Introduction

http://www.techpowerup.com/reviews/N...mages/logo.jpg

How time flies. The last time we had a bleeding-edge graphics card from NVIDIA to try by fire, was exactly a year ago, with the launch of GeForce GTX 590. Quite a bit changed since then. Rival AMD is just about done with the launch of its Radeon HD 7000 series "Southern Islands" GPU family, and NVIDIA's answer to that felt wanted, for the past 3 months or so. Well, it’s finally here – our GeForce GTX 680 Kepler review.

Both evolutionary and revolutionary changes have gone into making the GeForce GTX 680. It's evolutionary in that it's designed to be an upgrade over its immediate predecessor (and not something two generations behind), and revolutionary new features make sure GTX 680 is a worthy upgrade. The evolutionary part of course is that NVIDIA wants to reclaim the title of having the fastest GPU out there; and the revolutionary part is a vibrant feature-set that supposedly contributes to never before seen energy-efficiency levels, to accomplish those steep design goals.

http://www.techpowerup.com/reviews/N...ide1_small.jpg

Product Positioning

Speaking of positioning, NVIDIA is gunning for nothing short of the performance-crown for single-GPU graphics cards, with the GeForce GTX 680, and it wants to do so without having to compromise of energy-efficiency. In the past, we have seen both NVIDIA and AMD throw energy efficiency to the wind and come up with power-guzzling chips, in a blind pursuit of performance leadership. That's not the case with GeForce GTX 680.

http://www.techpowerup.com/reviews/N...ide2_small.jpg

See what we mean? To be fair to the HD 7970, it did impress us with its performance/Watt figures. What NVIDIA is looking to do is raise the bar with energy-efficiency. To accomplish that with the Kepler-based GTX 680, and go on to seek performance-leadership is a very tough ask, and takes some very gritty engineering.

One revolutionary change that allows GeForce GTX 680 to aim high, is an extremely smart self-tuning logic that fine-tunes clock speeds and voltages, on the fly, with zero user intervention, to yield the best possible combination of performance and efficiency for a given load scenario. The GTX 680 hence reshapes the definition of fixed load clock speed, with dynamic clock speeds. Think of it as a GPU-take on Intel's Turbo Boost technology, which works in conjunction with SpeedStep to produce the best performance-per-Watt for CPUs that feature it.

<table class="tputbl hilight" cellspacing="0" cellpadding="3">
<caption>
GeForce GTX 680 Market Segment Analysis
</caption>
<tr>
<th scope="col"> </th>
<th scope="col">Radeon <br />
HD 7870</th>
<th scope="col">GeForce <br />
GTX 580</th>
<th scope="col">Radeon <br />
HD 7950</th>
<th scope="col">Radeon <br />
HD 7970</th>
<th scope="col"><strong>GeForce <br />
GTX 680</strong></th>
<th scope="col">Radeon <br />
HD 6990</th>
<th scope="col">GeForce <br />
GTX 590</th>
</tr>
<tr>
<th scope="row">Shader Units</th>
<td align="right">1280</td>
<td align="right">512</td>
<td align="right">1792</td>
<td align="right">2048</td>
<td align="right"><strong>1536</strong></td>
<td align="right">2x 1536</td>
<td align="right">2x 512</td>
</tr>
<tr class="alt">
<th scope="row">ROPs</th>
<td align="right">32</td>
<td align="right">48</td>
<td align="right">32</td>
<td align="right">32</td>
<td align="right"><strong>32</strong></td>
<td align="right">2x 32</td>
<td align="right">2x 48</td>
</tr>
<tr>
<th scope="row">Graphics Processor</th>
<td align="right">Pitcairn</td>
<td align="right">GF110</td>
<td align="right">Tahiti</td>
<td align="right">Tahiti</td>
<td align="right"><strong>GK104</strong></td>
<td align="right">2x Cayman</td>
<td align="right">2x GF110</td>
</tr>
<tr class="alt">
<th scope="row">Transistors</th>
<td align="right">2800M</td>
<td align="right">3000M</td>
<td align="right">4310M</td>
<td align="right">4310M</td>
<td align="right"><strong>3540M</strong></td>
<td align="right">2x 2640M</td>
<td align="right">2x 3000M</td>
</tr>
<tr>
<th scope="row">Memory Size</th>
<td align="right">2048 MB</td>
<td align="right">1536 MB</td>
<td align="right">3072 MB</td>
<td align="right">3072 MB</td>
<td align="right"><strong>2048 MB</strong></td>
<td align="right">2x 2048 MB</td>
<td align="right">2x 1536 MB</td>
</tr>
<tr class="alt">
<th scope="row">Memory Bus Width</th>
<td align="right">256 bit</td>
<td align="right">384 bit</td>
<td align="right">384 bit</td>
<td align="right">384 bit</td>
<td align="right"><strong>256 bit</strong></td>
<td align="right">2x 256 bit</td>
<td align="right">2x 384 bit</td>
</tr>
<tr>
<th scope="row">Core Clock</th>
<td align="right">1000 MHz</td>
<td align="right">772 MHz</td>
<td align="right">800 MHz</td>
<td align="right">925 MHz</td>
<td align="right"><strong>1006 MHz+</strong></td>
<td align="right">830 MHz</td>
<td align="right">607 MHz</td>
</tr>
<tr class="alt">
<th scope="row">Memory Clock</th>
<td align="right">1200 MHz</td>
<td align="right">1002 MHz</td>
<td align="right">1250 MHz</td>
<td align="right">1375 MHz</td>
<td align="right"><strong>1502 MHz</strong></td>
<td align="right">1250 MHz</td>
<td align="right">855 MHz</td>
</tr>
<tr>
<th scope="row">Price</th>
<td align="right">$360</td>
<td align="right">$390</td>
<td align="right">$450</td>
<td align="right">$550</td>
<td align="right"><strong>$499</strong></td>
<td align="right">$700</td>
<td align="right">$750</td>
</tr>
</table>

Architecture

http://www.techpowerup.com/reviews/N...ide3_small.jpg http://www.techpowerup.com/reviews/N...shot_small.jpg
At the heart of the GeForce GTX 680, is the GeForce Kepler architecture. Its design goals are to raise performance and energy-efficiency over the previous generation "Fermi" architecture. GeForce Kepler's architecture more or less maintains the basic component-hierarchy of GeForce Fermi, which emphasizes on a fast, highly parallelized component load-out. Think of the hierarchy as a Bento container. At the topmost level is the PCI-Express Gen. 3.0 host interface, a 256-bit wide GDDR5 memory interface, and a highly-tweaked NVIDIA GigaThread Engine, which transacts processed and unprocessed data between the host and memory interfaces.

http://www.techpowerup.com/reviews/N...ture_small.jpg http://www.techpowerup.com/reviews/N.../gpc_small.jpg http://www.techpowerup.com/reviews/N.../smx_small.jpg
At the downstream of the GigaThread Engine are four Graphics Processing Clusters (GPCs). Each GPC is a self-contained GPU subunit, since it has nearly every component an independent GPU does. The GPC has one shared resource, and two dedicated resources, the shared resource is the Raster Engine, which handles high-level raster operations such as edge setup, and Z-cull. The dedicated resources are the next-generation Streaming Multiprocessor-X (SMX). A large chunk of architectural improvements have gone into perfecting this component. The Streaming Multiprocessor-X (SMX) is the GPU's number-crunching resource. It is highly-parallelized, to handing the kind of computing loads that tomorrow's 3D applications demand.

http://www.techpowerup.com/reviews/N...ster_small.jpg http://www.techpowerup.com/reviews/N...uler_small.jpg
The SMX further has shared and dedicated resources. The next-generation PolyMorph 2.0 Engine handles the low-level raster operations, such as vertex-fetch, tessellation, viewpoint-transformation, attribute setup, etc. The dedicated components are where the number-crunching action happens. Four Warp Schedulers marshal instructions and data between 192 CUDA cores. This is a six-fold increase over that of the GF110, and four-fold over that of the GF114. There are 16 texture memory units per SMX, which are cached. The Warp Schedulers are backed by a more efficient software pre-decode algorithm that reduces the number of steps needed to issue instructions. Essentially at shader compile time the shader compiler, which is a component of NVIDIA's driver, will evaluate the instruction stream, reorder instructions as needed and provide extra info to hardware by attaching additional info to instructions. It can do so because it has a complete view of the shader code.

http://www.techpowerup.com/reviews/N...less_small.jpg http://www.techpowerup.com/reviews/N...sseg_small.jpg
NVIDIA also innovated what it calls Bindless Textures. In the classical GPU model, to reference a texture, the GPU has to allocate it a slot in a fixed size binding table, which limits the number of textures a shader can access at a given time, which ended up being 128, with the previous-generation Fermi architecture. The Kepler architecture removes this binding step, a shader can reference textures directly in the memory, without needing a conventional binding table. So the number of textures a shader can reference is practically unlimited, or 1 million if you want to talk in figures. This makes rendering scenes that are as complex as the photo above, a breeze, because it can be done so with fewer passes.

To sum it all up, the GeForce Kepler 104 GPU has 192 CUDA cores per SMX, 384 per GPC, and 1536 in all. It has 128 Texture Memory Units (TMUs) in all (16 per SMX, 32 per GPC); and 32 Raster Operations Processors (ROPs) in all. At several levels, transactions between the various components are cached, to prevent wastage of clock cycles (in turn, translating to energy efficiency).

http://www.techpowerup.com/reviews/N...xaa2_small.jpg http://www.techpowerup.com/reviews/N...xaa3_small.jpg
NVIDIA also introduced a new anti-aliasing (AA) algorithm called TXAA. There have already been a few new AA algorithms introduced in recent past, such as FXAA, SMAA and SRAA, which raised the bar with quality, with lower performance-impact. TXAA seeks to raise it even further, with image quality comparable to high levels MSAA, with the performance penalty of 4x MSAA. TXAA is a hybrid between hardware multi-sampling, temporal AA, and a customized AA resolve. It has two levels: TXAA1 and TXAA2. The former offers image quality comparable to 8x MSAA, with the performance-penalty of 2x MSAA; the latter offers image quality beyond 8x MSAA, with the performance-penalty comparable to 4x MSAA. Since lower MSAA levels are practically "free" with today's GPUs, TXAA will wipe the floor with the competition, in terms of image quality, but there's a catch. Applications have to be customized to take advantage of TXAA. This is where NVIDIA's developer-relations muscle should kick in. We expect a fairly decent proliferation of TXAA among upcoming games.

NVIDIA has also added an FXAA option to the driver control panel, which enables it in all games without the need for any integration from game developers.

http://www.techpowerup.com/reviews/N...sync_small.jpg http://www.techpowerup.com/reviews/N...ync2_small.jpg
The last of the three big features is Adaptive V-Sync. The feature improves on traditional V-Sync, by dynamically adjusting the frame limiter to ensure smoother gameplay. Traditional V-Sync merely sends frame data to the screen after every full screen refresh. This means if a frame arrives slow, because the GPU took longer to render it, it will have to wait a full screen refresh before it can be displayed, effectively reducing frame rate to 30 FPS. If rendering a frame takes longer than two full refreshes, the frame rate will even drop down to 20 FPS. These framerate differences are very noticeable during gaming because they are so huge.
What Adaptive V-Sync does is, it makes the transition between frame-rate drop and synchronized frame-rate smooth, alleviating lag. It achieves this by dynamically adjusting the value that V-Sync takes into account when limiting frame-rates. I did some testing of this feature and found it to work as advertised. Of course this does not completely eliminate frame rate differences, but it makes them less noticeable.

Packaging

http://www.techpowerup.com/reviews/N...age1_small.jpg http://www.techpowerup.com/reviews/N...age2_small.jpg

NVIDIA's sample package does not include any accessories, I guess reviewers already have enough of those. Retail cards will come with the usual documentation, driver CD and adapters.

The Card

http://www.techpowerup.com/reviews/N...ard1_small.jpg http://www.techpowerup.com/reviews/N...ard2_small.jpg
NVIDIA's reference design cooler, follows the company's black and green cooler theme. The board looks immediately familiar, following the design of great products such as GTX 580.

http://www.techpowerup.com/reviews/N...ard3_small.jpg
The card requires two slots in your system.

http://www.techpowerup.com/reviews/N...puts_small.jpg
Display connectivity options include one dual-link DVI port (with analog VGA), one dual-link DVI port (digital only), one full size HDMI port and one full size DisplayPort. You may use all the outputs at the same time, thanks to NVIDIA's new display output logic that is introduced with Kepler. This also makes triple screen multi-monitor gaming easy now, since you need only one card.

An HDMI sound device is included in the GPU, too. It is HDMI 1.4a compatible which includes HD audio, support for 4K and Blu-ray 3D movies. The DisplayPort outputs are version 1.2 which enables the use of hubs and Multi-Stream transport.

http://www.techpowerup.com/reviews/N.../sli_small.jpg
You may combine up to four GeForce GTX 680 cards from any vendor in a multi-GPU SLI configuration for higher framerates or better image quality settings.

http://www.techpowerup.com/reviews/N...ront_small.jpg http://www.techpowerup.com/reviews/N...back_small.jpg
Pictured above are photos of the front and back, showing the disassembled board. High-res versions are also available (front, back). If you choose to use these images for voltmods etc, please include a link back to this site or let us post your article.

A Closer Look

http://www.techpowerup.com/reviews/N...ler1_small.jpg http://www.techpowerup.com/reviews/N...ler2_small.jpg
NVIDIA's heatsink may look simple on the outside, but as the second picture shows it is an elaborate construction that cools GPU, memory and voltage regulation circuitry at the same time.
One thing that I noticed is that the cooler emits a strong smell of solvents when the card is loaded. This is not the typical "new graphics card" smell, but something more like glue. Even after a week of testing, it has not completely gone away, but gotten much less intense.

http://www.techpowerup.com/reviews/N...ower_small.jpg
The card requires two 6-pin PCI-Express power connectors. This configuration is good for up to 225 W of power draw. NVIDIA claims that this stacked power connector configuration is better, but I personally don't like it. It makes plugging cables in and out more difficult and the bottom connector is rotated by 180° from what we are used to.

http://www.techpowerup.com/reviews/N...a219_small.jpg
Three of these INA219 power monitor chips are located on the board (the third one is on the other side). These provide voltage, current and power monitoring for 12V PCI-E power, and both 6-pin power connectors. They are used to provide realtime power consumption numbers to the driver, which will then use that info to enable dynamic overclocking and ensure the board does not go above its rated TDP.

http://www.techpowerup.com/reviews/N...vreg_small.jpg
NVIDIA has chosen to use a Richtek RT8802A voltage controller on their card. This is a fairly simple controller which does not offer any monitoring features or software voltage control. Voltages are controlled via VID pins that are directly connected to the GPU.
Also note how the chip sits on its own little PCB, which could hint that this is a modular design that will accept different, more advanced controllers, too.

http://www.techpowerup.com/reviews/N...mory_small.jpg
The GDDR5 memory chips are made by Hynix, and carry the model number H5GQ2H24MFR-R0C. They are specified to run at 1500 MHz (6000 MHz GDDR5 effective).

http://www.techpowerup.com/reviews/N.../gpu_small.jpg
NVIDIA's new GK104 graphics processor introduces the company's brand-new Kepler architecture. It is NVIDIA's first chip to be produced on a 28 nanometer process, at TSMC Taiwan. The transistor count is 3.54 billion.

Test System

<table class="tputbl">
<thead>
<tr>
<th colspan="2">Test System - VGA Rev. 16</th>
</tr>
</thead>
<tr>
<th width="120" scope="row">CPU:</th>
<td align="center">Intel Core i7 920 @ 3.8 GHz<br />(Bloomfield, 8192 KB Cache)</td>
</tr>
<tr class="alt">
<th scope="row">Motherboard:</th>
<td align="center">Gigabyte X58 Extreme<br />
Intel X58 & ICH10R</td>
</tr>
<tr>
<th scope="row">Memory:</th>
<td align="center">3x 2048 MB Mushkin Redline XP3-12800 DDR3 <br>
@ 1520 MHz 8-7-7-16</td>
</tr>
<tr class="alt">
<th scope="row">Harddisk:</th>
<td align="center">WD Caviar Blue WD5000AAKS 500 GB</td>
</tr>
<tr>
<th scope="row">Power Supply:</th>
<td align="center">Antec HCP-1200 1200W</td>
</tr>
<tr class="alt">
<th scope="row">Software:</th>
<td align="center">Windows 7 64-bit Service Pack 1</td>
</tr>
<tr>
<th scope="row">Drivers:</th>
<td valign="top" align="center">NVIDIA: 285.62<br />ATI: Catalyst 11.12<br />HD 7950 & 7970: 8.921.2 RC11<br />HD 7750 & HD 7770: 8.932.2<br />GTX 680: 300.99</td>
</tr>
<tr class="alt">
<th scope="row">Display:</th>
<td valign="top" align="center">
LG Flatron W3000H 30" 2560x1600<br /><img src="http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_680/images/zotac.jpg" width="120" height="40"></td>
</tr>
</table>
Benchmark scores in other reviews are only comparable when this exact same configuration is used.

All video card results were obtained on this exact system with the exact same configuration.
All games were set to their highest quality setting unless indicated otherwise.
AA and AF are applied via in-game settings, not via the driver's control panel.

Each benchmark was tested at the following settings and resolution:

1024 x 768, No Anti-aliasing. This is a standard resolution without demanding display settings.
1280 x 1024, 2x Anti-aliasing. Common resolution for most smaller flatscreens today (17" - 19"). A bit of eye candy turned on in the drivers.
1680 x 1050, 4x Anti-aliasing. Most common widescreen resolution on larger displays (19" - 22"). Very good looking driver graphics settings.
1920 x 1200, 4x Anti-aliasing. Typical widescreen resolution for large displays (22" - 26"). Very good looking driver graphics settings.
2560 x 1600, 4x Anti-aliasing. Highest possible resolution for commonly available displays (30"). Very good looking driver graphics settings.

Aliens vs. Predator

http://www.techpowerup.com/reviews/N...images/avp.jpg
Aliens vs. Predator is based on a merger of the Aliens and the Predators franchise: two legendary alien species that are in conflict with each other, fighting to the death with human marines caught in between. The first person shooter game was developed by Rebellion Studios, who also developed the first AVP PC title and released in February 2010. It is one of the first DirectX 11 games with support for new features like tesselation, which is why AMD heavily promoted it at the time of their DX 11 card launches. We use the AVP benchmark utility with tesselation and advanced DX11 shadows enabled.

http://www.techpowerup.com/reviews/N...p_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Batman: Arkham City

http://www.techpowerup.com/reviews/N...arkhamcity.jpg
Batman is back on the LCD screen with Arkham City, a sequel to Batman: Arkham Asylum, by Rocksteady Games and WB. It was released to the PC platform in November. Batman is imprisoned in Arkham City, an infamous district of the DC Universe that contains the scum of Gotham, most of which Batman helped get in there. In order to get out he must go through scores of baddies, and encounter many of the iconic super-villains along the way. He's not entirely alone.
Batman Arkham City uses the same Unreal Engine by Epic, as Arkham Asylum, but thanks to the engine's modularity, it has been overhauled, outfitted with the latest technologies, including a graphics engine that takes advantage of DirectX 11.

http://www.techpowerup.com/reviews/N...y_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Battlefield 3

http://www.techpowerup.com/reviews/N...images/bf3.jpg
Arguably the most anticipated online shooter title among real gamers - PC gamers, Battlefield 3 is the latest addition to some of the most engaging online multi-player shooter franchises. It combines infantry combat with mechanized warfare including transport vehicles, armored personnel carriers, main battle tanks, attack helicopters, combat aircraft, pretty much everything that goes into today's battlefields. The infantry combat is coupled with role-playing elements, which makes the experience all the more engaging. It also has a single-player campaign which added a few gigabytes to its installer.
Behind all this is a spanking new game engine by EA-DICE, Frostbite 2. It makes use of every possible feature DirectX 11 has to offer, including hardware tessellation, and new lighting effects, to deliver some of the most captivating visuals gamers ever had access to. Not playing this game on PC is grave injustice to what's in store. Faster PCs are rewarded with better visuals.

http://www.techpowerup.com/reviews/N...3_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

BattleForge

http://www.techpowerup.com/reviews/N...attleforge.jpg
BattleForge, a card based RTS, is developed by the German EA Phenomic Studio. A few months after launch the game was transformed into a Play 4 Free branded game. That move and the fact that it was included as game bundle with a large number of ATI cards made it one of the more well known RTS games of 2009. You as a player assemble your deck before game to select the units that will be available. Elemental force choices can be from forces of Fire, Frost, Nature and Shadow to complement each other.
The BattleForge engine has full support for DX 9, DX 10 and DX 10.1, we use the internal benchmark tool in DirectX 11 mode with highest settings to acquire our results.

http://www.techpowerup.com/reviews/N...e_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Call of Duty 4

http://www.techpowerup.com/reviews/N...mages/cod4.jpg
Call of Duty 4 is a first-person shooter that is built on the award winning Call of Duty Series. It is the first version to play in modern times. In a near-future conflict between the United States, Europe and Russia you get to play as a United States Marine and a British SAS operative. The engine is Infinity Ward's own creation and has true dynamic lighting, depth of field, dynamic shadows and HDR. Even though the game plot is scripted you will find yourself in intense battles, often working together with computer controlled team mates. Later installments of the Call of Duty Series use the same game engine, so this test is also representative of Call of Duty: Modern Warfare 3 performance.

http://www.techpowerup.com/reviews/N...4_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Civilization 5

http://www.techpowerup.com/reviews/N...mages/civ5.jpg
Sid Meier's Civilization V (or Civ 5 in common jargon), is the latest addition to the franchise of masterfully-crafted realtime strategy games that let you play God to a nascent civilization of your choice all the way up to the space-age. Civilization V uses large 3D worlds that are procedurally-generated, and takes advantage of hardware tessellation features offered by DirectX 11 to exponentially step up complexity of cities, models, terrains, and objects. It is also expected of this generation of GPUs to handle the larger texture loads that come with the eye-candy.

http://www.techpowerup.com/reviews/N...5_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Crysis

http://www.techpowerup.com/reviews/N...ges/crysis.jpg
After the tremendous success of Far Cry, the German game studio Crytek released their latest shooter Crysis in 2007. The game was by far the most hyped and anticipated game in 2007, and forums were full of "Can my system run Crysis?" threads because of the high hardware requirements of this game. Just like in Far Cry the plot evolves on a small island with a thick and richly detailed jungle world. A lot of attention has been given to small details like accurate physics. For example when you fire on a tree trunk, it will shatter and the tree will fall over leaving a stump behind. Enemies in a car can be stopped by shooting the tire of the car. The game graphics are top notch, even today, yet the game still runs well on most computers.

http://www.techpowerup.com/reviews/N...s_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Crysis 2

http://www.techpowerup.com/reviews/N...es/crysis2.jpg
Crysis 2 takes the player into an alien-infested New York City. The game adds a tactical options mode that allows several approaches to attack a heavily infested enemy location. The new Nanosuit 2.0 that the player uses offers more freedom in ability use, for example multiple abilities can be used at the same time. To better accomodate a given play style weapons can be customized with silencers, laser sights or even a sniping scope.
For rendering Crytek's CryEngine 3 is used which comes with reduced system requirements compared to the first Crysis game. Since Crysis 2 is a multi-platform game, with major development focus on console, the graphics on launch day were only DirectX 9. DirectX 11 functionality was added later in a patch. We use the DX11 version and the high-res texture pack for our benchmarking.

http://www.techpowerup.com/reviews/N...2_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

DiRT 3

http://www.techpowerup.com/reviews/N...ages/dirt3.jpg
The latest addition to the Collin McRae Rally franchise, DiRT 3, of multi-format rally motorsport. DiRT 3 introduced more of the same great racing experience Collin McRae DiRT 2 gave you, but with better gameplay, and the new Gymkhana freestyle motor-acrobatics mode, which you'll more likely love than hate. It uses a more polished, performance-optimized version of the EGO engine, version 2.0, which takes advantage of more DirectX 11 features than version 1.0 used on Collin McRae DiRT 2, did.

http://www.techpowerup.com/reviews/N...3_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Dragon Age II

http://www.techpowerup.com/reviews/N...dragonage2.jpg
Dragon Age II is the second game in BioWare's Dragon Age franchise and was released in March 2011. As player, named Hawke, you will be able to pick your hero from several classes and grow him over the course of the adventure. Gameplay takes you through a linear narrated story of Hawke's rise to become the legendary "Champion of Kirkwall".
BioWare's Lycium Engine has support for DirectX 11, using tesselation, advanced dynamic lighting and camera effects like depth of field. We benchmark the DX11 version with details set to highest.

http://www.techpowerup.com/reviews/N...2_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Hard Reset

http://www.techpowerup.com/reviews/N.../hardreset.jpg
Developed by Flying Wild Hog, a studio that prides itself with the fact that its creation is PC-exclusive (bless them), Hard Reset is a first person shooter that's set in a future cyberpunk setting of a dystopian world. It reintroduces many of the gameplay mechanics that made classics such as Quake wicked fun, which today's tactical military shooters eroded, creating a 'void' for.
The game uses the studio's in-house Road Hog Engine, which isn't particularly heavy on new-generation DirectX features, but can still get taxing with some GPUs.

http://www.techpowerup.com/reviews/N...t_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Metro 2033

http://www.techpowerup.com/reviews/N.../metro2033.jpg
Metro 2033 is a first-person shooter game that is set in a post apocalyptic Moscow - as the name suggests inside the metro system. You will fight mutants or other humans who like to take away your shelter. The game has many gameplay elements similar to STALKER, also the engine has similar features. This is because two STALKER engine programmers left GSC Game World and started their own company which is now making Metro 2033.
The engine has support for all the latest eye candy like DirectX 11 and Tesselation. Unfortunately it leaves a less than optimized impression, making it a candidate to surpass Crysis for the highest hardware requirements. We test in DirectX 11 mode with details set to "Very High".

http://www.techpowerup.com/reviews/N...3_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

STALKER: Call of Pripyat

http://www.techpowerup.com/reviews/N...stalkercop.jpg
STALKER: Call of Pripyat takes places shortly after the events of the previous game STALKER: Shadow of Chernobyl. The player is one of many stalkers who are attracted by the Zone in hope of finding fame, wealth and artifacts. Over the course of the game you meet Strelok, the protagonist of the first STALKER game and team up with him to progress through the Zone.
An updated X-Ray Engine 1.6 powers the game with support for DirectX 11 using Compute Shaders for improved shadow rendering and tesselation to improve model quality.

http://www.techpowerup.com/reviews/N...p_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

StarCraft II

http://www.techpowerup.com/reviews/N...images/sc2.jpg
StarCraft II, released in July 2010, is a sequel to Blizzard's award-winning strategy game StarCraft. In the 26th century three species Terrans, Protoss and Zerg are at war. The campaign takes you through many missions on different planets where you have to face the various enemy factions, sometimes several of them. StarCraft II features a similar number of units as the original game, some of them new. Due to the massive success of the first game, Blizzard chose to focus large aspects of the game on multiplayer combat through Battle.net. The campaign serves as a good introduction to units and concepts and competitive multiplayer is where the action is at.
The StarCraft 2 engine supports only DirectX 9, but several patches have improved rendering quality and available options considerably. We test using a recorded 1 vs. 1 multiplayer replay in the late game phase. Please note that Star Craft II is very CPU limited on high-end cards, especially on lower resolutions, so you may not see much scaling between some cards.

http://www.techpowerup.com/reviews/N...2_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Total War: Shogun 2

http://www.techpowerup.com/reviews/N...es/shogun2.jpg
Set in 16th century feudal Japan, Total War: Shogun 2 takes the player on a quest for domination to conquer and unite the warlords of Japan. Moving away from the European setting of previous Total War games, the game is now designed around principles of the brilliant Chinese general Sun Tzu and his book "The Art of War". Gameplay is switched between real-time battles during which units on the battlefield are controlled and turn-based strategy which enable diplomacy, economy and production management. Taking control of a castles is comprised of several different stages which adds more complexity to warfare.
We benchmark using the highest settings in DirectX 11 mode, which was added via patch after release.

http://www.techpowerup.com/reviews/N...2_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

The Elder Scrolls: Skyrim

http://www.techpowerup.com/reviews/N...ges/skyrim.jpg
This isn't just a game, it's a masterpiece. A very large sandbox game that rejects the quality-quantity inverse-proportionality. By genre a role-playing game, TES: Skyrim combines some of the best elements of older titles in the franchise, with some new sandbox elements to churn out an extremely engaging, and addictive game. It makes use of Bethesda's Creation Engine, which isn't visually-intensive in that it doesn't use taxing graphics features, but the game's presentation itself, with large open worlds, end up taxing your hardware. Faster GPUs result in smoother gameplay with most eyecandy turned on.

http://www.techpowerup.com/reviews/N...m_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

3DMark 11

http://www.techpowerup.com/reviews/N...s/3dmark11.jpg
3DMark 11 is the very latest from the house of Futuremark, which has given out some of the most comprehensive benchmark applications for PC enthusiasts and gamers. 3DMark 11, as the name might probably suggest, makes use of Microsoft DirectX 11 API, and puts every feature at its disposal to use, creating astonishingly-realistic visuals. In the process, it evaluates DirectX 11 compliant GPUs, and lets gamers know what to expect from games from the near future that make use of the API, in terms of visual realism. The tessellation and depth of field tests are particularly of interest here.

http://www.techpowerup.com/reviews/N...1_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Unigine Heaven 2.0

http://www.techpowerup.com/reviews/N...ges/heaven.jpg
Unigine Heaven was one of the first demos that supported DirectX 11. Heaven is a technology demonstration for Unigine engine which supports DirectX 9 through 11 and OpenGL too. Version 2.0 adds more scenes and optionally more complex tesselation features. While there is some controversy surrounding the benchmark whether it is an accurate representation of what to expect from future games in regards to DirectX 11 we still chose it as test to get an insight into potential future gaming.

http://www.techpowerup.com/reviews/N...n_1024_768.gif http://www.techpowerup.com/reviews/N..._1280_1024.gif http://www.techpowerup.com/reviews/N..._1680_1050.gif http://www.techpowerup.com/reviews/N..._1920_1200.gif http://www.techpowerup.com/reviews/N..._2560_1600.gif

Power Consumption

Cooling modern video cards is becoming more and more difficult, especially when users are asking for quiet cooling solutions. That's why the engineers are now paying much more attention to power consumption of new video card designs. An optimized fan profile is also one of the few things that board vendors can do to impress with reference designs where they are prohibited to make changes to the thermal solution or components on the card.

For this test we measure power consumption of only the graphics card, via PCI-Express power connector(s) and PCI-Express bus slot. A Keithley Integra 2700 with 6.5 digits is used for all measurements. Again, the values here reflect card only power consumption measured at DC VGA card inputs, not the whole system.

We chose Crysis 2 as a standard test representing typical 3D gaming usage because it offers: - very high power draw - high repeatability - is a current game that is supported on all cards due to its DirectX 9 nature - drivers are actively tested and optimized for it - supports all multi-GPU configurations - test runs a relatively short time and renders a non-static scene with variable complexity.

Our results are based on the following tests:

Idle: Windows 7 Aero sitting at the desktop (1280x1024 32-bit) all windows closed, drivers installed. Card left to warm up in idle until power draw is stable.
Multi-Monitor: Two monitors connected to the tested card, which use different display timings. Windows 7 Aero sitting at the desktop (1280x1024 32-bit) all windows closed, drivers installed. Card left to warm up in idle until power draw is stable.
Average: Crysis 2 at 1920x1200, Extreme profile, representing a typical gaming power draw. Average of all readings (12 per second) while the benchmark was rendering (no title/loading screen).
Peak: Crysis 2 at 1920x1200, Extreme profile, representing a typical gaming power draw. Highest single reading during the test.
Maximum: Furmark Stability Test at 1280x1024, 0xAA. This results in a very high non-game power consumption that can typically be reached only with stress testing applications. Card left running stress test until power draw converged to a stable value. On cards with power limiting systems we will disable the power limiting system or configure it to the highest available setting - if possible. We will also use the highest single reading from a Furmark run which is obtained by measuring faster than when the power limit can kick in.
Blu-ray Playback: Power DVD 9 Ultra is used at a resolution of 1920x1200 to play back the Batman: The Dark Knight disc with GPU acceleration turned on. Playback starts around timecode 1:19 which has the highest data rates on the BD with up to 40 Mb/s. Playback left running until power draw converged to a stable value.

NVIDIA's new Kepler Architecture emphasizes power consumption, which is cleary visible in our tests. In the past, ATI always had a clear performance per Watt advantage, but this has changed apparently.

In idle, multi-monitor power, typical gaming and max. gaming both GTX 680 and HD 7970 consume very similar amounts of power. The differences here are too small to be significant in any way. However, consider that GTX 680 is faster than HD 7970, which means it ends up with a better performance per Watt score.

Furmark maximum power is about 50W lower than on AMD's card, but this is due to different power limiting methods limiting the boards at different power levels. GTX 680 is specified for up to 225 W, HD 7970 for up to 300 W, so it is only logical that they end up at different worst-case power levels.

Blu-ray media playback power consumption of NVIDIA's latest card is almost half of what the HD 7970 consumes. Good job, NVIDIA!

One missing feature on the GTX 680 is AMD's ZeroCore power, which basically turns off the card when the monitor is switched off, resulting in a power consumption of around 1 W. NVIDIA's GTX 680 will consume full idle power in that case, which is 14 W. This might be an important point for office systems that do not get turned off overnight. Or media PC systems that are always on to provide storage/background downloading, yet are not actively used.

http://www.techpowerup.com/reviews/N...power_idle.gif http://www.techpowerup.com/reviews/N...r_multimon.gif http://www.techpowerup.com/reviews/N...er_average.gif http://www.techpowerup.com/reviews/N...power_peak.gif http://www.techpowerup.com/reviews/N...er_maximum.gif http://www.techpowerup.com/reviews/N...wer_bluray.gif

Fan Noise

In the past years users would accept everything just to get more performance. Nowadays this has changed and people have become more aware of the fan noise and power consumption of their graphic cards.
In order to properly test the fan noise a card emits we are using a Bruel & Kjaer 2236 sound level meter (~$4,000) which has the measurement range and accuracy we are looking for.

http://www.techpowerup.com/reviews/N...2236_small.jpg http://www.techpowerup.com/reviews/N...jaer_setup.jpg
The tested graphics card is installed in a system that is completely passively cooled. That is passive PSU, passive CPU cooler, passive cooling on the motherboard and a solid state drive.
This setup allows us to eliminate secondary noise sources and test only the video card. To be more compliant with standards like DIN 45635 (we are not claiming to be fully DIN 45635 certified) the measurement is conducted at 100 cm distance and 160 cm over the floor. The ambient background noise level in the room is well below 20 dbA for all measurements. Please note that the dbA scale is not linear, it is logarithmic. 40 dbA is not twice as loud as 20 dbA. A 3 dbA increase results in double the sound pressure. The human hearing is a bit different and it is generally accepted that a 10 dbA increase doubles the perceived sound level. The 3D load noise levels are tested with a stressful game, not Furmark.

Idle fan noise is acceptable, even though we have seen much quieter cards.

Fan noise under load is still too high in my opinion. Yes, it's lower than HD 7970, and just a bit higher than GTX 580, but we've seen custom cooling solutions for high-end cards that are much quieter. Let's hope NVIDIA's board partners will do the same and bring out reduced noise GTX 680s.

http://www.techpowerup.com/reviews/N...noise_idle.gif http://www.techpowerup.com/reviews/N...noise_load.gif

Performance Summary

The graphs on this page show a combined performance summary of all tests and resolutions from previous pages. Each graph shows the tested card as 100% and all other cards' performance relative to it. A sixth graph summarizes all tests in all resolutions to calculate the total relative performance of the review sample.

http://www.techpowerup.com/reviews/N...es/perfrel.gif

http://www.techpowerup.com/reviews/N...rfrel_1024.gif http://www.techpowerup.com/reviews/N...rfrel_1280.gif http://www.techpowerup.com/reviews/N...rfrel_1680.gif http://www.techpowerup.com/reviews/N...rfrel_1920.gif http://www.techpowerup.com/reviews/N...rfrel_2560.gif

Performance per Watt

Using the relative performance scores from the previous page and the typical gaming power consumption result, the following graphs show efficiency of the cards in our test group.

http://www.techpowerup.com/reviews/N...s/perfwatt.gif

http://www.techpowerup.com/reviews/N...fwatt_1024.gif http://www.techpowerup.com/reviews/N...fwatt_1280.gif http://www.techpowerup.com/reviews/N...fwatt_1680.gif http://www.techpowerup.com/reviews/N...fwatt_1920.gif http://www.techpowerup.com/reviews/N...fwatt_2560.gif

Performance per Dollar

If you are looking for the best bang for the buck, then you will love this graph. We looked up the current USD price of each card on the popular online shop Newegg and used it and the relative performance numbers to calculate the Performance per Dollar Index.

http://www.techpowerup.com/reviews/N...perfdollar.gif

http://www.techpowerup.com/reviews/N...ollar_1024.gif http://www.techpowerup.com/reviews/N...ollar_1280.gif http://www.techpowerup.com/reviews/N...ollar_1680.gif http://www.techpowerup.com/reviews/N...ollar_1920.gif http://www.techpowerup.com/reviews/N...ollar_2560.gif

Dynamic Overclocking Introduction

NVIDIA's GeForce GTX 680 introduces the biggest overclocking change since 2D/3D clocks.

So far, all graphics cards that have been released came with a specific clock frequency that was active during typical 3D rendering. There were some exceptions for when the card overheated or an application ran at extremely low load (like the Windows 7 Aero desktop), but overall 3D clocks were static and never changed.

Now NVIDIA is introducing a new concept called "Boost Clocks". These enable the GPU to run at increased frequency depending on a number of input factors that are used to determine whether it's a good idea to run at higher clocks or not. According to NVIDIA, their algorithm takes into account board power consumption, temperature, GPU load, memory load and more. With this information, it should be possible for the driver to make a decision whether a certain clock state is safe in terms of power usage and heat output. Obviously we don't want cards burning up due to automatic overclocking.

Instead of talking GPU clock, memory clock and shader clock (which are gone on Kepler) we will now talk about base clock, boost clock and memory clock.
- Base clock is the minimum guaranteed clock speed the card will run at in non-stress testing applications (Furmark, OCCT). For GeForce GTX 680 this is set to 1006 MHz.
- Boost clock is the average clock frequency the GPU will run under load during typical gaming. This is set to 1058 MHz for GTX 680. Please note that the actual clock frequency will often exceed the boost clock specification of 1058 MHz.
- Memory clock. This is unchanged and exactly the same as before, representing the speed the memory chips are running at.

As many overclockers know, increasing voltage increases overclocking potential. NVIDIA uses this to ensure that the GTX 680 will be stable over the whole range of (stock) GPU frequencies, which also increases yields as even GPUs that wouldn't normally make the clock qualification can now be used - at increased voltage.

For the following graphs we used a run of our benchmarking suite at 1920x1200 (to reduce time). GPU-Z was used to log both the GPU frequency and GPU voltage.

The following graph shows the result of this test at NVIDIA stock settings of 1006 MHz base clock and 1058 MHz boost clock. I gave each data point some transparency, so rarely used combinations are less visible, which helps guide the eye.

http://www.techpowerup.com/reviews/N...vs_voltage.gif

As you can see there is a clear correlation between GPU clock and voltage. When a higher GPU clock is selected by the driver, it will also pick an increased voltage to ensure maximum stability.

We can also observe that the typical range of clock speeds is between 1006 MHz and 1110 MHz. Voltages range from 1.075 V to 1.175 V. The voltage range will vary from card to card as Kepler uses a dynamic VID algorithm that selects a certain base voltage based on manufacturing properties of the GPU. For Fermi cards this ASIC Quality value can be read using GPU-Z (not supported on Kepler yet).

Please note that any increase in voltage will lead to increased power consumption and heat output, which requires that NVIDIA's clock algorithm monitors power and heat on a constant basis, to possibly lower the clock speeds again.

In the following testing we will investigate how the system works and check out what additional manual overclocking will do for the card.

Power Limit

Introduced first with NVIDIA's GeForce GTX 580, the power limiter is in full force again on the GTX 680. It can not be disabled according to NVIDIA and board partners must use it. In order to cater to overclockers, ASUS just left out the whole circuitry on their GTX 580 Matrix for example, whether this will be possible again with GTX 680 seems doubtful.

NVIDIA is using three INA219 power monitoring chips on their board, which provide the driver with power consumption numbers for 12V PCI-Express and the two 6-pin PCI-E power connectors. The driver then decides whether the board is running below or above the NVIDIA configured power limit. Depending on the result of the measurement, the dynamic overclocking algorithm will pick a new clock target roughly every 100 milliseconds (three frames at 30 FPS). AMD's power limiting system works slightly different. It does not measure any actual power consumption but relies on an estimate based on current GPU load (not just %, but taking more things into account). This makes the system completely deterministic and independent of any component tolerances and reduces board cost. AMD uses a hardware controller integrated in the GPU to update measurements and clocks many times per frame, in microsecond intervals, independent of any driver or software.

Each of these methods has its advantages and disadvantages. One thing that I like about NVIDIA's approach is that with a bit of hardware modding, the sensors might be tricked into thinking power is lower.

NVIDIA has set the adjustable range for the power limit to +32% up and -30% down. So far this is a hard limit, no software based method to circumvent it has been found.

I ran some tests with the power limiter configured at different levels to see how actual performance would change.

In the graph below, the red line shows the performance improvement (or reduction) of GTX 680 compared to the baseline value of 0% (black line). The green dotted line shows the average of the red line.

http://www.techpowerup.com/reviews/N...s/power_up.gif http://www.techpowerup.com/reviews/N...power_down.gif

Adjusting the power limit up by its maximum yields 0.9% in real life performance. Take this in comparison to the -8.3% that we see when setting the power limit as low as possible.

The conclusion of these tests is that NVIDIA picked a decent default power limit on their board which doesn't limit typical gaming by much. However, there is still a bit of headroom to maximize performance (again, without any manual overclocking, using just the default always-on dynamic overclocking).

Temperature

In the briefings NVIDIA mentioned that GPU temperature is taken into account for dynamic overclocking. Wait, what? Will the card run slower when it's running hot ? Yes it does.

The following graph shows how changes in GPU temperature affect the selected clock. We tested this with a static scene that renders the same scene each frame, resulting in constant GPU and memory load, which would otherwise affect this testing.

GPU clock is plotted on the vertical axis using the blue MHz scale on the left. Temperature is plotted on the vertical axis too, using the red °C scale on the right. Time is running on the horizontal axis.

http://www.techpowerup.com/reviews/N...ck_vs_temp.gif

We see a clearly visible downward step pattern on the clock frequency curve as the temperature increases. This is not a gradual change, but the steps happen at what looks like predefined values of 70°C, 80°C, 85°C, 95°C. Once temperature exceeds 98°C, thermal protection will kick in and GPU clock drops like a rock to 536 MHz and then even 176 MHz trying to save the card from overheating.

Each step is 13.5 MHz in size, which results in a total clock difference of 40 MHz going from below 70°C to 95°C - with the exact same rendering load, all happening transparently by the NVIDIA driver.

For end-users this means that to maximize dynamic overclocking potential, they would have to run at temperatures below 70°C. Otherwise they will end up with up to 40 MHz less if their card runs above 90°C. Even users who don't care about manual overclocking will have to consider this. The dynamic overclocking in the driver is always active and can not be turned off.

Performance now being based on temperature will pose an interesting challenge for system assemblers and case manufacturers as they will now have to focus even more on thermals, while still trying to keep noise levels acceptable. How will reviewers test their cards? With an open bench? a normal case? or a worst case [sic] ?

I ran some additional testing with the card's fan speed set to maximum, which results in much lower temperatures of the card, directly increasing performance (without any manual overclocking).

http://www.techpowerup.com/reviews/N...ges/fan100.gif

It looks like on average just setting the fan to 100% results in a 0.8% performance increase. Again, this is without any overclocking or other tweaking. 0.8% is not very significant, but it still shows that there is now another variable that needs to be considered when trying to maximize performance. It also means that cases with really bad ventilation will suffer from a (small) performance penalty when a GeForce GTX 680 is installed in the case.

Manual overclocking

On the previous page we discussed how NVIDIA's GPU Bost works. If you just skipped here, go back one page and read it. It's really important to understand the basics of the dynamic clock adjustments before trying to understand the implications for manual overclocking.

Manually setting a specific GPU clock on the GeForce GTX 680 is not possible. You can only define a certain offset that the dynamic overclocking algorithm will _try_ to respect. If the card runs into the power limit or something else comes up, the clocks will be lower than requested. Think of it more as a "best effort plzplz" value than a hard setting.

NVIDIA has defined a hard limit of +549 MHz for the clock offset, which will certainly upset some extreme overclockers with liquid nitrogen. As mentioned several times before, there is no way to turn off dynamic clocking, which means there is no way to go back to the classic overclocking method. Also directly accessing hardware to write clocks to the clock generator won't work as the algorithm in the driver will instantly overwrite it with what it thinks is right. The clock offset simply acts as additional input variable for the dynamic clock algorithm which also takes into account things like power consumption, temperature and GPU load.

This means that no matter how hard you try using clock offsets, power limits and voltage settings, the card will always reduce clocks when it thinks it has to.

Increased GPU Clock

I tried running at various clock offsets to check what real-life performance boost they offer, because the clock offset is only a target value and actual clocks during gaming may differ, despite the selected clock offset.

I picked +50 MHz and +100 MHz as both are rock stable and they provide some insight into how well clock scaling will work. Please note, both these runs were with board power limit set to default.

http://www.techpowerup.com/reviews/N...mages/oc50.gif
Using a small overclock of just 50 MHz we see immediately that performance gains are relatively small with only 1.3% faster than the GTX 680 running its default boost clock algorithm.

http://www.techpowerup.com/reviews/N...c50_clocks.gif
Given a selected clock offset of +50 MHz, we would expect clocks between 1056 MHz (base clock + 50 MHz) and 1150 MHz (highest dynamic clock + 50 MHz). While the majority of clocks are bunched up in that region indeed, we do see a good amount of clocks below 1056 MHz, all the way down to the default base clock of 1006 MHz and even below.
These unexpected clocks can be explained by dynamic overclocking reducing clock speeds because a certain game scene causes it to run into the TDP power limit, or similar situations. Increased temperature from overclocking alone can not account for the difference, as it can only reduce clocks by 40 MHz, which would still give us a lowest clock of 1016 MHz.

Next I tried increasing clocks further using an offset of +100 MHz.

http://www.techpowerup.com/reviews/N...ages/oc100.gif

Here we see 2.9% real life performance gained. Also note the increased amount of upward spikes in Shogun 2, BF3 and Dragon Age 2, which are caused by games that love to run higher clocks, yet are not blocked by the power limiter from doing so.

http://www.techpowerup.com/reviews/N...100_clocks.gif

The clock distribution shows an even more stretched out range of clocks. Again, ranging from below 1000 MHz, through 1106 MHz (which would be our new base clock with the offset taken into account), all the way up to a maximum of 1215 MHz.

Increased Memory Clock

I ran a quick test, bumping memory clock by +100 MHz to 1603 MHz using the clock offset option.

http://www.techpowerup.com/reviews/N...images/mem.gif

As the data shows, memory clock is constant during 3D, the drops you see are from when the card switches to 2D clocks between tests.
This means that dynamic overclocking is only active for the GPU clock, which will make things a bit easier.

Voltage Tuning

Adjusting voltage has become a little bit more difficult due to dynamic overclocking. It seems that the NVIDIA driver enforces a voltage minimum, which ensures dynamic OC is stable. Voltage selection to higher levels is possible, but limited to the driver maximum of 1.175 V. The voltage controller on the board has no support for I2C, so it can not be programmed directly.

For example, if the dynamic clock algorithm decides it wants to run the card at 1.125 V, then the voltage will be 1.125 V, despite any manually set voltage between 0.825 V (the available minimum) and 1.125 V.
Once you manually set 1.137 V (one step up from 1.125 V), this voltage will become active. However, should the dynamic OC algorithm decide that it needs more juice because it wants to set a higher clock, your setting will be overriden for the time that this clock/voltage combination is active.

As mentioned before, each card's individual set of voltages depends on a value stored in the GPU during production time, so actual voltages may differ. The mechanism of voltage control will be the same however.

Putting it all together

So now that we know how overclocking works, let's put it to some good use.

Finding maximum clocks is similar as before, except that you have to search for a new stress testing application as Furmark will run at significantly reduced clocks due to power limit, so the actual requested clocks will not be tested. Further, 3DMark11 does not seem a good choice either, as it let me run much higher clocks than were stable in other tests. Out of the games in our benchmark suite I found Unigine Heaven to be the best for stability testing. It also notifies you when the card crashes, causing a driver reset. If that happens, it's best that you restart the system to make sure it is fully stable and working well, before trying a slightly reduced clock frequency.

Personally I think with the latest generation of cards from both AMD and NVIDIA, that both have some sort of power limiting system that messes with your overclock, the best stress testing is playing your favorite games. So go ahead and find what looks like a stable set of clocks, set them and don't worry about them until you see rendering errors or crashes during your gaming, then take the clocks down a notch and repeat. This is also more efficient and more fun than staring at stability tests all day.

And don't forget to set the board power limit to +32%, to ensure the power limit will not drop down your clocks. We did so for all further OC testing.

Maximum OC

http://www.techpowerup.com/reviews/N...es/gpuz_oc.gif

Maximum stable clocks of our card are 1159 MHz GPU / +153 MHz clock offset (15% overclock) and 1833 MHz Memory (22% overclock). (HD 7970 review: +16% / +25%)

Both overclocks are great, especially seeing memory overclock so far is nice. Previous Fermi generation GPUs did much worse here. It's also good to see that there is still some headroom left in the cards for manual tweaking and dynamic OC didn't eat up the whole OC potential.

Please note that the dynamic overclocking mechanism will automatically adjust GPU clocks to lower level in case it thinks it is required to do so. Also it will increase GPU voltage to a level that's guaranteed to work with the selected clock - when possible. Once it hits the maximum voltage, it will not increase voltage any further. This implies that basic overclocking is more stable now, but once you go beyond the qualified clock range, stability testing is still required.

Overclocked Performance

Using these clock frequencies we ran the 1920x1200 resolution tests of our benchmarking suite to evaluate real life performance gains. I also added a manually overclocked (but not overvolted) HD 7970 reference design, so you can see what to expect when manual overclocking is taken into account for both cards. Both results are directly comparable, higher vertical position implies higher performance in that specific game. The 0% line is the performance that we saw from the non-overclocked GTX 680, just running the default dynamic OC algorithm, without any clock offsets or power limit change.

http://www.techpowerup.com/reviews/N...oc_vs_7970.gif

So even when both cards are manually tweaked to deliver maximum performance, NVIDIA's GeForce GTX 680 is still on top. Overclocking the Radeon HD 7970 does give it a nice performance boost though, which allows it to surpass the out-of-the-box GTX 680 experience.

To be honest, when I started overclocking GTX 680 I was extremely sceptical how well manual overclocking would work, and how much dynamic OC would interfere. As the previous pages have shown, the overclocking process is more complex, there are a few gotchas to consider, but it is still possible to get a significant amount of extra performance out of the card - in our case +11.9% vs. the stock GTX 680.

Temperatures

http://www.techpowerup.com/reviews/N...mages/temp.gif

Temperatures are decent. It would have been nice to see lower load temperatures on the GTX 680, but this doesn't seem viable given the noise levels of the card. As mentioned on the previous page, lower temperatures directly translate into higher performance, as dynamic overclocking reduces clocks when temperatures get high.

Clock Profiles

Modern graphics cards have several clock profiles that are selected to balance power draw and performance requirements.
The following table lists the clock settings for important performance scenarios and the GPU voltage that we measured. We measure on the pins of a coil or capacitor near the GPU voltage regulator.

<table class="tputbl">
<tr>
<th scope="col"></th>
<th align="center" scope="col" width="70">Core <br />
Clock</th>
<th align="center" scope="col" width="70">Memory <br />
Clock</th>
<th align="center" scope="col">GPU Voltage <br />
(measured)</th>
</tr>
<tr>
<th scope="row">Desktop</th>
<td align="right">324 MHz</td>
<td align="right">162 MHz</td>
<td align="right">0.99 V</td>
</tr>
<tr class="alt">
<th scope="row">Multi-Monitor</th>
<td align="right">550 MHz</td>
<td align="right">1502 MHz</td>
<td align="right">0.99 V</td>
</tr>
<tr>
<th scope="row">Blu-ray Playback</th>
<td align="right">549 MHz</td>
<td align="right">405 MHz</td>
<td align="right">0.99 V</td>
</tr>
<tr class="alt">
<th scope="row">3D Load</th>
<td align="right">1006 - 1111 MHz</td>
<td align="right">1502 MHz</td>
<td align="right">1.20 V</td>
</tr>
</table>

Value and Conclusion

<table width="100%" cellpadding="5" cellspacing="0" id="result">
<tr><th>http://www.techpowerup.com/images/dollar.gif</th>
<td>

NVIDIA GeForce GTX 680 is available for as low as US $499.

</td></tr><tr>
<th>http://www.techpowerup.com/images/thumbup.gif</th>
<td>

16% performance increase over GTX 580
Reasonable price
Massive leap forward in energy efficiency
Dynamic overclocking works well
Good additional OC potential
Native full-size HDMI & DisplayPort output
Adds adaptive VSync and new Anti-Aliasing modes
Up to four active displays now, makes surround possible with one card
Adds support for PCI-Express 3.0 and DirectX 11.1
Support for CUDA and PhysX

</td>
</tr>
<tr>
<th>http://www.techpowerup.com/images/thumbdown.gif</th>
<td>

Dynamic OC can't be turned off
Manual overclocking more complicated than before
Noisy in 3D
Voltage controller has no software voltage control neither monitoring
No technology similar to AMD's ZeroCore power
+500 MHz max OC and power limit won't work for extreme overclockers

</td></tr>
<tr><th>9.5</th>
<td>NVIDIA clearly has a winner on their hands with the GTX 680. The new card, which is based on NVIDIA's GK104 graphics processor, that introduces the Kepler architecture, is a significant leap forward both in terms of performance and GPU technology. Technically GK104, as its name reveals is an upper mid-range GPU, not a pure high-end part. Following NVIDIA's naming convention such a chip would be called GK100. This subtle difference makes the GeForce GTX 680 even more impressive. Technically we'd have to compare it to GTX 560 Ti, not GTX 580. Even when compared to GeForce GTX 580, the performance improvement of GTX 680 is excellent. Averaged over our testing it increases performance by 16% (+37% vs. GTX 560 Ti!), and easily beats AMD's HD 7970. Achieving such performance levels nowadays has to go with improved performance per Watt, as modern high-end graphics cards are limited by power consumption and heat output.<br />
NVIDIA claims massively improved power consumption with their latest architecture, which is confirmed by our testing. Compared to previous-generation cards, we see over 30% performance per Watt improvement. Every power measurement has gone down, yet performance has gone up. Very nicely done.<br />
Reduced power consumption is a foundation that NVIDIA's new dynamic overclocking algorithm is built upon. Instead of a single fixed GPU clock, NVIDIA now markets two values. Base clock (1006 MHz), which is the guaranteed minimum clock speed in normal gaming, and boost clock (1058 MHz) which is the average clock speed reached in those scenarios. Depending on real-time monitoring, which includes board power, temperature and GPU load, the driver decides on a clock target beyond base clock, up to 1110 MHz. This approach enables a performance boost in most games, yet ensures that the card does not overheat by drawing too much power in more demanding scenes or games. In this review we did extensive testing of NVIDIA's Boost Clock algorithm and can say that it works exceedingly well for consumers who just want to install their new card and start playing games. Things do get more complicated for enthusiasts who are looking to squeeze the maximum out of their card. We confirmed that dynamic overclocking did not eat up all the overclocking headroom on the GeForce GTX 680. With manual overclocking we were able to increase performance by another 12% over the out of the box experience, which already uses dynamic overclocking. There are a few gotchas left with dynamic OC, for example that it can not be turned off. Also boost levels are dependent on temperature, which means that there might be a small performance drop when the card is used in badly ventilated cases. Last but not least, being based on a physical measurements, means that dynamic OC is dependent on component tolerances. Two cards that you buy will perform differently out of the box. I am worried that we might see people buying ten cards, pick the fastest one and then return the rest, which will turn this into a nightmare for retailers. Remember, this applies even to cards where no manual overclocking is done.<br />
With lower power consumption usually comes reduced fan noise, too. NVIDIA did reduce noise levels of their cooler by a bit, but it's still far from what I would consider a quiet experience, especially in 3D. Custom design cards will hopefully address that. Another area that has seen improvement is NVIDIA's display output logic, which supports up to four active displays now. This makes it easier than ever before to setup a multi-monitor gaming system, with just one NVIDIA card. NVIDIA has also introduced new software features like adaptive V-Sync which will help reduce stutter and tearing in games, and the new anti-aliasing modes will make them look even prettier.<br />
Pricing of the GeForce GTX 680 is reasonable, with $499. When compared to AMD's $549 HD 7970, the GTX 680 is the clear winner. It offers better performance at similar power consumption and comes with more features. Looking at the board design I am sure that there is still plenty of headroom for price reductions - a price war with AMD would be great for customers, resulting in better pricing in these tough economic times. Let's hope it happens.</td></tr>
<tr><th></th><td>http://www.techpowerup.com/images/editorschoice.gif</td></tr>
</table>