Deus Ex: Human Revolution - A user review (Part II)


No ROM battery
This post continues from Deus Ex: Human Revolution - A user review (Part I)

The Game Art

One of the things that I found most interesting about DE:HR technology was its economical approach to DX11. Particularly with tessellation. The game makes use of it if the player has a DX11 compatible card and enables this technology in the game settings panel (you have to scroll down on the advanced video panel to find it, which may not be readily evident at first). Tessellation in DE:HR is being used just on character meshes and certain objects (according to the developers, although I haven't searched for evidence of tessellation on objects myself), while the actual level of tessellation applied is quite modest, done mostly to accentuate or fix curved lines. It's this economical approach that allows the game to run with tessellation enabled without any great loss of performance. Is this a good thing? Ask me by the end of next year or in 2013 and I will probably say no. But as it is, today, this is probably a wise decision. Tessellation can have a huge impact on visuals if fully applied. On the other hand it can bring a modern video card to its knees. By being modest about it, the game can be played by DX9/DX10 capable video cards without players feeling they are missing out much on art detail, thus broadening the game reach. Likewise, only the top end of cards today can handle heavily tessellated scenes with a modicum level of performance. Turning DE:HR into a "Can it run DE:HR?" meme not only doesn't sound so good as the former one, but would also probably irritate its fanbase who have been waiting for the past 11 years for this game. It would certainly irritate at least me.

Below are two screenshots showing some differences between tessellation enabled and disabled (make sure you click the image to show the larger version). You can see how lines are more curved, especially around Jensen's right shoulder and in his collar. Unnoticed in these images, but noticeable if you make an animated gif out of them both, is that the pads on Jensen's chest also gained some volume. The ability to fine tune the mesh (which is mostly what tessellation is all about) allows for volumetric improvements too, making rendered objects truer to their 3D nature in the hands of skilled artists. I can't wait for this technology to really takeoff and to forever put behind us nearly 20 years of polygonal meshes.


DE:HR is also the first(?) game to employ MLAA after AMD finally released it to developers and allowed them to employ this anti-aliasing technique regardless of graphics card vendor. Nvidia had already released their own FXAA earlier this year. We were just missing out on AMD's technique; Not anymore. At this point, however, it seems anti-aliasing options have become more confusing than ever. So here's a quick guide before we proceed with the review proper:

MSAA technique (you know, that's those 2x MSAA, 4x MSAA, etc) is probably becoming a thing of the past. FXAA and MLAA are their "replacements". Quoted because it's not really so. Both techniques can in fact -- and should sometimes? -- be used by developers in conjunction with MSAA. MSAA can handle subpixel features better, while FXAA or MLAA will deal with better edge gradients. But if that is the case, MSAA can go incognito to users and they won't be able to change it because, when employing both techniques, developers will gain more if they control MSAA with a fixed value. So it's possible that MSAA is going away, replaced by MLAA and FXAA. Some games will keep making use of MSAA under the hood while employing any of these two techniques.

Deus Ex: Human Revolution allows us to change between MLAA and 3 levels of FXAA. For all purposes and intents, we should only concern ourselves with the highest level of FXAA. The performance impact between the three levels is minimum and should not matter to anyone except those sporting borderline graphics cards that may throw their average FPS below 30 if choosing a higher setting. These should be however only on older entry-level cards. The interesting thing about both MLAA and FXAA is that they get generally better results with less requirements. The differences between both techniques are however frustratingly shallow and conflicting, meaning both are very alike, and where they differ means the user will have to sacrifice one thing for another. It's just a matter of fact that on the same game, MLAA will look better on some scenes, while FXAA will look better on others.

MLAA is the older technique, created by AMD to combat the notorious performance impact of MSAA and SSAA anti-aliasing techniques. Nvidia's FXAA came afterwards as a competitive answer to MLAA by introducing some manner of subpixel analysis (which MLAA lacks). Both techniques differ from earlier technologies by applying their anti-aliasing algorithms not during the rendering phase of the GPU pipeline, but as a post-processing technique (MLAA also makes use of direct computing and stream processing). This greatly reduces the cost anti-aliasing has on the whole scene rendering process, because it should be less computationally expensive and memory demanding to work on the final rendering. The results are stunning; Regardless of which one we choose, the impact on FPS is a fraction of what it used to be and the image quality only falls short to the king-of-the-hill family of anti-aliasing algorithms known as SSAA. MLAA and FXAA differ somewhat. But unless one is carefully looking for it, they won't notice these differences. It is just on the treatment both techniques give to subpixel rendering that there's a noticeable difference on some scenes. MLAA makes no use of it and FXAA tries to fill this gap by employing a modest subpixel analysis algorithm. For this reason, MLAA tends to better protect texture fidelity, whereas FXAA tends to lose on texture contrast. Conversely, MLAA can become too artificial. Its lack of sup-pixel analysis does tend to become visibly distracting on high contrast scenes, whereas FXAA can improve overall image quality by giving it a more fluid feeling. Where it will work and whether it will work for you, it will depend entirely on the scene being rendered and what your preferences are.

It's hard to say you will ever like one over the other on every game that comes with both. Also, needless to say that both techniques are still largely under development and have lots of room for improvement. It's a well known fact that anti-aliasing is one of the biggest unresolved issues of modern computer graphics. No technique we have so far, and that can run on our "modest" computers, is a great technique. Merely bearable.

Bellow are two screenshots that shows MLAA and FXAA (again, don't forget to click the image). Notice how FXAA seems to do a better job at eliminating the "jaggies" in the bottom portion of the thin black rim running across the length of the image (this is subpixel analysis at work). But then notice how the background area with the papers fixed to the wall loses some contrast and fidelity in comparison to MLAA. On some scenes this loss of contrast will be greater -- and particularly noticeable on scenes with less light -- and may not compensate the gain in anti-aliasing. If you look at the black thin line again, you'll notice that the top portion is basically the same for both MLAA and FXAA. So on scenes where subpixel rendering isn't introducing any noticeable advantage, the loss of contrast may make FXAA the less desirable option. But on the other hand, this loss of contrast can become instrumental to give the image a more realistic look. It's because of this that it's hard or impossible to say one is better than the other. Unless a developer poorly implements one of them and does a good job at implementing the other, I say close your eyes and take your pick.


The final matter of business in this game art section I feel like discussing is character design and acting.

It's a well known fact by now that this isn't the prettiest aspect of the game. In fact, it is the ugliest. Generally speaking characters look like they never went past the stage of being a prototype. Some faces even become hard to forget, for the worst reasons of course. Like the lady hostage you rescue (or not, depending on your dialog choices) early in the game. I could probably do better by just randomly throwing lumps of plasticine onto a log. Another character that got under my skin was Hugh Darrow. Generally speaking I am tempted to characterize DE:HR player characters as dangerously close to the uncanny valley of things. And it doesn't help one bit that those characters you can speak with, and that cut to a dialog scene, have an old school skeletal animation that is not only exaggerated in DE:HR, making most movements look almost like stop-motion, but also rather limited in scope. The palette of gestures available to them is short and invariably makes all characters behave exactly the same. A game that I remember doing this pretty well was Dragon Age: Origins. Dialog scenes had some of the most realistic character motions like nearly imperceptible head and shoulder movements that would greatly contribute to staying immersed in the game. The number of available gestures could remain small because they wouldn't significantly impact on the player visually and thus go largely unnoticed, while still subconsciously impacting the player, which is how we "read" gestures in real life.

Adam Jensen himself isn't the cutest kid on the block (in a way, "about damn time!"). There's something definitely wrong with his chin, for instance. On some angles it looks like the Wicked Witch with a grown goatee at the end of her shovel-like chin. The funny bit though is that at the end of the day I ended up feeling quite charmed by all this coarseness. I can't quite explain it. Just that it didn't affect me negatively as it normally would. Maybe it was just a matter of my subconscious protecting me against a game aspect that could turn an otherwise great experience into a bad one. But I think there's maybe more to it. I think that the characters end up feeling right at home in their dystopian world. It seems fitting that the characters share some of characteristics of the environment they inhabit. In the end, despite all that can be said about the quality of the character art, I feel two things: The first is that there is no doubt they could do better than this. They had the means, and the artists involved, all the necessarily skills and talent. So I like to think there was intention here. The second thing is that they just end up working. Maybe a throwback to older games, maybe not. But they just work and this didn't seem to have become an issue to most players who gladly kept putting the game on the first spot of sales reports.


But I can't say the same for the voice acting. Steve Shellen completely stole the show for everyone else. His voice acting of David Sarif makes this character the most memorable and lauded of the entire game. Way ahead of anyone else, including Jensen himself. And here's the thing, Elias Toufexis is no worse an actor. We are talking about the guy that played Andriy Kobin in Splinter Cell: Conviction. A trained stage actor too. And we can see his qualities in Adam Jensen's in many of his speech inflections, for instance, and the ability he has to convey minute emotional details. But because the video gaming industry can't get their head off their buttocks and still treats gamers as immature infants incapable of handling anything a notch higher than Dolph Lundgren, protagonist player characters have to have deep voices that could dry a well. For this reason, Adam Jensen gets lost in so much "cuteness" and lack of credibility, while David Sarif catapults itself to the top. Other voice actors contributed greatly to the game too. But there's one particular actress -- which I won't name -- that I would like to never again see lending her voice to a game, unless it is something out of a cartoon. With all due respect -- and I'm sure she's a fine person otherwise -- it was just horrible, horrible, acting on the two characters I remember her lending her voice.

This is the one aspect of the game that disappointed me. Unfortunately (for me) I'm way more caring of voice acting that I ever will be of graphics quality. So I wasn't exactly a happy camper through the whole experience and some portions of the game added to a loss of suspension of disbelief due to the poor voice acting. My own player character wasn't someone I was pleased to hear. Not my hero archetype. At all. But hey, it can be said that on this aspect DE:HR pays tribute to the original game! Deus Ex wasn't exactly complimented on its voice acting either. On the contrary.


Didn't realize this second part would be so long and the forums didn't allow me to fit everything in one post. So I had to divide this into three parts, from the initially planned two. The third and final part is coming up in a few minutes.
Last edited: