Watt vs. Frames: AMD R9 290X and NVIDIA GTX 980 in the efficiency test

The energy efficiency of the new GeForce GTX 900 cards has been praised everywhere since the introduction of the Maxwell-based GTX 980. In reality, however, chokes come into play here, which ensure that the Maxwell chip does not need more energy. Nevertheless it is very fast. What happens if you loosen these chokes at NVIDIA and set them at AMD? Both manufacturers have techniques that make this possible. Our test tries to answer these questions.

Intro

There is no question about it: after the introduction of the GeForce GTX 980 and GeForce GTX 970, the power consumption of these two extremely fast graphics cards has been praised on everyone's lips. On the other hand, the power consumption of the fastest AMD graphics cards meets with harsh criticism in general discussions. Reason enough to take a closer look at the whole thing.

However, we also asked ourselves to what extent the two technologies PowerTune (AMD) and Power-Target (NVIDIA) play an important role here. Most graphics cards from the board partners come with relaxed settings, which should generally result in higher power consumption, but also higher 3D performance. But to what extent do you benefit from such relaxed attitudes or even from your own interventions?

We want to clarify this question in today's article and use high-quality measurement technology to relate the determined power consumption of 18 game titles on our benchmark course to performance. A GeForce GTX 980 in the form of the EVGA GTX 980 Superclocked ACX 2.0 and a Radeon R9 290X in the form of the MSI R9 290X Gaming - also overclocked at the factory - serve as the basis for the comparisons.

Test environment

Hardware

EVGA GeForce GTX 980 SC ACX 2.0 (HT4U test / Amazon offers)
MSI Radeon R9 290X Gaming (HT4U test / Amazon offers)

We deliberately chose not to use reference graphics cards from the two chip manufacturers AMD and NVIDIA, but to models from the board partners, which had made a good impression in terms of cooling solution, volume and performance and also had clearly relaxed settings for power consumption go to the start. These criteria are met by the two models we have chosen.

Test station:

CPU: Intel Core i7-3820 - 4 x 3,6 GHz (Turbo / HT: off) [Find it on Amazon*]
Motherboard: ASUS P9X79 Pro (X79 chipset) - BIOS: 3202 12-2012 [Find it on Amazon*]
Memory: 8 GB (4 x 2 GB) ADATA Gaming Series - SPD operation: DDR3-1333 9-9-9-24-1T at 1,5 volts [Find it on Amazon*]
Cooler: Scythe Samurai ZZ Rev B LGA2011 [Find it on Amazon*]

Other hardware:

Seagate Barracuda ST2000DM001 2TB SATA 6 GBit / s [Find it on Amazon*]
LG GSW H20L (Blu-ray / DVD burner) [Find it on Amazon*]
be quiet! Dark Power Pro 950 Watts [Find it on Amazon*]
G.Skill 100 GB SSD as cache drive [Find it on Amazon*]

Teac Floppy Drive / USB Floppy Drive [Find it on Amazon*]
Dell 27 inch monitor [Find it on Amazon*]
Samsung U28590P 4K Monitor [Find it on Amazon*]
Lian Li T60 (open test stand) [Find it on Amazon*]

Meters

Anyone who knows us knows that we have always been and will always be very special in this area. We forego non-calibrated or poorly calibrated hand-held level meters for sound pressure or loudness and rely on one special system with a simulated anechoic room. We generally keep our hands off the 10 euro multimeter and prefer to rely on it calibrated measurement technology from Fluke, it's about volts or amps. If the information from a digitally transmitted temperature is not enough for us or if we want to see the temperature distribution, then we resort to one Thermographic Camera.

So be that as it may - for today's test, these devices are usually less important, because it is about the exact measurement of power consumption.

Up to now, we had relied on a PCI Express adapter that had been specially converted by our engineers and then monitored the current flow through the individual cable strands to the graphics card using clamp ammeters. We have been one step further for well over half a year and are working with a new toy, which now allows us to evaluate not only the minimum and maximum power consumption of a graphics card, but also the average consumption.

There is also a specially developed software that can log and evaluate the measured values. In a separate article we will go into exactly this topic - not today and not now, as we are currently working on Revision B of this new toy. Unfortunately we weren't prepared for the incredible power consumption of a Radeon R9 295X2, which made this step necessary;).

Software

Of course, the last, current drivers were used in the test from both manufacturers, in the present WHQL form. In the games, our 18 titles from the previous test course remain, which is as follows:

Year 2070 V. 1.04.7151 (DX 11 - savegame)
Assassin's Creed IV: Black Flag V. 1.06 (DX 11 - savegame)
Battlefield 4 (DX 11 - savegame)
BioShock: Infinite (DX 11 - savegame)
BRINK V. 1.023692.48133 (OpenGL - Savegame)
Call of Duty: Ghosts (DX11 - Savegame)
Crysis 3 V. 1.0 (DX 11 - savegame)
Far Cry 3 v1.4 (DX 11 - savegame)
Hitman: Absolution (DX 11 - savegame)
Max Payne 3 (DX 11 - savegame)
Metro: Last Light (DX 11 - savegame)
DiRT: Showdown (DX 11: Integrated Benchmark: Miami Route 0)
Splinter Cell: Blacklist (DX11 - Savegame)
The Elder Scrolls V: Skyrim V. 1.8.151.0.7 & HR-Texturepack (DX 9 - Savegames)
Thief (DX 11 - savegame)
The Witcher 2: The Assassins of Kings V. 1.35 (DX 9 - savegame)
tomb raider V. 748.0 (DX 11 - savegame)
Wolfenstein: The New Order* (OpenGL - savegame)

The revised benchmark course is geared towards new blockbusters, but remains with popular titles or representatives of their genre. Under no circumstances should one assume, however, that a strategy game like Anno 2070 may be considered representative of this genre. In most cases, for example, strategy games are designed to be CPU-heavy. Anno and our chosen game sequence represent the opposite. This also applies to other applications. While Sleeping Dogs may show great similarities to GTA, this should not mean that the results of the game can be transferred to GTA.

We are only making a cut over the selected applications and the scenes used for them. We try to make sure that the selected scene corresponds to what the game entails. If we encounter worst-case scenarios, we prefer to choose such a scene, because that is what makes the game flow.

Special features and power saving techniques

Now we could really bring a lot of information in this chapter, but the masses don't want to read it at all. So we limit ourselves to the essentials.

Boost is marketing

We can only repeat it like a mantra: Nothing is “boosted”! Boost is an advertising term from the marketing departments. There is only one maximum GPU clock, and it stays that way until the limits set by the manufacturer are reached.

At NVIDIA, for example, these limits are clearly defined with temperature and clock. There is the so-called temperature limit and the power limit. When these limits are reached, the GPU's clock rate and voltage are reduced. When the temperature target (GTX 980 = 80 ° C) is reached, it is throttled; when the target power consumption (GTX 980 = 180 watts) is reached, it is also throttled. This is done by a chip on the graphics card in conjunction with the graphics card driver.

AMD is a bit more complicated than that. The PowerTune “feature” mentioned there monitors the utilization of the chip by means of several units. An (now) factor that comes into play is the temperature (R9 290X = 94 °C), but also the power consumption (R9 290X = 250 watts).


AMD PowerTune	NVIDIA Power Target

However, the throttling takes place relatively differently. While with NVIDIA graphics cards it can be observed with this technology that correct step formations arise, which absorb the load limits, with AMD it is observed peaks that jump up and down in a very short time without noticing this in the game flow. In practice it is difficult to compare the two techniques.

Manual intervention

Manual interventions in NVIDIA-based models can be made relatively precisely, but only with external tools such as MSI Afterburner. In the meantime, almost every graphics card partner from NVIDIA has its own tool, but these are mostly based on the programming of the RivaTuner. No wonder, because “Unwinder” (inventor of the RivaTool) is the programmer and supplier for many manufacturers and their tools. And so this programming is hidden behind the variant from ASUS as well as behind that of EVGA or MSI. The tools are usually not linked to a particular manufacturer and can also be used on other NVIDIA graphics cards.

If you know exactly the power limit set by the board partner for the NVIDIA graphics card, you can manually reduce or increase the limit or increase within the permitted slider using such tools by specifying a percentage (quite precisely).

Unfortunately, AMD is completely different. There is even a percentage in the driver. But as AMD made clear earlier, this has nothing to do with a percentage bond to the power consumption.

Unfortunately, the user can only try it out here if he wants to reduce the power consumption manually. In most cases, the endeavors of the users of tools or driver options are mainly to achieve the maximum possible clock rate and the maximum possible performance.

Approach in the test

Game scenes

Of course we use our usual game course with a total of 18 titles and the respective scenes for today's consideration. For each game and the usual scene for us, we have determined the maximum power consumption of each of today's two test graphics cards with our measurement technology and will also report these results later.

But: Just because Crysis 3, for example, in this selected game scene with the test subject, claims 190 watts for itself in the following presentations, this statement should in no way be understood to mean that the game would not be more demanding. The scene only represents one area in the game that we selected and which we thought would be suitable for showing representative benchmarks. So we didn't set out to look for other places in the games where the measuring system shows a significantly higher level of power consumption.

Throttling of the cards to different TDP ranges

As mentioned at the beginning, we not only want to show the total or maximum power consumption, but also how the two test candidates behave at different levels of power consumption. To do this, we first determined the maximum using two applications (Anno 2070 and Tom Clancy's HAWX) and then reduced the power consumption through driver or tool intervention so that the desired throttling was achieved as a maximum.

So in order not to exceed a maximum power consumption of 9 watts for the MSI R290 4X Gaming 180G, we turned the power down using the PowerTune control until our measurement no longer showed any higher values. In practice, this meant a minus of 46 and a fluctuating GPU clock in the range of 820 to 850 MHz.

NVIDIA Power Target	Maximum power consumption	Clock rates EVGA GTX 980 SC ACX 2.0
124 %	224 watts,	1.418 MHz
117 %	210 watts,	1.380 to 1.400 MHz
108 %	195 watts,	1.366 to 1.380 MHz
100 %	180 watts,	1.329 MHz
92 %	165 watts,	1.278 to 1.291 MHz

Test scenes: Anno 2070 / Tom Clancy's HAWX

AMD PowerTune	Maximum power consumption	Clock rates MSI R9 290X Gaming 4G
0	300 watts,	1.030 MHz
- 15	260 watts,	1.010 to 1.030 MHz
- 30	225 watts,	940 to 960 MHz
- 36	210 watts,	905 to 920 MHz
- 41	195 watts,	880 to 895 MHz
- 46	180 watts,	820 to 847 MHz
- 50	170 watts,	790 to 820 MHz

Test scenes: Anno 2070 / Tom Clancy's HAWXPlease do not misunderstand and generalize the above information on the clock rates! The information relates purely to the two titles mentioned. Games, for example, which are less demanding in terms of power consumption, are of course fired at higher clock rates by the technology. Let's take the OpenGL title Brink as an example - AMD's PowerTune set clock rates of around 50 to 900 MHz with the -920 setting, but stayed within the targeted level of power consumption.

evaluations

Total power consumption

Our current game test course comprises a total of 18 games, which we all consulted, but each linked to our test scene. As already mentioned, we limit ourselves to a resolution of 2.560 x 1.440 here. With some titles we tested them in different settings (with MSAA / SSAA or without). We averaged the results of individual games in different settings.

The two graphics cards from EVGA and MSI used are inherently overclocked models, with the MSI R9 290X Gaming 4G only overclocking its GPU by 30 MHz, the EVGA representative in the form of the GTX 980 SC ACX 2.0 clocks around 100 MHz higher the GPU applies.

In order to be able to carry out the following measurements at all, we made sure by manual intervention that the graphics cards are not throttled, i.e. that they stay on their clock. In the case of the NVIDIA representative, we only had to increase the power limit; the AMD representative did not need to intervene at all - the card's clock rate was never throttled, which is due to the different technologies of AMD and NVIDIA.

So we then determined the maximum power consumption of the graphics card for each of our test scenes and each game, whereby we only created a time span of 15 minutes.

The interventions in the throttling mean that the EVGA representative can use a maximum power consumption of 224 watts for both cards. The MSI card, on the other hand, can show up to 300 watts without any problems (but it doesn't in the examples here). In practical terms, however, there are only a few applications that bring this GTX 980 close to the throttle, and there are no representatives (or test scenes) that drive the R9 290X to its maximum.

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)	EVGA GTX 980 SC ACX 2.0 (FPS)	AMD MSI R9 290X Gaming (FPS)
Year 2070	224	262	69	60
Assassin's Creed: Black Flag	184	235	41	41
Battlefield 4	223	236	61	50
BioShock: Infinite	177	251	64	61
Brink	191	198	120	85
CoD: Ghosts	194	232	50,5	40
Crysis 3	190	236	67	63
DiRT: Showdown	182	216	110	99,5
Far Cry 3	190	246	51	45
Hitman: Absolution	188	247	52	51
Max Payne 3	185	238	56,5	47
Metro: Last Light	210	252	57,5	44
SC: Blacklist	170	251	41	34,5
TES V: Skyrim	206	225	106	97
The Witcher 2	190	235	68	63,5
Thieves (2014)	180	226	70,5	51,5
tomb raider	204	251	47	38
Wolfenstein: The New Order	170	174	38	31

In order to be able to average the results fairly and evaluate them as a whole, we set the results of the AMD representative at 100% and determined the percentage difference to NVIDIA. In the overall comparison there is no surprise then. Without throttling, the EVGA GTX 980 SC ACX 2.0 beats the MSI R9 290X Gaming 4G with around 17% advantage in performance. The immediately determined advantage in power consumption is also evident with a little more than 17% for the EVGA graphics card.

As a result, AMD and NVIDIA cards are principally provided by the manufacturer with a throttle - in both cases with different technology. The maximum power consumption of a GTX 980 graphics card would be 180 watts, typically only 165 watts after a short time (via the throttling). AMD's specification for the R9 290 is 250 watts. Based on this information, NVIDIA's GTX 980 would have advantages of around 39% with the maximum TDP and even advantages of 51% with the help of the clock throttling. Our results - as shown - do not provide that for the time being.

In the following individual evaluations, we will go into how interventions in Power-Target or PowerTune can affect the performance of the models, because in principle AMD could also slow down its GPU in such a way that it only plays at the level of the power consumption of NVIDIA. Where is the sweet spot and does it even exist?

Individual evaluation: Brink

Game	Brink
Developer	Splash damage
Publisher	Bethesda Softworks
publication	13 May 2011
Genre	Ego shooter
Graphics engine	modified id-tech 4
DirectX path / API	OpenGL
Age rating USK	From 16 years
Benchmark measurement	Fraps / savegame
Test area	Hostage rescue
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail

In-game test scene

	Brink
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	120,01
EVGA GTX 980 [Max. 210 watts]	119,97
EVGA GTX 980 [Max. 195 watts]	119,94
EVGA GTX 980 [Max. 180 watts]	119,94
EVGA GTX 980 [Max. 165 watts]	118,11
MSI R9 290X [Max. 300 watts]	81,20
MSI R9 290X [Max. 260 watts]	81,10
MSI R9 290X [Max. 225 watts]	81,07
MSI R9 290X [Max. 210 watts]	81,05
MSI R9 290X [Max. 195 watts]	78,86
MSI R9 290X [Max. 180 watts]	76,47
MSI R9 290X [Max. 170 watts]	74,06

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Brink	191	198

The test scene shows that the power consumption of both graphics cards is about the same - just under 200 watts - but of course with certain advantages for the NVIDIA graphics card. With the power consumption shown, it is also clear that both representatives can no longer improve beyond the 200 watt mark in terms of performance. Deviations can be clearly explained by measurement inaccuracies.

It is therefore clear that the NVIDIA representative has the biggest difference between 165 and 180 watts, but there is no longer a larger jump to 190 watts. The performance is increased by 2%, but the power consumption is increased by 9% (180 W) or 15% (unthrottled). The sweet spot with the GTX 980 clearly seems to be the NVIDIA default setting.

With AMD, the difference between the minimum set power consumption and the maximum looks a little different. The AMD representative can gain a maximum of 9% in performance. However, the difference in power consumption shows an increase of around 17%. The sweet spot of the AMD card would be 195 watts in this area. However, it must be noted that the AMD graphics card scales equally well with increasing power consumption.

Individual evaluation: Wolfenstein: The New Order

Game	Wolfenstein: The New Order
Developer	Machine Games
Publisher	Bethesda
publication	May 2014
Genre	Ego shooter
Age rating USK	From 18 years
Graphics engine	id Tech 5
DirectX path	OpenGL
Benchmark measurement	Fraps / savegame
Test area	Chapter 9, intro
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail
	HT4U test
	Find on Amazon*

Image: AMD's Tonga GPU - Radeon R9 285 in the test

Test scene of the game

	Wolfenstein: The New Order
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 210 watts]	38,18
EVGA GTX 980 [Max. 195 watts]	38,16
EVGA GTX 980 [Max. 224 watts]	38,16
EVGA GTX 980 [Max. 180 watts]	38,14
EVGA GTX 980 [Max. 165 watts]	37,87
MSI R9 290X [Max. 180 watts]	30,88
MSI R9 290X [Max. 195 watts]	30,82
MSI R9 290X [Max. 260 watts]	30,80
MSI R9 290X [Max. 225 watts]	30,68
MSI R9 290X [Max. 300 watts]	30,67
MSI R9 290X [Max. 210 watts]	30,26
MSI R9 290X [Max. 170 watts]	28,51

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Wolfenstein: The New Order	170	174

Wolfenstein: The New Order can be power-hungry, but there are no real challenges to be found in the area of power consumption itself, because with around 170 watts in our test scene, it is also clear that the two representatives can hardly show any differences, despite manual restrictions.

Despite its throttle, NVIDIA should mostly be able to call up the maximum performance here. Unthrottled, there is only a 1% increase in performance. A default model without restrictions will manage this at any time - as long as the temperature limit does not intervene.

If we throttle the AMD representative into the range of 170 watts, there is a significant limitation in performance. Loosening the TDP then allows an 8% increase in 3D performance. But there is no more information to be derived from Wolfenstein here.

Individual evaluation: The Elder Scrolls V: Skyrim

Game	The Elder Scrolls V: Skyrim
Developer	Bethesda Game Studios
Publisher	Bethesda Softworks
publication	(March 2012)
Genre	role playing game
Age rating USK	From 16 years
Graphics engine	Creation engine
DirectX path	DirectX 9
Benchmark measurement	Fraps / savegame
Test area	Steinhuebel
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail, FXAA, High Resolution Texture Pack
	Order from Amazon*

Benchmark scene in the test

	TES V - Skyrim
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	106,53
EVGA GTX 980 [Max. 210 watts]	106,43
EVGA GTX 980 [Max. 195 watts]	105,75
EVGA GTX 980 [Max. 180 watts]	104,51
EVGA GTX 980 [Max. 165 watts]	102,16
MSI R9 290X [Max. 300 watts]	96,99
MSI R9 290X [Max. 260 watts]	96,74
MSI R9 290X [Max. 225 watts]	92,52
MSI R9 290X [Max. 210 watts]	89,59
MSI R9 290X [Max. 195 watts]	86,80
MSI R9 290X [Max. 180 watts]	84,27
MSI R9 290X [Max. 170 watts]	80,05

	FPS

	TES V - Skyrim
	2560 x 1440 [8xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	104,92
EVGA GTX 980 [Max. 210 watts]	104,87
EVGA GTX 980 [Max. 195 watts]	103,72
EVGA GTX 980 [Max. 180 watts]	101,05
EVGA GTX 980 [Max. 165 watts]	99,05
MSI R9 290X [Max. 260 watts]	92,12
MSI R9 290X [Max. 300 watts]	92,07
MSI R9 290X [Max. 225 watts]	89,17
MSI R9 290X [Max. 210 watts]	86,25
MSI R9 290X [Max. 195 watts]	83,03
MSI R9 290X [Max. 180 watts]	81,00
MSI R9 290X [Max. 170 watts]	77,63

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
TES V: Skyrim	206	225

With Skyrim, the power consumption of the two test subjects is now slightly further apart for the first time. AMD has to pay for its best performance with 225 watts, NVIDIA already 206 watts are sufficient here - mind you, always based on our test scene.

The NVIDIA representative shows the best scaling in the range between 165 and 195 watts. In these levels you can increase by up to 5% - in order to still be able to take the last performance point, you need a relaxation to 210 watts, but you only gain 1%. The typical 180 watts of a GTX 980 are basically well chosen here. In this case, the GPU still scales equally well, or even a little better, up to 195 watts.

The Radeon representative continues to scale identically within the scope of its throttle levels. Actually, at 225 watts, there should be no throttling at all, but there are obviously still fluctuations here that PowerTune collects and then disappeared in the next level of relaxation. Starting with the lowest power consumption level of 170 watts up to around 230 watts, the MSI graphics card increases its performance by a whopping 19%. A sweet spot cannot therefore be identified.

Individual evaluation: The Witcher 2

Game	The Witcher 2: The Assassins of Kings
Developer	CD Projekt RED
Publisher	CD project, Atari
publication	17 May 2011
Genre	RPG, fantasy
Graphics engine	RED engine
DirectX path	DirectX 9
Age rating USK	From 16 years
Benchmark measurement	Fraps / savegame
Test area	barricade
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail

In-game test scene

	Witcher 2 - Assassins of Kings
	2560 x 1440 [4xSSAA / 16xAF]

MSI R9 290X [Max. 300 watts]	40,00
MSI R9 290X [Max. 260 watts]	40,00
EVGA GTX 980 [Max. 224 watts]	39,76
MSI R9 290X [Max. 225 watts]	39,76
EVGA GTX 980 [Max. 195 watts]	39,75
MSI R9 290X [Max. 195 watts]	39,75
EVGA GTX 980 [Max. 210 watts]	39,74
MSI R9 290X [Max. 210 watts]	39,74
EVGA GTX 980 [Max. 180 watts]	38,60
MSI R9 290X [Max. 180 watts]	38,60
EVGA GTX 980 [Max. 165 watts]	37,51
MSI R9 290X [Max. 170 watts]	37,51

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
The Witcher 2	190	235

The Witcher 2 also shows clear differences in power consumption between the two rivals. The AMD representative needs around 235 watts for the best performance, the EVGA card with NVIDIA GPU is content with around 190 watts. However, the performance is pretty much the same and this game has some nice information to offer when we look at it.

The EVGA GTX 980 SC scales identically from the smallest throttle level up to the 195 watt level, and similar to the previous one, no sweet spot can be seen in terms of power consumption for this GTX 980 implementation. After that, nothing happens, precisely because the game scene is no longer demanding on the power consumption of the card.

Although this test scene demanded about 235 watts unthrottled from the AMD representative, this is not the same when it comes to scaling. From the lowest point (around 170 watts) to the throttle level, you scale the same and practically identically to the NVIDIA rival. After that, however, more or less nothing happens. A maximum of 1% can still be increased in performance, but then you have to throw 25 watts more into the balance. So here you could definitely see a sweet spot at 195 watts of power consumption.

But: This circumstance clearly seems to be due to the in-game supersampling! If we look at runs without supersampling, the picture with NVIDIA remains very similar, but the R9 from MSI now scales more or less the same across all throttle levels:

	Witcher 2 - Assassins of Kings
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	95,90
EVGA GTX 980 [Max. 210 watts]	95,28
EVGA GTX 980 [Max. 195 watts]	95,26
EVGA GTX 980 [Max. 180 watts]	92,60
EVGA GTX 980 [Max. 165 watts]	90,36
MSI R9 290X [Max. 300 watts]	87,58
MSI R9 290X [Max. 260 watts]	87,58
MSI R9 290X [Max. 225 watts]	84,47
MSI R9 290X [Max. 210 watts]	82,57
MSI R9 290X [Max. 195 watts]	80,07
MSI R9 290X [Max. 180 watts]	78,14
MSI R9 290X [Max. 170 watts]	73,59

	FPS

Individual evaluation: Anno 2070

Game	Year 2070
Developer	Related Designs / Ubisoft Blue Byte
Publisher	Ubisoft
publication	17 November 2011
Genre	strategy game
Age rating USK	From 6 years
Graphics engine	InitEngine
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	On the trail of the truth
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail

In-game test scene

	Year 2070
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	68,73
EVGA GTX 980 [Max. 210 watts]	68,20
EVGA GTX 980 [Max. 195 watts]	67,35
EVGA GTX 980 [Max. 180 watts]	65,69
EVGA GTX 980 [Max. 165 watts]	63,55
MSI R9 290X [Max. 300 watts]	58,75
MSI R9 290X [Max. 260 watts]	58,73
MSI R9 290X [Max. 225 watts]	54,57
MSI R9 290X [Max. 210 watts]	52,45
MSI R9 290X [Max. 195 watts]	50,02
MSI R9 290X [Max. 180 watts]	47,67
MSI R9 290X [Max. 170 watts]	43,92

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Year 2070	224	262

Anno 2070 pushes both test subjects to their limits. The selected scene reaches the maximum possible power consumption of 224 watts with the EVGA card, and at least 262 watts with the MSI card.

With the GeForce GTX 980 representative, however, the scaling comes to a standstill as soon as the throttle level is loosened above 195 watts. Below that, you immediately improve in performance. So 195 watts are the best point in performance, but are bought at the price of 18% higher power consumption with only 6% performance gain.

The MSI representative shows that there is no really explosive point in terms of power consumption. It can be pointed out that the 10-watt increase from 170 to 180 watts results in a performance increase of 9% and the subsequent increases are usually only 4 to 5%. Overall, however, the graphics card can increase its performance by around 34%, and this is quite linear. The difference in power consumption between 170 and 250 watts is then 47% - justifiable in terms of the performance increases.

From an energy point of view, 180 watts would be optimal, but 3 watts would be optimal with regard to the best 250D performance.

Individual evaluation: Assassin's Creed IV: Black Flag

Game	Assassin's Creed IV: Black Flag
Developer	Ubisoft
Publisher	Ubisoft
publication	Nov 2013 (PC)
Genre	Action adventure
Age rating USK	From 16 years
Graphics engine	AnvilNext
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Sequence 4 - Reminder 2
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail, DirectX 11; PhysX: Off
HT4U test	Order from Amazon*

Picture: NVIDIA GeForce GTX 750 Ti in the test

In-game test scene

	Assassins Creed IV: Black Flag
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	40,8
EVGA GTX 980 [Max. 210 watts]	40,8
EVGA GTX 980 [Max. 195 watts]	40,8
MSI R9 290X [Max. 300 watts]	40,6
MSI R9 290X [Max. 260 watts]	40,5
EVGA GTX 980 [Max. 180 watts]	40,2
MSI R9 290X [Max. 225 watts]	39,6
EVGA GTX 980 [Max. 165 watts]	39,5
MSI R9 290X [Max. 210 watts]	38,8
MSI R9 290X [Max. 195 watts]	37,6
MSI R9 290X [Max. 180 watts]	36,4
MSI R9 290X [Max. 170 watts]	35,1

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Assassin's Creed IV: Black Flag	184	235

Once again, we see a very large difference between the two representatives in terms of maximum power consumption with maximum performance. The NVIDIA representative needs around 185 watts here, the AMD representative around 235 watts. Of course, this has a clear effect on the course of our following diagram for scaling.

From 165 to 180 watts, the NVIDIA graphics card can increase performance by 2%, in the next level up to 195 watts only 1% is added. NVIDIA's target of 180 watts as maximum board power seems to be confirmed by this title as "reasonably chosen".

With AMD, the scaling from 170 to 225 watts is constant, and only then does the diagram curve flatten out clearly. The representative brings out the best performance at 210 to 225 watts.

Individual evaluation: Battlefield 4

Game	Battlefield 4
Developer	EA Digital Illusions CE
Publisher	Electronic Arts
publication	Dezember 2013
Genre	Ego shooter
Age rating USK	From 18 years
Graphics engine	Frostbite 3
DirectX path	DirectX 10 / DirectX 11 / Mantle
Benchmark measurement	Fraps / savegame
Test area	Level 6: Tashgar - Checkpoint 5
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail, DX 11
HT4U test	Order from Amazon

In-game test scene

	Battlefield 4
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	70,0
EVGA GTX 980 [Max. 210 watts]	68,4
EVGA GTX 980 [Max. 195 watts]	66,4
EVGA GTX 980 [Max. 180 watts]	64,4
EVGA GTX 980 [Max. 165 watts]	61,4
MSI R9 290X [Max. 300 watts]	55,9
MSI R9 290X [Max. 260 watts]	55,8
MSI R9 290X [Max. 225 watts]	53,7
MSI R9 290X [Max. 210 watts]	52,8
MSI R9 290X [Max. 195 watts]	51,5
MSI R9 290X [Max. 180 watts]	51,2
MSI R9 290X [Max. 170 watts]	48,2

	FPS

	Battlefield 4
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	52,5
EVGA GTX 980 [Max. 210 watts]	52,3
EVGA GTX 980 [Max. 195 watts]	51,4
EVGA GTX 980 [Max. 180 watts]	50,3
EVGA GTX 980 [Max. 165 watts]	49,3
MSI R9 290X [Max. 300 watts]	44,5
MSI R9 290X [Max. 260 watts]	44,4
MSI R9 290X [Max. 225 watts]	43,8
MSI R9 290X [Max. 210 watts]	41,9
MSI R9 290X [Max. 195 watts]	40,9
MSI R9 290X [Max. 180 watts]	40,7
MSI R9 290X [Max. 170 watts]	38,9

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Battlefield 4	223	236

In Battlefield 4 and our test scene, the maximum power consumption of our test candidates is much closer together. While the EVGA graphics card practically reached its maximum of 224 watts, the MSI R9's 236 watts.

The NVIDIA representative can once again say that the scaling in the individual levels is very identical and that practically no performance can be implemented until the last power choke (results with 4-fold MSAA).

We cannot explain with certainty why the 180 to 195 watt choke on the AMD graphics card does not show a clear scaling here. AMD's PowerTune doesn't just work as a choke for the power consumption and could be the explanation for the result, because regardless of this measured value, we see a continuous increase in performance up to the 225 watt level. Only then does the result turn out to be unspectacular and dispensable. 180 watts show up here as the best point from the point of view of energy consumption; It is 225 watts with a view to the 3D performance.

Individual evaluation: BioShock: Infinite

Game	BioShock: Infinite
Developer	Irrational Games, 2K Marin, Human Head Studios
Publisher	2K Games
publication	March 26, 2013
Genre	First person shooter with fantasy elements
Graphics engine	Unreal engine 3
DirectX path	DirectX 10/11
Age rating USK	From 18 years
Benchmark measurement	Fraps / savegame
Test area	Finkton Proper
Runtime benchmark	10 seconds
Benchmark settings	System settings Maximum & FXAA
	HT4U test
	Order from Amazon*

Image: NVIDIA GeForce GTX Titan from ASUS and Gigabyte in the test

Benchmark scene in the test

	Bioshock: Infinite
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 210 watts]	63,47
EVGA GTX 980 [Max. 224 watts]	63,42
EVGA GTX 980 [Max. 195 watts]	63,36
EVGA GTX 980 [Max. 180 watts]	63,35
EVGA GTX 980 [Max. 165 watts]	62,50
MSI R9 290X [Max. 300 watts]	61,15
MSI R9 290X [Max. 260 watts]	61,14
MSI R9 290X [Max. 225 watts]	57,58
MSI R9 290X [Max. 210 watts]	55,52
MSI R9 290X [Max. 195 watts]	53,45
MSI R9 290X [Max. 180 watts]	51,31
MSI R9 290X [Max. 170 watts]	46,74

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
BioShock: Infinite	177	251

Rarely in this comparison are the differences in maximum power consumption as great as with BioShock. The EVGA graphics card needs a maximum of 180 watts, the AMD graphics card a maximum of around 250 watts.

As a result, NVIDIA's curve shape barely shows up in the diagram - but how should it, based on the stated maximum consumption? 165 watts compared to 180 watts mean just 1% performance advantage. You can do without that because we are not even dealing with an fps.

The AMD card is completely different. Throttled to 170 watts it is not at all. A loosening of 10 watts already brings 10% more power with it. Then you scale gently in about 4% steps up to the 260 watt choke. If you want to speak of a sweet spot with this R9-290X implementation, you could then at most name 180 watts in terms of energy. The best overall performance is only available beyond 250 watts.

Individual evaluation: Call of Duty: Ghosts

Game	Call of Duty: Ghosts
Developer	Infinity Ward
Publisher	Activision
publication	November 2013
Genre	Ego shooter
Age rating USK	From 18 years
Graphics engine	IW engine / Havok
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Level: The Hunted - Checkpoint 3
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail, DX 11
HT4U test	Order from Amazon

In-game test scene

	Call of Duty: Ghosts
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	61,15
EVGA GTX 980 [Max. 210 watts]	61,02
EVGA GTX 980 [Max. 195 watts]	60,91
EVGA GTX 980 [Max. 180 watts]	60,18
EVGA GTX 980 [Max. 165 watts]	59,07
MSI R9 290X [Max. 300 watts]	47,83
MSI R9 290X [Max. 260 watts]	47,82
MSI R9 290X [Max. 225 watts]	46,71
MSI R9 290X [Max. 210 watts]	45,21
MSI R9 290X [Max. 195 watts]	43,67
MSI R9 290X [Max. 180 watts]	42,73
MSI R9 290X [Max. 170 watts]	41,11

	FPS

	Call of Duty: Ghosts
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	39,48
EVGA GTX 980 [Max. 195 watts]	39,44
EVGA GTX 980 [Max. 210 watts]	39,44
EVGA GTX 980 [Max. 180 watts]	38,37
EVGA GTX 980 [Max. 165 watts]	37,06
MSI R9 290X [Max. 260 watts]	32,27
MSI R9 290X [Max. 300 watts]	32,26
MSI R9 290X [Max. 225 watts]	31,64
MSI R9 290X [Max. 210 watts]	30,73
MSI R9 290X [Max. 195 watts]	29,80
MSI R9 290X [Max. 180 watts]	29,09
MSI R9 290X [Max. 170 watts]	28,13

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
CoD: Ghosts	194	232

Clear differences in the maximum power consumption can also be seen in Call of Duty: Ghosts and the selected test scene. The GCN architecture of the Radeon graphics card allows itself around 230 watts, the Maxwell architecture of the GTX graphics card is satisfied with around 190 watts.

With the value of 194 watts as the maximum in this test, it is also clear that no changes can occur at levels above 195 watts (outside of measurement inaccuracies). The NVIDIA graphics card tastes the 180 watts clearly better than the 165 watts (typical power consumption). Nevertheless, the relaxation to 195 watts still creates an increase of 3%, but starting from the minimum you gain the least 3D performance in the last level.

The AMD representative shows an image that is seen all too often and increases quite linearly to around 250 watts, which is the correct value with regard to the 3D performance and the requirements of the game.

Individual evaluation: Crysis 3

Game	Crysis 3
Developer	Crytek
Publisher	Electronic Arts
publication	21 February 2013
Genre	Ego shooter
Graphics engine	CryENGINE 3
DirectX path	DirectX 9/11
Age rating USK	From 18 years
Benchmark measurement	Fraps / savegame
Test area	Mission 5: River - Red Star Rising
Runtime benchmark	10 seconds
Benchmark settings	Default system and textures: high
	Order from Amazon

In-game test sceneIn the following diagrams, 1 x AA stands for deactivated antialiasing and the post-processing filter FXAA. 2 x AA stands for the special level 4 x SMAA. The game relies on double, regular anti-aliasing (MSAA) and additional filters. The designation 4 x AA corresponds to the usual quadruple anti-aliasing (MSAA).

	Crysis 3
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	85,50
EVGA GTX 980 [Max. 210 watts]	85,37
EVGA GTX 980 [Max. 195 watts]	84,97
EVGA GTX 980 [Max. 180 watts]	84,55
EVGA GTX 980 [Max. 165 watts]	83,76
MSI R9 290X [Max. 300 watts]	78,36
MSI R9 290X [Max. 260 watts]	78,31
MSI R9 290X [Max. 225 watts]	75,60
MSI R9 290X [Max. 210 watts]	72,41
MSI R9 290X [Max. 195 watts]	71,17
MSI R9 290X [Max. 180 watts]	68,20
MSI R9 290X [Max. 170 watts]	65,96

	FPS

	Crysis 3
	2560 x 1440 [2xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	60,95
EVGA GTX 980 [Max. 210 watts]	60,91
EVGA GTX 980 [Max. 195 watts]	60,71
EVGA GTX 980 [Max. 180 watts]	60,57
EVGA GTX 980 [Max. 165 watts]	59,85
MSI R9 290X [Max. 300 watts]	57,59
MSI R9 290X [Max. 260 watts]	57,54
MSI R9 290X [Max. 225 watts]	56,12
MSI R9 290X [Max. 210 watts]	54,76
MSI R9 290X [Max. 195 watts]	53,51
MSI R9 290X [Max. 180 watts]	52,53
MSI R9 290X [Max. 170 watts]	50,25

	FPS

	Crysis 3
	2560 x 1440 [4xAA / 16xAF]

MSI R9 290X [Max. 260 watts]	54,06
MSI R9 290X [Max. 300 watts]	53,89
MSI R9 290X [Max. 225 watts]	52,88
EVGA GTX 980 [Max. 224 watts]	52,60
EVGA GTX 980 [Max. 195 watts]	52,48
EVGA GTX 980 [Max. 210 watts]	52,42
EVGA GTX 980 [Max. 180 watts]	51,97
MSI R9 290X [Max. 210 watts]	51,78
EVGA GTX 980 [Max. 165 watts]	51,20
MSI R9 290X [Max. 195 watts]	50,33
MSI R9 290X [Max. 180 watts]	49,31
MSI R9 290X [Max. 170 watts]	47,85

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Crysis 3	190	236

Again we see a roughly the same difference in power consumption - a maximum of around 190 watts for NVIDIA, a maximum of around 230 watts for AMD. In some titles this picture is repeated very clearly. Again it becomes clear that we can hardly derive any results in the following diagram from the 195 watt level.

The best performance in connection with the TDP brings the NVIDIA graphics card again in the range of the TDP specification of the GPU manufacturer of 180 watts. However, at the 165 watt level you lose little and no feel at all. However, we have to make it clear that this only applies to the selected game scene. In Crysis 3, with this preselection with such a GTX 980, the frame range can sometimes drop into the 40 fps region, and those who cannot / do not want to play with such frame rates are better off at 180 Watt canceled.

On the other hand, the best cut between performance and power consumption by the AMD representative is 225 watts. Up to this power consumption, the test subject then again scales cleanly and linearly with the increasing TDP preselection. 210 watts would still be acceptable, but AMD models have so far not given any options to select these criteria yourself.

Individual evaluation: Far Cry 3

Game	Far Cry 3
Developer	Ubisoft
Publisher	Ubisoft
publication	November 2012
Genre	Ego shooter
Age rating USK	From 16 years
Graphics engine	Dunia Engine 2 and Havok Physics
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Fruits of the jungle
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail (Ultra), SSAO: SSAO, DirectX 11
	Order from Amazon*

Benchmark scene in the test

	Far Cry 3
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	69,0
EVGA GTX 980 [Max. 210 watts]	68,9
EVGA GTX 980 [Max. 195 watts]	68,9
EVGA GTX 980 [Max. 180 watts]	67,9
EVGA GTX 980 [Max. 165 watts]	67,1
MSI R9 290X [Max. 300 watts]	62,0
MSI R9 290X [Max. 260 watts]	61,9
MSI R9 290X [Max. 225 watts]	58,7
MSI R9 290X [Max. 210 watts]	57,1
MSI R9 290X [Max. 195 watts]	55,2
MSI R9 290X [Max. 180 watts]	53,5
MSI R9 290X [Max. 170 watts]	50,7

	FPS

	Far Cry 3
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 210 watts]	48,9
EVGA GTX 980 [Max. 195 watts]	48,9
EVGA GTX 980 [Max. 224 watts]	48,9
EVGA GTX 980 [Max. 180 watts]	48,0
EVGA GTX 980 [Max. 165 watts]	47,2
MSI R9 290X [Max. 260 watts]	41,6
MSI R9 290X [Max. 300 watts]	41,6
MSI R9 290X [Max. 225 watts]	39,6
MSI R9 290X [Max. 210 watts]	38,4
MSI R9 290X [Max. 195 watts]	37,1
MSI R9 290X [Max. 180 watts]	35,9
MSI R9 290X [Max. 170 watts]	33,8

	FPS

	Far Cry 3
	2560 x 1440 [8xAA / 16xAF]

EVGA GTX 980 [Max. 195 watts]	36,4
EVGA GTX 980 [Max. 210 watts]	36,3
EVGA GTX 980 [Max. 224 watts]	36,2
EVGA GTX 980 [Max. 180 watts]	35,6
EVGA GTX 980 [Max. 165 watts]	34,8
MSI R9 290X [Max. 300 watts]	30,2
MSI R9 290X [Max. 260 watts]	30,1
MSI R9 290X [Max. 225 watts]	28,8
MSI R9 290X [Max. 210 watts]	28,0
MSI R9 290X [Max. 195 watts]	27,1
MSI R9 290X [Max. 180 watts]	26,3
MSI R9 290X [Max. 170 watts]	25,0

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Far Cry 3	190	246

Once again, we see a clear discrepancy in the power consumption between the two representatives under this title and the selected test scene. In the worst case, the EVGA model is satisfied with 190 watts, the MSI graphics card also manages in the 250 watt region.

This means that the results of the GTX 980 in our diagram only scale up to the throttle point of 195 watts - up to that point, however, the performance increases evenly, and so it cannot be deduced here whether a certain TDP range is more useful than that brings other or special advantages.

With the AMD representative, we once again observed that the 170 watt throttle does not taste good at all and the jump in performance to 180 watts with a 6% increase in performance shows the strongest deflection. Nevertheless: The R9 290X also scales in an exemplary manner with the further loosening - up to 23% increase in performance can be seen up to the level of 250 watts.

Individual evaluation: DiRT: Showdown

Game	DiRT: Showdown
Developer	Codemasters southam
Publisher	Codemasters
publication	May 2012
Genre	Racing simulation
Age rating USK	From 7 years
Graphics engine	EGO engine
DirectX path	DirectX 9/10/11
Benchmark measurement	Integrated benchmark
Test area	Miami Routes
Runtime benchmark	85 seconds
Benchmark settings	Highest level of detail, DirectX 11
	Order from Amazon

Benchmark scene in the test

	Dirt: Showdown
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	113,56
EVGA GTX 980 [Max. 210 watts]	113,48
EVGA GTX 980 [Max. 195 watts]	113,37
EVGA GTX 980 [Max. 180 watts]	112,94
EVGA GTX 980 [Max. 165 watts]	111,14
MSI R9 290X [Max. 300 watts]	100,73
MSI R9 290X [Max. 260 watts]	100,49
MSI R9 290X [Max. 225 watts]	99,62
MSI R9 290X [Max. 210 watts]	96,99
MSI R9 290X [Max. 195 watts]	93,87
MSI R9 290X [Max. 180 watts]	91,99
MSI R9 290X [Max. 170 watts]	89,08

	FPS

	Dirt: Showdown
	2560 x 1440 [8xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	105,51
EVGA GTX 980 [Max. 210 watts]	105,49
EVGA GTX 980 [Max. 195 watts]	105,49
EVGA GTX 980 [Max. 180 watts]	104,87
EVGA GTX 980 [Max. 165 watts]	103,02
MSI R9 290X [Max. 300 watts]	95,76
MSI R9 290X [Max. 225 watts]	95,73
MSI R9 290X [Max. 260 watts]	95,72
MSI R9 290X [Max. 210 watts]	93,28
MSI R9 290X [Max. 195 watts]	91,31
MSI R9 290X [Max. 180 watts]	89,22
MSI R9 290X [Max. 170 watts]	86,55

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
DiRT: Showdown	182	216

Our test scene in DiRT: Showdown does not turn out to be particularly demanding for either graphics card in terms of power consumption, and the difference between the two graphics cards is significantly lower than in some of the other applications shown. But there is still a difference of more than 30 watts.

The NVIDIA representative only reacts with a rash in the 165 and 180 watt measurements. After that, practically nothing changes. Again, we only see a performance increase of 2% with the increase, which one could of course do without at such high frame rates.

However, high frame rates can also be seen in the R9 290X in this game, although this is of course secondary to the general approach. As expected, the Radeon graphics card scales relatively linearly in the individual levels up to 225 watts with the individual loosening.

Individual evaluation: Hitman: Absolution

Game	Hitman: Absolution
Developer	IO Interactive / Nixxes software
Publisher	Square Enix
publication	November 2012
Genre	Action shooter
Age rating USK	From 18 years
Graphics engine	Glacier 2
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Rock slope
Runtime benchmark	10 seconds
Benchmark settings	DirectX 11; Highest levels of detail
Post-processing filter	FXAA
Anti-aliasing	4 x / 8 x MSAA
	Find on Amazon*

Picture: VTX3D Radeon HD 7870 Black Edition in the test

In-game test scene

	Hitman: Absolution
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	66,6
EVGA GTX 980 [Max. 210 watts]	66,5
EVGA GTX 980 [Max. 195 watts]	66,5
EVGA GTX 980 [Max. 180 watts]	65,9
MSI R9 290X [Max. 300 watts]	64,6
MSI R9 290X [Max. 260 watts]	64,4
EVGA GTX 980 [Max. 165 watts]	64,3
MSI R9 290X [Max. 225 watts]	62,5
MSI R9 290X [Max. 210 watts]	61,0
MSI R9 290X [Max. 195 watts]	59,1
MSI R9 290X [Max. 180 watts]	57,0
MSI R9 290X [Max. 170 watts]	53,8

	FPS

	Hitman: Absolution
	2560 x 1440 [4xAA / 16xAF]

MSI R9 290X [Max. 300 watts]	49,4
MSI R9 290X [Max. 260 watts]	49,3
MSI R9 290X [Max. 225 watts]	48,3
EVGA GTX 980 [Max. 210 watts]	47,6
MSI R9 290X [Max. 210 watts]	47,6
EVGA GTX 980 [Max. 195 watts]	47,5
EVGA GTX 980 [Max. 224 watts]	47,3
EVGA GTX 980 [Max. 180 watts]	46,9
EVGA GTX 980 [Max. 165 watts]	46,1
MSI R9 290X [Max. 195 watts]	46,1
MSI R9 290X [Max. 180 watts]	45,0
MSI R9 290X [Max. 170 watts]	43,0

	FPS

	Hitman: Absolution
	2560 x 1440 [8xAA / 16xAF]

EVGA GTX 980 [Max. 195 watts]	39,6
EVGA GTX 980 [Max. 224 watts]	39,3
EVGA GTX 980 [Max. 180 watts]	39,0
EVGA GTX 980 [Max. 210 watts]	38,8
EVGA GTX 980 [Max. 165 watts]	38,3
MSI R9 290X [Max. 260 watts]	38,0
MSI R9 290X [Max. 300 watts]	38,0
MSI R9 290X [Max. 225 watts]	37,8
MSI R9 290X [Max. 210 watts]	37,2
MSI R9 290X [Max. 195 watts]	35,8
MSI R9 290X [Max. 180 watts]	34,8
MSI R9 290X [Max. 170 watts]	34,5

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Hitman: Absolution	188	247

Hitman then again presents us with a serious difference in the maximum power consumption of the two candidates. The MSI card is almost 60 watts ahead of the EVGA GTX 980. The latter shows a maximum of almost 190 watts in the test scene.

The relaxation of the power level from 165 to 180 watts again shows us a performance gain of 2%, the throttle to 195 watts then only increases by 1%.

It is interesting to note that the AMD representative does not show any fluctuations in performance beyond the 225 watt throttling. Up to this mark one shows good increases in performance, which then simply fail. However, this statement then only applies to the results with 8 x MSSA, which is probably responsible for the behavior. Without MSSA or with only 4 x MSAA, the Radeon graphics card can increase performance by another 225% from 250 to 3 watts. It can be assumed that at this point it is the graphics card memory that leads to the limitations shown.

Individual evaluation: Max Payne 3

Game	Max Payne 3
Developer	Rockstar studios
Publisher	Rockstar Games
publication	May 2012
Genre	Action shooter
Age rating USK	From 18 years
Graphics engine	RAGE (Rockstar Advanced Game Engine)
DirectX path	DirectX 9/10/11
Benchmark measurement	Fraps / savegame
Test area	Chapter V - Control Point 10
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail, FXAA high
	Order from Amazon*

In-game test scene

	Max Payne 3
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	68,6
EVGA GTX 980 [Max. 210 watts]	68,5
EVGA GTX 980 [Max. 195 watts]	68,5
EVGA GTX 980 [Max. 180 watts]	68,3
EVGA GTX 980 [Max. 165 watts]	67,6
MSI R9 290X [Max. 300 watts]	61,2
MSI R9 290X [Max. 260 watts]	61,1
MSI R9 290X [Max. 225 watts]	59,0
MSI R9 290X [Max. 210 watts]	57,2
MSI R9 290X [Max. 195 watts]	55,4
MSI R9 290X [Max. 180 watts]	53,8
MSI R9 290X [Max. 170 watts]	50,9

	FPS

	Max Payne 3
	2560 x 1440 [8xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	42,9
EVGA GTX 980 [Max. 210 watts]	42,9
EVGA GTX 980 [Max. 195 watts]	42,9
EVGA GTX 980 [Max. 180 watts]	42,8
EVGA GTX 980 [Max. 165 watts]	42,1
MSI R9 290X [Max. 300 watts]	32,6
MSI R9 290X [Max. 260 watts]	32,4
MSI R9 290X [Max. 225 watts]	31,4
MSI R9 290X [Max. 210 watts]	30,7
MSI R9 290X [Max. 195 watts]	29,8
MSI R9 290X [Max. 180 watts]	29,1
MSI R9 290X [Max. 170 watts]	28,0

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Max Payne 3	185	238

Max Payne 3 demands the MSI graphics card with Radeon GPU once again much more than the NVIDIA representative in the form of the EVGA GTX 980 SC.

Due to the low power consumption of the NVIDIA representative, we no longer see any actual fluctuations after the 180 watt level. From 165 to 180 watts it is the often seen 2%.

The MSI R9 290X Gaming 4G scales relatively evenly up to 250 watts - again, the easing from 170 to 180 watts is most noticeable.

Individual evaluation: Metro: Last Light

Game	Metro: Last Light
Developer	4A Games
Publisher	Deep Silver
publication	17 May 2013
Genre	Ego shooter
Graphics engine	4A engine
DirectX path	DirectX 10/11
Age rating USK	From 18 years
Benchmark measurement	Fraps / savegame
Test area	Chapter infection
Runtime benchmark	10 seconds
Benchmark settings	System settings: Very high - Tess: High
	Find it on Amazon*

In-game test scene

	Metro: Last Light
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	74,61
EVGA GTX 980 [Max. 210 watts]	74,38
EVGA GTX 980 [Max. 195 watts]	73,29
EVGA GTX 980 [Max. 180 watts]	72,17
EVGA GTX 980 [Max. 165 watts]	69,94
MSI R9 290X [Max. 300 watts]	56,80
MSI R9 290X [Max. 260 watts]	56,33
MSI R9 290X [Max. 225 watts]	54,12
MSI R9 290X [Max. 210 watts]	53,54
MSI R9 290X [Max. 195 watts]	50,65
MSI R9 290X [Max. 180 watts]	49,23
MSI R9 290X [Max. 170 watts]	47,31

	FPS

	Metro: Last Light
	2560 x 1440 [4xSSAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	39,72
EVGA GTX 980 [Max. 210 watts]	39,53
EVGA GTX 980 [Max. 195 watts]	39,07
EVGA GTX 980 [Max. 180 watts]	38,50
EVGA GTX 980 [Max. 165 watts]	37,69
MSI R9 290X [Max. 300 watts]	30,82
MSI R9 290X [Max. 260 watts]	30,81
MSI R9 290X [Max. 225 watts]	30,11
MSI R9 290X [Max. 210 watts]	29,12
MSI R9 290X [Max. 195 watts]	28,07
MSI R9 290X [Max. 180 watts]	26,77
MSI R9 290X [Max. 170 watts]	25,67

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Metro: Last Light	210	252

Metro: Last Light challenges the NVIDIA representative a little more than some of the previous titles. After all, 210 watts have to be made here in order to achieve maximum performance. AMD brings it with the MSI representative to just over 250 watts.

Our EVGA test card behaves as usual. You scale in two percent steps per loosening. From 195 watts to 210 watts, we then only see an increase of 1%. In terms of performance, the 195 watt mark seems to offer the best performance here.

And the AMD representative? It also has no surprises in store and scales quite linearly in the individual levels, up to 250 watts. From 225 watts, the gain in performance is a little lower, which is then only due to the choice of 4-fold supersampling, as the measured values without this option show.

Individual evaluation: Splinter Cell: Blacklist

Game	Splinter Cell: Blacklist
Developer	Ubisoft Toronto, Montreal, Shanghai
Publisher	Ubisoft
publication	August 2013
Genre	Sneak game, action adventure
Age rating USK	From 18 years
Graphics engine	Unreal Engine 2.5 / LEAD / Havok Physics
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Level: Benghazi - Libya - Mission 1
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail, DX 11
HT4U test	Order from Amazon*

In-game test scene

	Splinter Cell: Blacklist
	2560 x 1440 [4xAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	51,44
EVGA GTX 980 [Max. 210 watts]	51,36
EVGA GTX 980 [Max. 195 watts]	51,36
EVGA GTX 980 [Max. 180 watts]	51,28
EVGA GTX 980 [Max. 165 watts]	50,42
MSI R9 290X [Max. 260 watts]	42,58
MSI R9 290X [Max. 300 watts]	42,52
MSI R9 290X [Max. 225 watts]	42,16
MSI R9 290X [Max. 210 watts]	41,61
MSI R9 290X [Max. 195 watts]	40,46
MSI R9 290X [Max. 180 watts]	39,88
MSI R9 290X [Max. 170 watts]	38,65

	FPS

	Splinter Cell: Blacklist
	2560 x 1440 [4xSSAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	29,61
EVGA GTX 980 [Max. 210 watts]	29,36
EVGA GTX 980 [Max. 195 watts]	29,16
EVGA GTX 980 [Max. 180 watts]	29,12
EVGA GTX 980 [Max. 165 watts]	28,14
MSI R9 290X [Max. 300 watts]	25,85
MSI R9 290X [Max. 260 watts]	25,78
MSI R9 290X [Max. 225 watts]	24,61
MSI R9 290X [Max. 210 watts]	23,89
MSI R9 290X [Max. 195 watts]	22,74
MSI R9 290X [Max. 180 watts]	21,94
MSI R9 290X [Max. 170 watts]	20,68

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
SC: Blacklist	170	251

We see the biggest discrepancy so far in the difference in maximum power consumption in the benchmark scene from Splinter Cell. The two test candidates today are separated by over 70 watts.

The consequence of this is, of course, that the NVIDIA representative only shows swings when easing from 165 watts to 180 watts - the 2%, which has now been seen often, is shown again.

The commentary on the AMD representative remains largely the same. Linear up to 225 watts. The relaxation to 250 watts then no longer shows a large increase in performance. A circumstance that could be observed more often.

Individual evaluation: Thief (2014)

Game	Thieves (2014)
Developer	Eidos
Publisher	Square Enix
publication	February 2014
Genre	Action adventure / stealth game
Age rating USK	From 16 years
Graphics engine	Unreal engine 3
DirectX path	DirectX 9/11
Benchmark measurement	Fraps / savegame
Test area	Stone Market
Runtime benchmark	10 seconds
Benchmark settings	Highest levels of detail
	HT4U test
	Find it on Amazon*

In-game test scene

	Thief
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	82,93
EVGA GTX 980 [Max. 210 watts]	82,42
EVGA GTX 980 [Max. 195 watts]	82,28
EVGA GTX 980 [Max. 180 watts]	81,60
EVGA GTX 980 [Max. 165 watts]	81,40
MSI R9 290X [Max. 300 watts]	59,56
MSI R9 290X [Max. 260 watts]	59,48
MSI R9 290X [Max. 225 watts]	58,53
MSI R9 290X [Max. 210 watts]	56,61
MSI R9 290X [Max. 195 watts]	55,46
MSI R9 290X [Max. 180 watts]	53,72
MSI R9 290X [Max. 170 watts]	52,11

	FPS

	Thief
	2560 x 1440 [2xSSAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	57,88
EVGA GTX 980 [Max. 210 watts]	57,63
EVGA GTX 980 [Max. 195 watts]	57,45
EVGA GTX 980 [Max. 180 watts]	57,29
EVGA GTX 980 [Max. 165 watts]	56,35
MSI R9 290X [Max. 300 watts]	43,25
MSI R9 290X [Max. 260 watts]	43,05
MSI R9 290X [Max. 225 watts]	42,12
MSI R9 290X [Max. 210 watts]	41,16
MSI R9 290X [Max. 195 watts]	39,78
MSI R9 290X [Max. 180 watts]	38,84
MSI R9 290X [Max. 170 watts]	37,16

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
Thieves (2014)	180	226

The differences in the maximum power consumption of both representatives are not as extreme as in the blacklist, but of course still clearly present at around 46 watts. The GTX 980 hardly needs more than 180 watts in this scene, and it is therefore already clear that the following scaling will hardly bring any new information with it.

And so we see an increase in performance by the previously typical 2% when loosening from 165 to 180 watts and then no more reactions.

The AMD representative is once again most impressed by the relaxation from 170 to 180 watts, but also scales up to 250 watts.

Individual evaluation: Tomb Raider (2013)

Game	tomb raider
Developer	Crystal Dynamics & Eidos Montreal
Publisher	Square Enix
publication	March 5, 2013
Genre	Action-Adventure
Graphics engine	Crystal engine
DirectX path	DirectX 9/11
Age rating USK	From 18 years
Benchmark measurement	Fraps / savegame
Test area	at about 26 percent of the game
Runtime benchmark	10 seconds
Benchmark settings	Highest level of detail, TressFX, shadow: Ultra
HT4U test	Order from Amazon*

In-game test scene

	Tomb Raider (2013)
	2560 x 1440 [No AA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	65,63
EVGA GTX 980 [Max. 210 watts]	65,54
EVGA GTX 980 [Max. 195 watts]	65,47
EVGA GTX 980 [Max. 180 watts]	64,36
EVGA GTX 980 [Max. 165 watts]	63,20
MSI R9 290X [Max. 300 watts]	53,59
MSI R9 290X [Max. 260 watts]	52,95
MSI R9 290X [Max. 225 watts]	52,86
MSI R9 290X [Max. 210 watts]	51,20
MSI R9 290X [Max. 195 watts]	48,59
MSI R9 290X [Max. 180 watts]	48,40
MSI R9 290X [Max. 170 watts]	45,67

	FPS

	Tomb Raider (2013)
	2560 x 1440 [2xSSAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	43,52
EVGA GTX 980 [Max. 210 watts]	43,50
EVGA GTX 980 [Max. 195 watts]	43,17
EVGA GTX 980 [Max. 180 watts]	41,69
EVGA GTX 980 [Max. 165 watts]	40,91
MSI R9 290X [Max. 260 watts]	35,57
MSI R9 290X [Max. 300 watts]	35,07
MSI R9 290X [Max. 225 watts]	34,09
MSI R9 290X [Max. 210 watts]	32,91
MSI R9 290X [Max. 195 watts]	31,68
MSI R9 290X [Max. 180 watts]	30,15
MSI R9 290X [Max. 170 watts]	29,17

	FPS

	Tomb Raider (2013)
	2560 x 1440 [4xSSAA / 16xAF]

EVGA GTX 980 [Max. 224 watts]	29,98
EVGA GTX 980 [Max. 210 watts]	29,79
EVGA GTX 980 [Max. 195 watts]	29,05
EVGA GTX 980 [Max. 180 watts]	29,00
EVGA GTX 980 [Max. 165 watts]	28,36
MSI R9 290X [Max. 300 watts]	25,10
MSI R9 290X [Max. 260 watts]	24,89
MSI R9 290X [Max. 225 watts]	23,77
MSI R9 290X [Max. 210 watts]	22,75
MSI R9 290X [Max. 195 watts]	21,87
MSI R9 290X [Max. 180 watts]	21,59
MSI R9 290X [Max. 170 watts]	20,33

	FPS

Game	EVGA GTX 980 SC ACX 2.0 (watt)	MSI R9 290X Gaming (watts)
tomb raider	204	251

In Tomb Raider, too, the two test subjects are separated by more than 40 watts at maximum power consumption. The NVIDIA card shows swings up to a maximum of 204 watts, the Radeon candidate wanders just over 251 watts. Again, the findings from the last diagram are not new.

EVGA's GTX 980 SC ACX 2.0 scales up to 210 watts in the usual range and shows the usual 2% performance plus between 165 and 180 watts. At the next level of 195 watts you increase - quite unusually so far - as much as 4%, and in the last step to 210 watts with only 1%. Tomb Raider as a game shows - as one of the few titles - that the performance of the GTX 980 in the 195 watt range would be best chosen.

AMD's R9 290X is Tomb Raider like almost every title. The GPU scales up to the level of 260 watts and can increase the 22D performance by up to a maximum of 3%. We cannot identify a special “favorite area” here.

Summary

After today's measurements and analyzes, we have to clearly state and attest that the two graphics chip manufacturers AMD and NVIDIA probably know their GPUs best and that the selected default settings for power consumption are no coincidence. You have certainly thought about this based on many of your own tests and measurements and AMD has probably made the right choice with 250 watts - NVIDIA with 180 watts.

However, our tests have also shown that across the board we cannot say that these are the sweet spots for both chips from the manufacturers, because in some cases a sweet spot with AMD might be more like 225 watts, with NVIDIA more than 195 watts. But AMD has probably recognized that in a large number of cases you can offer a good increase in performance up to this range, and NVIDIA may have realized that the scaling is no longer quite as optimal. You could run through 100 other applications as you like and would always find extreme outliers in the result.

With this, however, one can come to the conclusion that an averaged value over a large number of applications is the best indicator when one needs a general decision value. If you want to tie things like power consumption and 3D performance to your personal needs, you have to go much deeper into research and the question of your own needs.

Generally speaking - based on our measurements - the implementation of the EVGA GTX 980 SC ACX 2.0 shown today is around 17% faster than the MSI R9 290X Gaming 4G shown and also offers around 17% better power consumption. Subjectively, however, the EVGA graphics card is not representative of all GTX 980 models on the market, just as the MSI card is not for Radeon R9-290X cards. There are applications and resolutions in which an R9 290X can sometimes be equal to a GTX 980 - but in most cases not really.

Generally speaking, today's test also clearly indicates that AMD is under pressure with future chips. Increased performance is required, but also a well-balanced mix of power consumption, because NVIDIA has presented here with the GTX 980 and Maxwell technology. AMD's R9 380X is expected this year - but at the moment rumors are again pointing to a 250 watt solution. The expected R9 390X represents the top chip from AMD. NVIDIA's top model - the GM200 - has not yet been presented or announced. However, after the presentations so far, it should be obvious that you will not go above a maximum board power of 225 watts - the manufacturer's throttle options will probably intervene much earlier. But our crystal ball is currently too cloudy for that, and we would rather wait for the facts.

[pg], February 13, 2015

About David Maul

David Maul is a qualified business IT specialist with a passion for hardware

All contributions by David Maul →