Logo   Banner   TopRight
TopUnder
Transparent
The State of DirectX 11 - Image Quality & Performance
Transparent
Date: May 12, 2010
Section:Graphics/Sound
Author: Michael Lin
Transparent
The Promise of DirectX 11
Touted as one of the biggest milestones in Windows games development in recent years, DirectX 10 ultimately didn't live up to the hype as far as end-users were concerned. A large part of the disappointment was perhaps due to a misunderstanding of what DirectX 10 was supposed to be. DirectX 10 was a fundamental refresh of the entire API and it brought a much-needed reboot to an aging API model. The reboot removed a lot of the legacy clutter left over from over a decade of previous DirectX developments. DirectX 10 was never really about the end users at all, it was really about the developers.

It also didn't help that DirectX 10 was tied to Windows Vista, which was unpopular with gamers from launch. On top of that, due to the revamped API model which required developers to adjust how they programmed games, DirectX 10 received sluggish developer adoption. Considering the lukewarm reception DirectX 10 received a few short years ago, it isn't too surprising that many gamers are now approaching DirectX 11 with some caution.



In terms of technology, DirectX 11 isn't the API defining, behind-the-scenes reboot that DirectX 10 was. It's an incremental update, much in the same way that Windows 7 is an update of Vista. At the end of the day, DirectX 11 is basically just DirectX 10 with a bunch of new features stuck on. However, we'd speculate that DirectX 11 should have a much larger impact on the end-user experience than DirectX 10 did because it brings several new features to the table that will make creating advanced graphical effects seldom seen in past games, much easier.

Some critics have said that Windows 7 is everything that Vista was supposed to be. Is DirectX 11 everything DirectX 10 was hyped up to be?  We couldn't let such a loaded question go unanswered, even if we did pose it ourselves. Now that DirectX 11 hardware is readily available and compatible games are on the shelves, it's high time we took stock of what DirectX 11 has to offer to the end-user. On the following pages, we're going to take a long, hard look into the state of DirectX 11 and what it means to the community as a whole.
Transparent
What's new in DirectX 11?

The DirectX 11 API is a superset of DirectX 10, including all of DirectX 10's features inside of it, in addition to a couple new ones. The new features are the addition of native support for tessellation, improved multi-threading support, two new texture compression algorithms, increased texture cache, Shader Model 5.0, and the DirectCompute API. Out of this list, the only features that will have any real impact on gamers are hardware tessellation, better multi-threading and the DirectCompute API. The rest will be welcomed by developers, but will have little direct impact on end users.

Multi-Threading
Support for multi-threaded processing has been implemented separately by ATI and NVIDIA for several years in their respective developer API's, allowing multi-core CPUs to communicate with the GPU in parallel. However, DirectX 11 takes this a step further, allowing for even more granular multi-threading. Now, the application, graphics driver and DirectX runtime can all run in separate threads. While this won't directly allow for any new image quality enhancements, it should result in better performance, especially moving into the future as processors become even more heavily multi-threaded.

Tessellation
The characters in modern games are essentially complex meshes that an artist creates using an authoring tool. The artists produce high resolution character meshes which are extremely complex. However with increased complexity comes high data and bandwidth requirements. The original high resolution characters meshes are simply too big, so the models are thinned out, reducing complexity, when they are imported into a game. This reduces the data requirements to a level that will fit within the bandwidth envelopes of currently available hardware, but at the cost of reduced detail. Tessellation is one method of getting some of that detail back. This is accomplished with procedural generation techniques right on the GPU, which neatly avoids the bandwidth problem.



Tessellation Example: Fully Rendered



Tessellation Example: Wireframe

The two tessellation example comparison images are from an AMD tech demo. It illustrates how tessellation can be used to greatly increase the geometry detail of a game character. In AMD's tech demo, the female character's head gear, armor and clothing are heavily tessellated, adding a significant amount of geometry detail, making the tessellated areas appear more defined with greater depth. However tessellation isn't limited to characters, it can be applied to any geometry objects in a game, including the terrain.

Tessellation has been implemented in hardware by ATI since 2001. However few developers chose to implement it in their games due to general lack of support from programming APIs and it didn't help that NVIDIA wasn't playing ball at the time. DirectX 11 changes all of that by making hardware tessellation support part of the standard. This takes away ATI's "edge" but at least there will be more titles actually implementing tessellation. It's worth noting that ATI's older hardware tessellator found in its older generation hardware is not compatible with DirectX 11, since the DX11 tessellator specification is a superset of ATI's approach.

DirectCompute API
Perhaps the biggest boon for gamers is the introduction of the DirectCompute API. Launched with DirectX 11 and now part of the DirectX family, this API provides support for general-purpose computing on graphics processing units (GPGPU). You've probably heard quite a bit about GPGPU already and perhaps know about NVIDIA's CUDA and AMD's Stream initiatives. DirectCompute is Microsoft's entry into this arena.

DirectCompute allows for more general computing processing to happen right on the GPU. This can both boost performance by offloading some processing from the CPU and allow for new graphics effects that were previously impossible or too performance intensive on a CPU. The list of applications for DirectCompute is lengthy, it currently includes artificial intelligence, ambient occlusion, physics, post-processing, ray tracing, and video transcoding, among many other possibilities. DirectCompute is also available on DirectX 10 and 10.1, though some of the features will be unavailable, reducing its effectiveness.



Screen-space Ambient Occlusion Example


Depth of Field Example

DirectCompute offers a lot of possibilities that will help make existing effects more efficient and new effects possible. One of the hot graphical features of the last two years is definitely screen-space ambient occlusion (SSAO). The effect has been a staple in 3D rendering for some time, but was first introduced in a game by Crysis. Ambient occlusion is a shading method that allows for more realistic light and reflection models. Another graphical effect we'll be seeing a lot of is depth of field, which attempts to approximate the way a camera lens focuses on objects. While both of these effects were possible before DirectCompute, they were very computationally expensive. Expect to see these effects, and others, much more often from now on.

Now that we've taken a look at the new features DirectX 11 has to offer, it's time to take a look at some currently available games on some currently available hardware and see if it delivers in practice.
Transparent
Scope, Test Setup & Methodology
As previously mentioned, we intend to explore the state of DirectX 11  in this article. That's a very tall order and one we cannot hope to fulfill unless we set out some boundaries and determine a specific scope for our examination. It would be nearly impossible and rather frivolous to encompass an evaluation of DirectX 11 from the perspective of everyone affected by its introduction -- such as the game developers, hardware manufacturers and end consumers -- into a single article.

To help keep the article at a manageable scope, we are only going to explore the current state of DirectX 11 from the perspective of the end consumer. Anything that is or should be transparent to the consumer, such as API optimizations that let developers program more efficiently, will not be covered. Note that this is not necessarily the same as exploring the current end user experience. We want to focus on examining the potential of currently available DirectX 11 software. We will focus our attention on two fronts; performance and image quality. Specifically, we are interested in examining the differences in performance between DirectX 9.0c and DirectX 11 and the image quality enhancements of DirectX 11, if any.

HotHardware's Test System
System Specifications, Video Cards Used & Games Tested

AMD Phenom II X4 955
(3.20GHz Quad-Core)


AMD 785G Chipset Motherboard

2GB DDR3 1333MHz CL7

On-Board 10/100/1000 Ethernet
On-Board Audio

WD "Raptor" 150GB Hard Drive
10,000 RPM SATA

Windows Vista Ultimate
Catalyst 10.3
DirectX 9, 10, 11




Video Cards:

Diamond Radeon HD 5770 1024MB
Diamond Radeon HD 5870 1024MB



Games:
Aliens vs Predator
BattleForge
Battlefield: Bad Company 2
Dirt 2
Unigine Heaven Benchmark


Performance Tests:
For our performance tests, we used a single system configuration (see above). Our goal is to compare the performance of DirectX 11 rendering paths in currently available games with other rendering paths like DirectX 10 and especially DirectX 9.  The April 2010
Steam hardware survey reveals that Windows XP 32bit is still the most popular OS with 36.79% of all Steam users who participated in the survey, although Windows 7 64bit is quickly catching up with 26.39%, a 2.1% gain from the last survey. So it would seem that a lot of users are still stuck with using DirectX 9 rendering paths, even if most of them have DirectX 10 and 11 hardware. Therefore we focused on the comparison between DirectX 9 and 11 for most of our testing, though we also tested DirectX 10 and we'll be reporting any interesting results there too.

Since the focus of the test is on contrasting the different DirectX rendering paths, very little attention will be given to the hardware running the tests. This is not a hardware focused article. Our only goal with hardware selection was to choose a respectable gaming system that many gamers upgrading to DirectX 11 will likely have. We are interested in how these games and DirectX rendering paths perform in a real-world mid-range system.

While our goal was to use a single consistent setup for all tests, we ultimately chose to test two different video cards since there seems to be two separate price 'sweet spots'. The first sweet spot appears at around $150, and it is represented here by the Diamond Radeon 5770. Video cards in this price point are currently an excellent value and hard to beat without shelling out significantly more, making it our value sweet spot. The other sweet spot is at the high end at $400, represented here by the Diamon Radeon 5870. This is the edge of single-card performance and a more powerful setup will require multiple cards which significantly bumps up the price, making this price point the high-end sweet spot. We expect both of these price points and video cards to be very popular and chose them to represent our hypothetical "average" PC gamer.

It's worth noting that this article was conceived before NVIDIA's DirectX 11 products were available and they are not represented here for that reason alone. You may also notice that the test system is a full AMD platform, which is largely due to convenience at the time of article conception. We'd like to remind everyone that this article is about software, and the hardware chosen is purely for the purposes of representing popular price points in order to illustrate a hypothetical "average" gamer. This article makes no further comment about hardware.

Image Quality Tests: For each of the games in our test, we compared the image quality of the game's DirectX 9 rendering path with its DierctX 11 rendering path. All image quality tests were performed with the Diamond Radeon 5870 and the same system configuration used in the performance tests. The video settings for each game were set to their highest settings for all of the image quality tests.

There are cases where a certain game will not offer certain features, like anti-aliasing or screen-space ambient occlusion (SSAO), in a DirectX 9 rendering path. This means that if both the DirectX 9 and 11 rendering paths were configured to their maximum available settings, the DirectX 11 rendering path would be doing significantly more work, perhaps losing some performance and also gaining image quality. We view this as a necessary part of comparing the different rendering paths and made no attempt to perform a direct setting-for-setting comparison.

Image quality was judged both in-game and with the aid of a large series of screenshots. Screenshot of each game were taken using the FRAPS utility in both DirectX 9 and 11 rendering modes. Studious effort was put into ensuring that screenshots were taken in matching sets, so the DirectX 9 screenshots would have matching DirectX 11 counterparts which are the best recreations of the DirectX 9 shot we could manage. Several dozen sets of matching screenshots were produces for each game and only the best 4 for each game are included in this article. The screenshots were judged on how well they illustrate differences between DirectX 9 and 11.

The screenshots were taken at 1920x1200 resolution and saved by FRAPS as uncompressed bitmaps. The screenshot sets chosen for inclusion in the article were cropped and resized in Adobe Photoshop and saved as high quality JPEGs, no other alterations were made to the images.While using JPEG does introduce a certain amount of compression artifacts to the images, we felt that they did not interfere with the illustration of the image quality differences. After all, if the difference is so subtle that simple JPEG compression artifacts could prevent them from being noticed in a direct side-by-side comparison, how are you supposed to notice it in-game?
Transparent
Aliens vs Predator: Image Quality


The Aliens and Predator franchises should require no introduction. The two franchises have been featured in countless movies, comics and games over the last three decades, both separately and together in several Aliens and Predators match-ups including the classic "Aliens versus Predator" PC game from 1999. For those that remember the game, it featured one of the most atmospheric experiences of its time, despite being saddled with DirectX 6 era graphics. The developers of the 1999 classic, Rebellion Developments, are back at it again with another Aliens vs Predator game, this time in glorious DirectX 11.

While lens flares was one of the hot must-have graphical features at the time of Rebellion's original Aliens versus Predator game, the new game embraces DirectX 11 and implements hardware tessellation and the latest dynamic lighting and shadow algorithms with DirectCompute. With the DirectX 11 rendering path enabled and the settings cranked up, Aliens vs Predator is a very good looking game. Check out the following video for an overview of Aliens vs Predator's DirectX 11 graphical features.



It's worth noting that anti-aliasing is only available with the DirectX 10 and 11 rendering paths. If you are using DirectX 9, the option simply won't be available for selection in-game, although you could still force FSAA through your drivers.

We chose to do our image quality comparisons for Aliens vs Predator between Dx11 and Dx9, with the AA turned on in Dx11 mode. We chose this because AA in Aliens vs Predator is implemented deep in the rendering path and disabling it also effects other graphical features such as lighting. AvP isn't the only game to do this and we had to make similar concessions with Bad Company 2.

   

Image quality differences appear to be minimal at first glance. The game looks great in both Dx9 and Dx11 modes. However upon closer inspection, we see that Dx11 rendering offers some subtle graphical improvements. Lighting effects are more pleasing and the shadows have more depth. The graphics in Dx11 seems to have more contrast overall, the shadows are darker and the lights are brighter.

As the video illustrates, Aliens vs Predator makes extensive use of hardware tessellation in the DirectX 11 rendering mode. In fact, the geometry detail in AvP is dynamic. Depending on how far a player is from an in-game object, the game engine will dynamically alter the level of geometry the object possesses. Tessellation is used to seamlessly alter the level of geometry detail. This is easy to spot when viewing the game with wire frames turned on, but when the geometry is covered in mapped and lighted textures and the wire frames are turned off, we had a hard time spotting geometry differences between Dx9 and Dx11.

   

We didn't bother including any comparison screenshots for DirectX 10, simply because the graphical differences become even smaller. In side-by-side screenshot comparisons of DirectX 11 and DirectX 10 with Aliens vs Predator, we really couldn't notice any differences. Any differences that may exist would likely be even less noticeable in-game when the objects are in movement and motion blur filters are applied. The main graphical feature you will miss out on with DirectX 10 compared to 11 is hardware tessellation. DirectCompute, and therefore all of the lighting effects, will still be available.

Overall, Aliens vs Predator is a very good looking game. The DirectX 11 rendering mode certainly adds some extra visual interest but the difference is hardly night and day. The added image quality enabling DirectX 11 features brings adds some extra realism, that is, if you have time to stop and spot them in the mayhem. For the best experience, we would definitely recommend DirectX 11, but it would hardly be a disaster if you had to play the game in DirectX 9.

Transparent
Aliens vs Predator: Performance & Verdict


When it comes to system requirements, Aliens vs Predator is a relatively demanding game. The recommended systems requirements call for a Core 2 class processor or a newer Athlon X2. While the 2GB memory requirement should be easily met, the 17GBs of free hard drive space the game will require is quite steep. The recommended graphics requirements call for a NVIDIA GeForce 8800 class card or an ATI Radeon HD 2900. This is roughly equivalent to current generation mid-range cards.



For our benchmarks, all graphics settings were turned up to their highest levels. It's worth noting that AvP seems to perform its anti-aliasing using compute shaders and the operation is tied to lighting. As a result, anti-aliasing is not available in DirectX 9 and you will need Dx10 or Dx11 to enable it. This isn't as large a disadvantage as it may first sound, since the game does not suffer from especially obvious aliasing to begin with. For our performance tests, we used anti-aliasing where available and we did not force it for Dx9 through the drivers, although this may be an option for end-users. We chose not to disable anti-aliasing in the performance benchmarks, since the point here is to compare the performance of the two rendering paths at their respective maximum image quality settings.

Aliens vs Predator does not have a built-in automated benchmark. Instead we manually ran through a level while recording the game's performance with the Fraps application's benchmarking function. The benchmark numbers presented below are the average frame rates recorded by Fraps during our benchmark runs. We chose the first level of the alien campaign as our benchmark.

Throughout the first level of the alien campaign are multiple enemies and rendered in-game cut-scenes. This created a very dynamic setting for the benchmark and each run invariably ended up slightly different from the next, despite our best efforts. In light of this, each benchmark setting shown below is the average of the results of 5 separate benchmark runs. Benchmark runs that resulted in strange values that did not correlate with the rest of the results were discarded and attempted a second time. This process, though painstaking, allowed us to generate relatively accurate and repeatable benchmark results.

Aliens vs Predator Performance Comparison
DirectX 9 and DirectX 11 Performance Compared
 






The results from our testing are presented above. The large performance delta between DirectX 9 and 11 is immediately apparent. No doubt a large share of the performance difference is due to the application of anti-aliasing. With anti-aliasing enabled in Dx11, the game benchmarked at nearly half the performance of the Dx9 rendering path. However it's worth noting that even despite the handicap largely created by enabling anti-aliasing, both video cards tested were able to produce playable numbers.

As we've previously mentioned, our primary goal here is to compare the performance of the two rendering paths at their respective maximum image quality settings. However, we did experiment with benchmarking DirectX 11 with anti-aliasing disabled. We found that, in all three resolutions, AvP's Dx11 rendering path actually performed slightly better than Dx9. For example, the Radeon 5770 at 1920x1200 performed 6.32 frames per second better in Dx11 than Dx9, when anti-aliasing was disabled.

AvP Dx11 Verdict: We noted in our image quality comparison on the previous page that Aliens vs Predator does gain a noticeable image quality improvement from using the Dx11 rendering path with all its bells and whistles. While the performance delta between Dx9 and Dx11 at their respective maximum settings is large, we must consider that many of the game's graphical bells aren't available in Dx9. It's also important to note that despite the precipitous performance drop, Dx11 performance remains playable with our mid-range card, the Radeon 5770. Overall, we would recommend enabling Dx11 in Aliens vs Predator if possible. The image quality gained is worth the considerable performance drop and the game remains playable in nearly all cases except possibly with budget and integrated graphics chips.
Transparent
BattleForge: Image Quality


Unlike most of the games we're looking at in this article, BattleForge is the only game from a completely new franchise. It's also the only strategy game out of the bunch. Though to be more accurate, we should say BattleForge is a card based real-time strategy game. If a collectible card game and a traditional RTS collided and mixed together, the result would be BattleForge, which takes elements from both genres.

The concept is actually simpler than it may sound, instead of training units like in traditional RTS', you play cards. Each card produces a unit or squad of units. The cards are collectible in the standard playing card style, you can purchase booster packs and trade/sell individual cards in the built-in marketplace. There are hundreds of cards in total and new editions which add new card sets and themes to the game are released periodically (currently 3rd edition), but you can only bring 20 cards with you into a given match. The rest of the gameplay is standard RTS fare and winning depends on the cards you bring with you into the match as well as how you use them in-game.

Most interesting of all, BattleForge is the only "free" DirectX 11 game currently available. BattleForge is an EA "Play4Free" title, which means the basic game is completely free to download and play. The catch is you will need to shell out cash in exchange for BattleForge Points which can be used to buy cards, which you'll need to do eventually since the free game only comes with 32 basic cards.



BattleForge wasn't released as a DirectX 11 title, this functionality was added in a patch. As a result, the image quality improvements offered to DirectX 11 users is minimal. Most of the improvements come in terms of performance as the video, which shows the built-in benchmark, clearly shows. The video shows the performance of DirectX 11 against the performance of DirectX 10.1 in BattleForge.

In contrast, we performed our image quality comparison between DirectX 11 and DirectX 9. This was done because DirectX 10 hardware is still able to run many DirectX 11 features in software and BattleForge, along with many of the other games featured in this article, use the same rendering path for both APIs. However DirectX 9 must use a different rendering path as the API is very different under the hood. In other words, there wouldn't have been any image quality difference between DirectX 11 and DirectX 10.

 

BattleForge takes advantage of the DirectCompute API and the game's screen space ambient occlusion (SSAO) effect is compute shader accelerated. The game may also make use of hardware tessellation, although we can't confirm if this is offered in the current version of the game since it may have been removed to improve performance as the effect is not noticeable from a typical RTS camera perspective.

In games, high levels of geometry tessellation is generally only used in a very limited scope, determined by the distance from the camera to the object. The effective result is that objects close to the camera will contain a higher level of geometry detail while objects further away will be less detailed. This works great since it's harder to notice geometry detail the smaller (further away) the object is on screen and despite hardware acceleration, tessellation is still a relatively expensive operation that would slow things down if it were used on the full scene.

   

Overall BattleForge is a good looking game, although the cartoony style and fantasy setting don't appear especially graphically impressive, technically speaking. It also doesn't help that the RTS camera, in general, tends to hide graphical detail. While we did notice some impressive shader effects, especially in the particle and water effects, they were barely noticeable in-game. This is one game were DirectX 9 users aren't really missing out. However it's worth noting that the game is still being expanded and new features and expansions are being added periodically so we may see more DirectX 11 graphical perks in the future.
Transparent
BattleForge: Performance & Verdict


BattleForge has the dubious honor of sporting the lowest recommended system requirements out of the games featured in this article. Requiring only a low-end Pentium 4 class processor or AMD equivalent (ie. Athlon 64) and 1GB of memory, the base specifications should be easy to meet. The graphics requirements, which ask for a NVIDIA GeForce 6000 series or ATI Radeon 9500 series, aren't much more intensive and can actually be met by the latest integrated chipsets from AMD and NVIDIA, although current Intel and older ATI and NVIDIA integrated chipsets may not be enough. This is perhaps the only game featured here that can be run on a typical non-gaming oriented laptop at respectable graphics settings.

It's worth noting that the official recommended requirements call for 10GB of hard drive space, but we found that the game fit into approximately 7GB. Your milage may vary.



Scanning through BattleForge's graphics settings, the only graphical setting not available in Dx9 is screen-space ambient occlusion (SSAO), which will hurt image quality a bit, though shouldn't be too obvious in-game due to the game's standard RTS camera perspective. Unlike some of the other games featured, BattleForge provides native anti-aliasing in all DirectX rendering paths, so we left it enabled for our performance tests in both Dx9 and Dx11.

BattleForge features a built-in benchmark function found in the graphics settings menu. The benchmark is a short 30 second run through of a scripted game sequence. The benchmark features a fly-through camera and plenty of continuous on-screen action. The benchmark provides solid, consistent results, though admittedly may not represent a perfect picture of real-world gameplay performance, though it is plenty sufficient for our purposes. The numbers presented below are the average of 5 benchmark results for each tested setting.

BattleForge Performance Comparison
DirectX 9 and DirectX 11 Performance Compared
 






In our look at BattleForge image quality performance on the previous page, we featured a video from AMD which showed BattleForge performing significantly better in Dx11 then Dx10.1. Unfortunately we did not see a similar performance delta between Dx9 and Dx11 in our testing. As is clear from our benchmarks above, Dx11 performed slightly worse than Dx9 in all three resolutions. We speculate that the performance drop is due to the addition of screen-space ambient occlusion (SSAO) in Dx11.

It's worth noting that despite having the lowest recommended system requirements of the lot, when all the settings are pushed to their maximum levels, BattleForge doesn't perform much better (in terms of raw numbers) than most of the other games featured in this article. The Radeon 5770 actually seems to have had some trouble in 1920x1200, posting borderline playable numbers in Dx11. Thankfully in-game performance disagrees with the numbers, and gameplay was still smooth despite the relatively low 23 average FPS. Overall, we would definitely call this one playable at all tested settings.

BattleForge Dx11 Verdict: As we noted in our image quality comparison on the previous page, BattleForge doesn't gain much from enabling DirectX 11. While Dx11 does enable screen-space ambient occlusion, perhaps the hot graphical feature of the last two years, the game's top-down RTS style camera tends to hide any gain in image quality. Combined with the note-worthy drop in performance compared to DirectX 9, we ultimately have to recommend that Dx9 users should stick to their guns for now, but perhaps not for long as there are rumors floating of further graphical updates just over the horizon.
Transparent
Bad Company 2: Image Quality


Hot on the heels of Modern Warfare 2 is DICE's direct competitor in the squad-based modern combat first-person shooter sub-genre, Bad Company 2. A direct sequel to the original Bad Company, BC2 offers a host of improvements and a graphical face lift. Most important to us for this article, BC2 adds DirectX 11 support.

Bad Company 2 is built on the Frostbite game engine, the same one used by the original Bad Company. Frostbite is DICE's own proprietary tech, and it was originally built for current-gen consoles. For Bad Company 2, the Frostbite engine was substantially overhauled to accommodate the special requirements of PCs. In addition to adding support for Dx11, DICE also integrated full support for ATI's EyeInfinity and NVIDIA's 3D Vision technologies.

As the video below illustrates, DICE also spent a lot of time tuning the PC version of the game to avoid that "console port" feel PC gamers have come to associate with bad ports. They have gotten rid of the 10-foot interface and claustrophobic field of view of the console version, which are great when the screen is far away from the player but inappropriate for PC use where the user is sitting much closer.



Just like with Aliens vs Predator, anti-aliasing is only available with DirectX 10 and 11 and isn't available with DirectX 9. You may still force FSAA through the drivers, though at significant cost to performance, and you'll still miss out on some of the image quality benefits that the in-game method provides. Just like with Aliens vs Predator, we chose to conduct our image quality comparison between DirectX 11 and DirectX 9, since the graphical differences between DirectX 10 and 11 are minimal.

   

With the graphics cranked up, the differences in image quality are subtle. One of the major graphical benefits of DirectX 11 is the addition of soft shadows. Another major graphical addition is the implementation of horizon-based ambient occlusion (HBAO), a method of screen-space ambient occlusion (SSAO) originally introduced by NVIDIA at Siggraph '08. This can be observed in nearly every scene but is perhaps most obvious when viewing the water effects and on the weapons in the second screenshot (above, right).

In DirectX 9, without HBAO, the light reflecting from the water doesn't effect the scene much, leaving many areas in deep shadow and dark. With HBAO enabled in DirectX 11 (also available in Dx10), the light reflected from the water is incorporated into the rest of the on-screen scene and areas like the rocks and the player's weapon are brighter as a result of the additional light.

   

Another area where the new graphical effects are readily observed is in the last screenshot (above, right), on the edge of the beanie cap worn by the soldier. In DirectX 9 the cap produces a sharp straight edge against the soldier's face, with no transition or shadow. In DirectX 11 the lip of the cap produces a slight shadow over the soldier's face which gives the appearance of added depth and enhances realism.

Overall, Battlefield: Bad Company 2 is a very good looking game with highly realistic graphics and good usage of DirectX 11 features. While Bad Company 2 definitely incorporates many Dx11 graphical features, like Aliens vs Predator, the ultimate result is still relatively subtle. This is especially true when things are in motion, as they would be in-game. When the virtual bullets are flying, it's not often that you can find the time to stop and notice the extra graphical enhancements.
Transparent
Bad Company 2: Performance & Verdict


Battlefield: Bad Company 2 features the highest recommended system requirements out of the games featured here. It's the only game to recommend a quad-core processor, though the minimum system requirements are much more forgiving. Two gigabytes of memory and 10GB of hard drive space is fairly standard and shouldn't be hard to meet. The recommended graphics requirements call for at least last-gen hardware like the NVIDIA GeForce GTS 250 or ATI Radeon 4850. This is pretty steep, though just like the processor requirements, the minimum system requirements are much more lenient.



Bad Company 2's graphical setting menu reveals that Dx9 users will be missing out on HBAO, a form of screen-space ambient occlusion, soft shadows and anti-aliasing. Just like Aliens vs Predator, Dx9 users will have to do without anti-aliasing unless they force FSAA through their drivers. Bad Company 2 is one of the few games which allows the user to force a Dx10 rendering path on Dx11 hardware, so we have included Dx10 results along with the usual Dx9 and Dx11 numbers.

Bad Company 2 does not have a benchmarking feature so all our performance testing was conducted using manual benchmark runs. For each benchmark setting tested, we played through the entire first level of the single-player campaign 5 times while recording the average frames per second performance using Frap's benchmarking feature. The results presented below are an average of all 5 benchmark runs for each setting.

Battlefield: Bad Company 2 Performance Comparison
DirectX 9 and DirectX 11 Performance Compared
 






The most prominent feature of the set of benchmark results above is the extreme performance delta between Dx9 and the other rendering paths. In all three resolutions, with both video cards, we see that Dx9 performed nearly twice as well as Dx10 and Dx11. However, note that Dx9 is missing multiple image quality features which are not available and cannot be enabled. Out of the missing features, the most significant to performance is likely the absence of anti-aliasing. Screen-space ambient occlusion is also missing, which accounts for some of the performance delta as well.

Without anti-aliasing and HBAO, Dx9 suffers in terms of image quality but at a tremendous performance gain. Despite the performance drop, performance with the Radeon 5770 remains acceptable at very playable levels, even at 1920x1200. It's interesting to note that Dx10 and Dx11 performance are very similar. This is likely because most of the Dx11 features are also available in the Dx10 rendering path, though without the assistance of hardware acceleration.

Bad Company 2 Dx11 Verdict: Get it. While the game suffers from a massive performance drop with all of the DirectX 10/11 exclusive image quality settings cranked up, the overall performance remains very playable. If you're stuck with Dx9 in WinXP, you're not missing out too much since the image quality difference is hardly night-and-day, but it is definitely there and noticeable. Considering the game remains very playable in Dx11 with all the goodies enabled, we'd say it's worth it.
Transparent
Dirt 2: Image Quality


One of first DirectX 11 games to be released in North America, Dirt 2 also boasts one of the largest compliments of DirectX 11 features. The sequel to the original Dirt, Dirt 2 has already won high critical acclaim for its solid gameplay and excellent graphics. While some have commented that the game is less simulator and more arcade than its predecessor, most agree that the gameplay is fun and captures the enjoyment of rally racing well.

Dirt 2 is built on Codemaster's Ego Game Technology Engine, a modified version of the Neon game engine which powered the original Dirt. Unlike many of the other games featured in this article, Dirt 2's anti-aliasing is implemented in a way which allows AA in all DirectX levels. The game makes use of a large number of DirectX 11 features, as the following video illustrates.



Dirt 2 makes extensive use of tessellation, most notably in the cloth objects, like flags and banners, and in the spectator crowds. Unfortunately, both of these items play very small rolls in the game and the player is often only treated to brief glances of flags and crowd in their struggle for rally domination. We found that during normal gameplay, cloth objects like flags and banners, as well as the spectator crowd, were too far away from the camera for the benefits of tessellation to be noticeable. We actually had a hard time capturing a image quality difference in screenshots through normal gameplay, so it is not represented below in the comparison images.

   

Besides tessellation, Dirt 2 also implements a few other DirectX 11 goodies. Two other notable DirectX 11 features implemented by Dirt 2 are screen-space ambient occlusion (SSAO) and some post-processing filters. Dirt 2 uses high definition ambient occlusion (HDAO) which is AMD's implementation of SSAO. These effects can be observed on all manner of objects. They add extra depth to flat surfaces and their effects can be observed on the car bodies. In the first comparison image (above, left), the headlights of the car in DirectX 9 are relatively bright, despite being recessed in the body of the car. In DirectX 11, the HDAO algorithm correctly depicts the headlights as being in shadow, due to being blocked by the hood of the car.

These effects can also be observed in the second comparison image (above, right), where the body of the car reflects differently in DirectX 9 compared to 11. Note how the bottom of the "wing" on the rear of the car correctly reflects the ambient light reflecting from the top of the trunk in DirectX 11. Also note the lack of reflections on the car's rear window in DirectX 9. Finally, Dirt 2 performs some post-processing, and this can be observed in the shadow transitions, which are smooth in DirectX 11 but appear to be composed of discrete overlapping layers in DirectX 9.

   

While the implemented crowd and cloth tessellation was barely worth noting, we definitely did notice the tessellated water effects. The tracks were frequently covered by small pools of water which the cars were more than happy to splash through. The result is a tessellated wave which actually rose up and rippled. In contrast, in DirectX 9, the water remained relatively flat due to the lower level of geometry detail without tessellation. While this effect is best viewed in motion, like in the video, it can be observed in our last comparison image (above, right).

Overall, Dirt 2 is a very visually impressive game. The near photo-realistic graphics really lent a strong atmosphere to the game which helped to impart a real sense of "being there" to the player. While many of the tessellation effects won't be missed by DirectX 9 users, the lack of screen-space ambient occlusion and post processing will bring the image quality down a notch.
Transparent
Dirt 2: Performance & Verdict


Like Aliens vs Predator, Dirt 2 has relatively demanding recommended system requirements, although not quite at the levels that Bad Company 2 recommends. According to Codemasters, the game is best run on at least a mid-range Core 2 or equivalent Athlon X2 processor. Two gigabytes of memory and 11GB of hard drive space is typical. At least on paper, the game has similar graphical demands as Aliens vs Predator, asking for a NVIDIA GeForce 8800 series or ATI Radeon X1950 series card.



The primary features that DirectX 11 users will be getting are tessellation, screen-space ambient occlusion and Dx11 exclusive post-processing. However, anti-aliasing is available to all DirectX rendering paths. We chose to test the game with anti-aliasing set to 4x for both Dx11 and Dx9. All other settings were set to their highest available levels.

Dirt 2 features a built-in benchmark function hidden at the bottom of the graphics menu. This option is unavailable unless you create and log into a campaign account. It's one of the better built-in benchmarks we've seen in a game since it replicates the real-world gaming experience very well and therefore generates results in-line with real-world gaming performance. The benchmark itself is basically a short race on a track from the game. This isn't your typical scripted benchmark however, since all cars are AI controlled. It seems like they simply took a typical race from the game, ejected the player and replaced him/her with AI. You get to watch the action during the benchmark from the perspective of your AI replacement from the game's standard game camera.

Overall Dirt 2's built-in benchmark produces consistent results that are a good representation of real-world game performance. We used it for our performance tests. All of the results below are the average frame rates produced by the game's built-in benchmark. Each tested setting was benchmarked 5 times and the results were averaged.

Dirt 2 Performance Comparison
DirectX 9 and DirectX 11 Performance Compared
 






There is an unmistakable performance delta between DirectX 9 and 11, although it is much smaller than the delta seen with some of the other games. This time around, the performance difference has nothing to do with anti-aliasing since it is available and enabled for both rendering paths. The game's use of tessellation, ambient occlusion and DirectX level-specific post processing effects are to account for the performance drop in Dx11.

Despite having relatively high recommended system requirements, Dirt 2 performed very well at all tested settings. While the 20-30 FPS delta is hardly negligible, the game's overall performance remains good across our tested resolutions and on both video cards. With every image quality setting pushed up to maximum levels, the Radeon 5770 had no trouble cranking out a respectable and very playable 40 frames per second (average).

Dirt 2 Dx11 Verdict: This is one of the games where you will definitely want to enable DirectX 11 rendering features when available. The performance drop hardly impacts gameplay and the game is smooth even with mid-range cards. While you may be able to salvage some extra frames per second with DirectX 9, it's not worth the image quality drop and you probably won't notice the extra performance in-game except with budget graphics chips.
Transparent
Unigine: Image Quality


The last game we'll be looking at in this article isn't really a game at all, but a benchmark and demo for a game engine. The game engine in question is called Unigine and it's being developed by Unigine Corp. for licensing to other developers. They are also developing their own game using the engine. Unigine developed from the Frustum open source project which has made a few showings at Siggraph over the years. The engine is currently being licensed to over two dozen clients, over a dozen of which are using it for game development. While no currently available retail titles make use of Unigine, several upcoming titles will be built on it including Afterfall Universe and Primal Carnage.

Despite not having any shipping titles under its belt, Unigine has kept a relatively high profile by releasing a string of compelling benchmark demos, the latest of which, called Heaven, shows off the latest iteration of the engine which offers full DirectX 11 support. Unigine's Heaven benchmark features one of the most extreme and noticeable implementation of tessellation out of the games featured in this article, as demonstrated by the video below.



The Heaven benchmark features an extreme level of reliance on tessellation to achieve its image quality. This reliance on tessellation can be noticed nearly everywhere, but is most noticeable on the stairs, cobblestone roads and roof shingles. The geometry of the stairs is entirely dependent on tessellation and actually appear as a flat ramp in DirectX 9 when tessellation is unavailable. While the extreme reliance on tessellation would be unacceptable in a retail title since it significantly decreases the image quality for DirectX 9 users, Heaven is a benchmark and tech demo, and as a result we finally have a vivid example of what tessellation can do for the graphics in a game.

While our comparison images are with DirectX 9 and 11, it's worth noting that tessellation is also unavailable in DirectX 10. Overall, Unigine Heaven's DirectX 10 mode looks roughly similar to its DirectX 9 mode, in terms of overall image quality.

   

Other areas where tessellation plays a key role in providing image quality are the stone blocks which frame the bridge in the second to last comparison image (below, left) and the cobblestone roads in the last image (below, right). Without tessellation, the DirectX 9 rendering makes the bridge and road appear relatively smooth while the tessellated DirectX 11 rendering is extremely bumpy with significantly elevated geometry detail.

Tessellation certainly isn't the only DirectX 11 feature implemented by Unigine, it also claims to implement screen-space ambient occlusion (SSAO) and makes use of the DirectCompute API and the new Shader Model 5.0 features like compute shaders. However, the only DirectX 11 feature we really noticed in the Heaven benchmark is the tessellation effect. If SSAO and the other features are implemented in Heaven, we didn't notice them.

   

It's worth noting that we performed our image quality comparison with the first version of the Heaven benchmark, since we performed our Unigine Heaven image quality analysis a while back. But before this article was completed, a second version of Heaven was released. Heaven 2.0 extends the first edition with the addition of a new airship area to the map, which may implement some of the features we found lacking in the first edition of Heaven. You can view a video of Heaven 2.0 here, or (especially if you have DirectX 11 hardware) head over to the Unigine website and try the benchmark on your own system.

Overall, the Unigine Heaven benchmark is a very good looking tech demo. It certainly makes the Unigine engine look very impressive and we can't wait for retail titles built on Unigine to start appearing. Unigine also offers us the first truely vivid example of tessellation in action, but many of the other DirectX 11 image quality features, many of which are implemented by the other games featured in this article, are oddly missing from the benchmark. Or perhaps they are so subtle that they were lost among the extreme use of tessellation.
Transparent
Unigine: Performance & Verdict


Unlike the rest of the games featured in this article, Unigine Heaven isn't actually a game, but a tech demo for the Unigine game engine. As a result, the recommended system requirements should be taken with a grain of salt. Since it's not a full-fledged game and contains no gameplay elements like scripting or AI, Unigine Heaven will have much lower processor and memory requirements than any game built on the Unigine engine, like the upcoming  games Primal Carnage and Syndicates of Arkon. While the processor and memory requirements are low due to the lack of game elements, Unigine Heaven is a heavy-weight graphics benchmark that calls for relatively recent hardware, like the NVIDIA GeForce 7000 series and ATI Radeon HD 2000 series, at a bare minimum.



The Unigine Heaven benchmark is built primarily to showcase the engine's DirectX 11 implementation. As a result there are a considerable number of features that are exclusive to the Dx11 rendering path. The most noticeable feature which is Dx11 exclusive is the extensive use of hardware tessellation. While anti-aliasing is available for all rendering paths, we chose to leave it disabled since the Heaven demo is extremely graphically demanding, on purpose.

Since the entire program is one big tech demo and benchmark, we used the built-in benchmarking feature for all of our performance tests. The benchmark is a fly-through of Unigine Heaven's single level. There is no AI or even characters to be found in Heaven, only landscape and buildings. Since it was designed as a benchmark, it produces consistent results that suited our purposes just fine.

Unigine Heaven is the only game/program featured in this article that also has a OpenGL rendering path. Since it is designed as a tech demo and benchmark, we chose to benchmark all available rendering paths. All results shown below are the average of 3 benchmark runs.

Lastly, it's worth mentioning that we performed our tests using the first version of the Unigine Heaven demo. Since we performed our testing but before this article could be published Unigine released a second version of the Heaven demo which features a whole new section to the demo's single level and a different benchmark fly-through path. Check out Heaven v2.0 for some extra eye candy, but beware that it will produce benchmark results incompatible with the ones presented in this article.

Unigine Heaven v1.0 Performance Comparison
DirectX 9 and DirectX 11 Performance Compared
 






It's worth noting that as a tech demo, Unigine Heaven makes very few performance concessions and very little is done to ensure image quality scales consistently, compared to a production game. Its primary purpose is to show off what the Unigine game engine is capable of. As a result it has the lowest overall performance out of the games featured in this article.

The largest performance delta is between Dx11 and the rest of the pack. This is largely due to the application of tessellation, which is used extensively by the Heaven demo, as we saw in our image quality comparison. The performance of DirectX 9, 10 and OpenGL were relatively similar by comparison. However the DirectX 9 rendering path performed best overall, followed by DirectX 10 and finally OpenGL. It's worth noting that during our tests, the OpenGL rendering path experienced buggy behavior with frequent artifacting, which may have been fixed in Heaven v2.0.

Unigine Heaven Dx11 Verdict: Since this is merely a tech demo and not a production game, it's hard to make conclusions about its graphical performance. While it's plain to see that Unigine Heaven benefits extensively from Dx11 in terms of image quality and the associated performance drop isn't especially bad, it's impossible to say if we will see similar results in production games built on the Unigine engine. After all, Heaven is Unigine's Dx11 tech demo and is built with the purpose of showing off the engine's Dx11 features. Overall, if you want to check out all of the demo's bells and whistles, you'll definitely want to run it with the Dx11 rendering path.
Transparent
The State of DX11
A few years ago, we discussed the state of DirectX 10 and came back unimpressed. Not only did the available games of the time perform poorly under DirectX 10 on available hardware, image quality gains were nearly non-existent. Switching from DirectX 9 to DirectX 10 dropped performance by half with almost no discernible image quality improvement. Most of the DirectX 10 image quality perks we spotted in the games we checked out were due to specific developer decisions and had little to do with the new API. Now, nearly three years later, we're sorting through yet another mountain of data for yet another new DirectX API iteration, the 11th in the series. Thankfully things are looking up this time around.



We've tested five of the earliest available DirectX 11 titles and we're happy to report that the situation this time around is looking quite favorable for early adopters. While we observed performance hits when switching to DirectX 11 from DirectX 9 in all five games, we also observed a noticeable corresponding image quality improvement. The drop in performance for DirectX 11 in our tests can be attributed to the added image quality. Effects like tessellation, screen-space ambient occlusion, advanced post processing and DX11 exclusive anti-aliasing in some games brought the performance down in our testing, but they also boosted image quality.

What's more, it would seem that some of the games actually performed noticeably better in DirectX 11, compared to DirectX 10 and 9, when the extra image quality features exclusive to DirectX 11 were not enabled, as we observed in Alien vs. Predator. This suggests that DirectX 11 rendering paths can be faster than their DirectX 9 counterparts while maintaining the same image quality level in some scenarios.

In terms of image quality, the new effects and graphical tricks we saw with DX11 added some extra realism not available in DirectX 9. Effects like screen-space ambient occlusion, depth of field, and tessellation really brought the graphics up a notch, adding extra atmosphere and realism to games like Aliens vs Predator, Bad Company 2 and Dirt 2. However, these effects were generally quite subtle and are nothing as dramatic as the image quality changes brought on by DirectX 8 or 9 in the past, which both offered major milestones in shader technology. It's also worth noting that, except for tessellation, these effects are all technically possible without DirectX 11 and compute shaders, as proven by Crysis.

DirectX 11's hardware tessellation and DirectCompute API have great potential, but currently they're being used to implement effects we've seen before since Crysis was introduced. While that certainly doesn't sound very impressive, keep in mind that although the graphical effects are nothing new, DirectCompute now makes them much more efficient. So efficient, that all five games were able to run at full HD resolution (1920x1200) with all the graphics setting cranked to max without much trouble, even on the affordable $150 Radeon 5770. This means that game developers will be much more willing and likely to implement these effects in their titles, so watch for them in the coming months. Overall, performance in currently available DirectX 11 games is excellent.

Is DirectX 11 Worth Your Time?
Ultimately DirectX 10 will be remembered as a stepping stone. It was a necessary evil that helped rid the DirectX API of over a decade of legacy dead weight. We predict it will be overshadowed in the memories of most gamers by DirectX 11 which looks like it will finally make much of the hype originally created for DirectX 10 a reality. Currently available DirectX 11 titles and hardware really delivers, both in terms of performance and image quality, though arguably not quite at the same time. If you've held out on Vista and DirectX 10, and have been waiting to see if something better comes along before leaving your beloved DirectX 9 WinXP gaming platform behind... well, it appears to be here and it's called DirectX 11.


Content Property of HotHardware.com