminate the disadvantages in terms of high noise levels and temperatures.