So here we are, with one of the biggest releases of this year. Crysis 2 finally comes storming out through the gates after a myriad of techinical demonstrations and effects showcases designed to big up the CryEngine 3 to the gaming fathfull. Crytek are masters of producing high-end visuals that require high-end hardware to run. But what about designing the same cutting-edge content to run on what can only be considered five year old, low-end tech?
Well, that is exactly what we're here to find out as we lay out a triple platform tech analysis of the developers latest visual spectacle, Crysis 2. First we begin with the consoles, before moving onto our more direct PC comparison.
Crytek have made it no secret that their CryEngine 3 technology has been made in such a way as to scale between different platforms, each with varying specs while keeping the core components (GI lighting, advanced shader effects, real-time shadows etc) intact. Instead, compromises have been made in other areas, from shadow quality, resolution, LOD, right down to how perameters for each of these components operate.
While PC owners will eventually get the full, untapped potential of the engine (The game only supports Direct X9 at launch), console users on the other hand get a scaled back revision that impressively implements some of the high-end features found only in the upper settings of the computer version. In that respect they get a nicely balanced blend of compromised image quality at the expense of some loverly GI-based lighting, god rays, and other cool touches.
But how does each one fair? Lets get on with it...
As always we start off by taking a look at the framebuffer of both console versions and see how they hold up. While PC owners obviously get a choice of native rendering resolution - regardless of actual hardware specs - on consoles it isn't quite so simple. The framebuffer is restricted by both available processing power and memory bandwidth; both of which are a limited commodity on consoles compared to the constantly shifting nature of PC hardware.
Crysis 2 renders in 1152x720 on the Xbox 360 and 1024x720 on the PS3, with both versions getting the same use of temporal 2xMSAA (multi-sampling anti-aliasing). And as you can see in the opening screenshots above, the two games aren't far off of each other with regards to image quality. The 360 game is a tad sharper owning to less horizintal upscaling taking place, but at the same time the difference can often be barely noticeably in motion, and neither build features the clearer IQ of a true 720p game.
Interestingly, the HUD elements in the PS3 game appear to be upscaled rather than rendered over the final framebuffer. Quite why this is the case isn't exactly known for sure. But, what we do know is that the RSX GPU provides extremely low cost horizontal scaling, and in order to render the HUD after the FB they'd need more memory to do so.
Regardless both versions still look good however, and It seems that edge post-process effects along with the game's use of anti-aliasing attributes somewhat to its soft look. More so on the PS3 by the looks of things due to the additional upscale taking place, but in practice the difference appears less pronounced than in still shots, and thus less impactful in general.
Crytek's anti-aliasing solution - a temporal form of 2xMSAA which appears to be selective in its implementation on surfaces throughout the scene - has little impact in terms of providing high-level amounts of edge smoothing in highly detailed outdoor scenes. But its effects seem to be variable, with some areas - particuarly inside - faring better than others. Crytek's AA solution also works on various parts of the scene a little differently, using depth buffer info and edge detect on close objects.
Also, as a result of a frame blending technique used to create the AA samples from two seperate frames, we see that a ghosting effect is present during movement, like in the demo. Additionally, it appears that the use of frame-blending, along with edge-post process effects tend to blur the image somewhat on all versions of the game.
The killer point about Crysis 2 on both consoles and on the PC, is the inclusion of Crytek's much talked about single-bounce global illumination (GI) lighting system, whereby sun lighting features a singular, real-time bounce that for the most part accurately resembles real-world light occlusion.
This also means that all shadows and light sources are rendered in real-time through the game. The effects of which are outstanding as a whole. Ambient light and shadows are cast, while the main light-bounce creates a depthy, atmospheric look to the proceedings. Along with this we get the usual lens flaire and bloom effects, plus the addition of real-time, 'proper' sun-shafts too. All of which are equally represented on both PS3 and 360. Bar, except for some mild additional light occlusion in places on both the PC and PS3 codes, which usually darken the scene but seem to add a mildly stronger light bounce in places.
Crysis 2 uses deffered rendering in the form of differed lighting passes in order to deliver many dynamic lights onscreen at a lower overall cost than incurred by traditional forward rendering techniques. It is also an easier and more convenient way for artists to light every scene - they don't have to wait for hours of pre-computation in order to see the end results.
However, use of GI also comes with additional costs - namely memory bandwidth and computational power. Real-time shadows and occlusion means more alpha on screen, while having to calculate the lighting bounce on the fly means more processing power needed per-pixel. As a result we have already seen a reduction in the resolution of the framebuffer on each version, but there are many other parts of the game - visually - that have seen soem compromises in order to accomodate what is arguably the most impressive use of lighting in any console game to date.
By far the most obvious of these is Crysis 2's use of LOD and texture/object steaming. While there is a slight sense that LOD has seen an improvement over the 360 multiplayer demo, we still see many objects that pop-in noticeably as you approcah them. For example, foliage transitions between low and high quality assets fairly close to the player, buildings and other geometric objects too are also affected. Often this can look more than a tad unsightly, but sometimes can also go by without much notice - it all depends on just what the engine is rendering at any given time.
In our demo analysis we also mentioned in closing that a pre-release config file mentioned LOD perameters that were similar, or perhaps even identical across both PS3 and 360. Usually, it is the PS3 code that comes off worse in this area, with less memory and bandwidth to accomodate the same levels of draw distance and LOD update as the 360 in most cases. But in the finished retail game we find that not to be the case - bar one or two oddities that only ocasionally stand out in certain circumstances.
'High' settings
The most noticeable difference is how shadows are handled on both formats. Shadow LOD appears to be a tad stronger on the PS3, with some elements either being rendered in much later, or not at all. Self-shadows too also suffer from the same problem.
Additionally, SSAO initially seems to be cut back on in the PS3 game. Notice in the above shot, by the concrete blocks on the ground, that the effect is present on the 360 but fails to load in at all on the PS3. However, this isn't actually the case at all - SSAO in the PS3 game is in fact a closer match to the PC game.
The PS3 code also benefits from some advantages in other areas. First up, and as you can see below, shadows are filtered using a higher quality implementation compared to on the 360.
'High' settings
And secondly, texture filtering has also been given a significant boost, with bias towards certain surfaces giving ground and enviromental textures a cleaner look compared to on the 360.
According to Crytek, both games use dynamic AF (anisotropic filtering), but in the PS3 version we can see what looks like between 2x-8x filtering compared with much lower amounts in many places on the 360 - what looks like about 4x max, from what we can see. Officially, Crytek say that the PS3 game can switch between using 2x and up to 16x levels of AF.
While these differences are perhaps minor in nature, there's no question that the PS3 game's use of AF and higher quality shadow filtering makes a small tangible differnce. Of course, the lower framebuffer resolution partially cancels out the AF - owing to a blurrier image in general - but there's also a real sense that the two versions are remarkably close to each other given the immense task of rendering a hugely detailed environment, and then lighting it all up in real time.
Far more important however, is how they both perform whilst delivering such intricate visual complexity. And in this respect neither are particulaly excellent, with plenty of impactful frame-rate drops reducing controller response down to unacceptable levels. But this is only half of the story. Performance in Crysis 2 is heavily bound by load, with the 360 version leading in a general sense, but with the PS3 one occasionally doing the same in some chaotic scenarios.
Crytek are targeting a 30fps update for the console releases of Crysis 2, and going by the 360 multiplayer demo code at least, they were doing a rather good job of maintaining it, with very little in the way of frame-rate drops and no screen tearing. However, the single-layer mode is vastly different; there's much, much more going on. And all of this has a might impact on how well the engine can cope as a result.
While the results aren't pretty; both versions are at times very close. It's not always the case that one has a distinct advantage over the other, and both have various ups and downs with regards to keeping that 30fps but in different scenarios. Although, it's often the PS3 version that falls short more than the 360, but not always so.
On both platforms Crysis 2 run with an uncapped frame-rate and with v-sync enabled. However neither version hits the targeted 30fps mark for very long and barely goes above this. In fact, they both constantly fail to reach it in many situations, with or without heavy load. In addition, there is a small difference in how both versions deal with holding v-sync: the PS3 maintains this it throughout, while the 360 game seems to drop it very briefly, perhaps in order to maintain a slightly higher refresh. Tearing however, only occupies the very top of the screen and isn't noticeable during play.
As the video suggests, performance in Crysis 2 can be hugely variable; one minute the game is plowing along smoothly at the desired screen refresh, and then, in just the blink of an eye will drop right down to the mid to low 20's for extended periods of time. The main cause seems to be combat situations. Although, not neccesarily heavy combat - slowdown occurs when small, one or two man firefights break out, and increases dramatically when more starts going on.
Thankfully, somehow Crytek have still made the game playable regardless - in practice due more down to having to work your way around situations rather than going through the whole game and treating it as one giant shooting gallery. But seeing as playing it any other way than with ample thought and especially, a good bit of sneaking/enemy avoidance, makes the challenge just a touch too much to bare, so this negates the issue somewhat.
In any case, the fact that the game still throws everything but the kitchen sink into the console code, means that having it run at all in a playable state - even with the above listed shortcomings - is a pretty mean feat regardless. That said, one could easily argue that performance should have been better, with more careful optimisations, and maybe a few more scant cut backs here and there in order to better achieve that 30fps target.
Despite a few shortcomings, including a loss in IQ, some obvious LOD and streaming issues, plus a decidedly unwheldy frame-rate, Crysis 2 is a technical marvel on consoles. Say what you will about those comprosmises, but the fact that the game provides both scale and detail whilst delivering a fully real-time, shadowing and GI lighting system is it self an incredible feat. The game looks abosutely stunning!
Of course, there are smoother games out there, ones which perhaps excel in other particular areas. But on the whole, Crytek's latest shows that in many ways these five year-old consoles cans till handle - albeit at a significant cost - similarly high-end visuals that grace today's top of the range PC's.
More impressive - although not really much of a surprise considering that a earlier config file hinted as much - is the how well the PS3 code actually matches up to the 360 and PC versions. Its basically as good in many ways, with some very small improvements over the 360 game, and one or two slight (but hardly impacting) cut backs, non of which really impact on the game as a whole in any meaningful way.
As to which one you should pick up. Well, it's fair to say that both versions come recommended, but perhaps the 360 build edges it ever so slightly with mildly better image quality and a smoother frame-rate. But really, it makes very little difference, and I personally found aiming a little easier when using the PS3's analogue sticks despite the sometimes heavier drops in framerate.
So, Crysis 2 has successfully made the translation to console, but how well does it hold up compared to the cutting-edge PC game, and is there any evidence to suggest that it is this version that has in fact been downgraded in order to support the consoles? Moving on to our triple format comparison, let's take a closer look.
'High' settings
The first thing to note, is that unlike the console versions of the game Crysis 2 can be made to run in a range of different resolutions natively, and as a result, even at 720p the game looks noticeably sharper than its Xbox 360 and PlayStation 3 counterparts. In addition, there are three distinct graphical settings designed to scale various features in order to gain better performance on lower spec systems, or, to deliver extra visual impact on more powerful, higher-end machines.
Crytek state that the PC game on it's lowest setting 'high' is a good approximate of what the actual console versions look like, but with further tayloring of effects specifically for the computer platform. Although, both console versions actually shape up really, really well in comparison - minus the obvious drop in resolution.
As you can see above, the benefits of the PC game's true 720p output delivers a noticeable sharpness advantage, but we can also see that lighting has been given an upgarde. More light sources are visible with the lighting bounce having more of an overall imapct, casting more shadows as a result.
The other big difference comes in the form of how the PC versions handles LOD - it's far less impactful with details being loaded in much further away from the player. This is down to the fact that on the PC, the game doesn't stream in any assets at all. Instead, it takes advantage of the much larger amouts of memory available on the format in order to avoid doing this, preserving object quality across the entire game.
'High' settings
'Very High' settings
'Extreme' settings
The difference though, isn't alarmingly huge, and there are many aspects of the PC code which look virtually like for like across all three formats, with texture detail being the prime example. In order to really see how far Crytek have really taylored their engine when it comes down to running on mid to high-end hardware, we have to look at the game's highest graphical setting, 'extreme'.
Here we start to see the PC version push ahead with a small wealth of graphical upgrades, although some are pretty subtle in nature and could easily be overlooked given the quality of both console conversions. The most obvious change, is that both lighting and shadowing have both been upgraded. Not only does every light practically cast a shadow, but lighting in general seems to be more refined and accurate in nature. SSAO too has also been given a higher precision implementation, looking a touch cleaner as a result.
'High' Settings
'Very High' settings
'Extreme' settings
Additionally, Crytek's custom temporal MSAA solution offers better coverage across the entire scene, albiet with an incease in image blur. There's less in the way of visible aliasing giving the game a cleaner look on the whole. However, this upgrade is only present when running the PC game in the 'extrme' setting. When playing in either 'high' or 'very high' the game appears to use less samples in order to generate the AA, with similar levels of edge-smoothing as found on the PS3 and 360 versions. But the game looks sharper as a result.
'High' settings
'Extreme' settings
In other areas we see a few more subtle upgrades. The game's representation of water for example appears more complex in nature. As you can see above, there are greater amounts of ripples and waves on show, while the effect is animated with more accuracy than on consoles; a direct result of both an incrase in both geometry and better normal map blending being used to recreate the effect.
Furthermore, other elements such as motion blur and depth of field benefit from higher quality implementations. In particular DOF features an additional layer in the distance not present in the console versions, whilst its resolution is also higher too. Motion blur is more impacting, but the results appear cleaner than either console game while increasing the amount of screen distortion present.
'Extreme' settings
Moving on, and one of the most important factors of PC gaming is being able to bump up the resolution, plaster on all the top-end effects and still run the game smoothly. And while performance across both 720p and 1080p modes (hardware dependant of course) is noticeably better than on the 360 and PS3, there's also a real sense that bumping up the resolution does very little to improve the visuals compared to selecting one of the higher graphics modes.
Below we've listed some shots to show our findings. The two are grabs from the game running in 'extreme' mode, showcasing how 1080p fares in comparision. And as you can see, the differnce isn't particular spellbinding. Granted, on a native 1080p HDTV the jump in sharpness is easily apparent, but at the same time the additional clarity is somewhat reduced by the game's seemingly restrictive art assests. It seems like the baseline visuals were optimised in order to run across a really broad range of specifications, and thus we see no real leap in detail from moving up the resolution chain.
'Extreme' settings - 720p
'Extreme' settings - 1080p
However, without a doubt, Crysis 2 running in 1080p on mid-range or mid high-end hardware is a class act indeed. Graphically, although the improvements are sometimes subtle at best, there's still a sense that you are seeing the game in its most polished form; a form that is unobtainable on any other hardware outside that of a decent gaming PC.
Case in point, running on a i5 CPU and a NVIDIA GTX460 GPU at 720p (1280x720) with v-sync enbabled, and we get a near constant 60fps update when using the 'extreme' setting. While running in 1080p, v-synced and on 'high' you get a solid 60fps update. All of which amounts to an experience way beyond the home gaming systems.
But that said, we can't help but feel that the PC game is in fact being held back slightly by both console versions. Clearly, Crytek have focused on designing an engine around running on age-old hardware with most of its effects correct and present, than in producing high-end tech that requires the most obsecne PC's in order to run. What substantiates this, is that on day one there is no Direct x11 support at all. By contrast, the original Crysis featured a beefed up Direct X10 mode on launch, with further upgrades separating it even further from the older DX9-based configurations contained within.
My best guess is that due to spending so much time on delivering two suitable impressive - and very comparable - consoles expereiences, that a bleeding-edge, highest-end graphical mode simply wasn't ready in time for luanch. Crytek have already confirmed that an 'ethusiast mode' is on its way, along with Direct X11 support, but whether or not it shall yield the same level of technical superiority that Crysis 1 had over its peers is unknown at this point.
Either way, there's still no question that Crysis 2 is a remarkably stunning release on the PC, and what it lacks in pure graphical advantage, it more than makes up for with vastly superior performance. Running the game at 720p, v-synced, and at the console target of 30fps should be no problem for most decent, older PC's, while people with mid-range and cutting edge hardware will easily be able to blow away both console versions entirely.
Overall, it's pretty obvious that the PC version of Crysis 2 is the one to go for given the choice - and you don't even need the very best hardware to do a good job of running it - but you shouldn't discount the PS3 and 360 builds out entirely. What Crytek have manged to achieve on those consoles is nothing short of legendary, and although there are still a fair share of problems which impact on the experience as a whole, they're both beautiful looking games regardless.
Thanks go out to AlStrong for the pixel counting and Richard Leadbetter for use of Digital Foundry's analysis tools.
Sunday, 27 March 2011
Tuesday, 22 March 2011
Tech Analysis: Homefront (360 vs PS3)
THQ's ambitions as a publisher have been laid out on a plate for all to see. With Homefront they aim to compete on the same level as Activision and EA with their top-tier military shooters - namely Call Of Duty and Battlefield - thus making their own mark within a vastly overcrowded genre. But doing so is no easy task. And in order to achieve their lofty goal they need to deliver something as production-line polished as those titles, but with greater scope and a bigger bang.
Sadly Kaos Studios take on things leaves a lot to be desired. While the multiplayer modes are filled with plentiful options and neat touches which balance out the game for both the hardcore elitist and the casual player, the campaign mode of Homefront feels rough around the edges and tightly pre-scripted, let down by poor design choices and the unwillingness to let the player really take control in driving forward the action. It is rather atmospheric however.
On the other hand, Kaos' use of the Unreal Engine 3 commands at least some respect. The developer has taken a engine which is most suited for smaller, more enclosed spaces and customised it to handle much larger draw distances, and wider expanses of terrain which deliver an added sense of scale to the proceedings. It's not the first time that this has been done: Frontlines: Fuel of War used the engine as a building block for something bigger and better, whilst Epic's own Gears Of War 2 showed how the tech was capable of breaking free from its original constraints.
A fair amount of work has also been put into the multiplatform aspect of the game's development, tayloring certain graphical elements for each console. As a result we see some noticeable plus points for the PS3 over the 360, and vice-versa. However, both games are visually underwhelming, with a range of issues from sub-HD framebuffers to a few poorly implemented effects.
It's fair to say that neither version holds a candle to the very best shooters which utilise the very same engine, though not without some merit.
The game world itself provides much in the way of atmosphere, with the developer's impressive use of lighting, and the building up of a detailed, middle American town scape ravaged by war both helping to connect you with what's going on inside your TV screen. However, that sense of immersion is disrupted by the game's inconsistent visual nature.
The framebuffer for example, comes in at a lowly 1024x576 (basically PAL DVD resolution) on both formats. There's a distinct blur over the entire image which helps hide intricate texture details, along with plenty of noticeable edge shimmering. Pretty much everything looks hideously fuzzy, marred by the hefty upscale that's going on here.
Xbox 360 owners get the benefit of 2xMSAA helping in smoothing over some of the jaggies and reducing a few upscaling artefacts. Whilst on PS3, the game runs without any kind of anti-aliasing whatsoever, thus making it look even rougher as a result. Both are far from delivering anything close to native HD goodness, but its clear that the 360 version features slightly better IQ.
The drop in resolution may come as a shock to some, but in the context of producing a highly detailed and reasonably open landscape, there are significant costs involved - especially when you take into account the game's stellar HDR lighting implementation and real-time shadowing. The additional memory bandwidth and GPU cycles spared by rendering in a sub-HD resolution allow for longer draw distances and considerable more intricate environments. And therein lies the compromise.
Of course, in upping the amount of rendering you have to do for any given scene, this often means carefully implementing LOD set-up and object/texture steaming in order to maintain a high level of detail without impacting on the core graphical make-up of the game. Older titles that used the UE3 featured prominent texture streaming issues, whereby higher resolution textures would pop-in extremely close to the player. Although since then, this has become far less of an issue, with changes in the way transitions between textures are displayed.
In Homefront we see that these refinements aren't present at all, and the end result is that both versions feature some noticeable texture pop whist the engine loads in the higher quality assets. You can see this in the screenshots above. But in this case it is the PS3 game which suffers far less from these issues - sometimes higher quality assets never load in on the 360 either.
We can almost certainly say that this is a direct result from the game imposing a mandatory install onto the system's hard drive, thus allowing content to be streamed in much faster than directly of off the DVD on 360.
On the flip side we find that most of the game's art assets are basically identical, with only a few instances of less detailed textures gracing the 360 version (sometimes it appears that higher resolution assets never load in). Use of texture filtering on 360 also appears to be better too, with higher levels of AF (anisotropic filtering) being present. Although, this does very little to improve the look of the game.
Moving on, there are a few other differences between the two games.
Starting with the foliage, and we can see above that certain plants and bushes are being rendered in a lower resolution in the PS3 game, whilst others are completely missing from the scene altogether. As only a few small objects have been cut back on, it's often hard to see any difference, which thankfully means that it has little to no impact on the look of the PS3 game in any meaningful way.
Secondly, there is the use of lower resolution alpha buffers on both formats. Effects have been filtered accordingly to look smooth, but some - like fire in particular - look very flat and unconvincing. Smoke other the other hand, has a reasonable amount of depth to it.
The main reason looks to be centered around the heavy bandwidth requirements of the customised UE3, and Kaos's use of a modified lighting system, both of which impacts on both system's available resources.
A look at Homefront's performance, and it's fair to say that most of these compromises are based around getting the game up and running at a relatively smooth frame-rate - at the usual target of 30fps.
As is now standard practice here at IQGamer, we've put together an analysis video showcasing performance across a range of scenes from both versions of the game.
The results are hardly complementary, revealing a slew of frame-rate and screen tearing issues pertaining to both builds. What's clear however, is that across the run of play the 360 game manages to maintain a smoother frame-rate, but tears far more often as a direct consequence. Whilst on PS3, the game drops under the 30fps mark slightly more frequently but features a still noticeable, although reduced amount of tearing.
Performance profiles differ across both platforms. The 360 game runs without v-sync and an uncapped frame-rate - thus we see near constant tearing as the engine delivers loads of unfinished frames to be displayed. But this results in the smoothest controller response possible outside of the game slowing down. By contrast, on the PS3 Homefront employs v-sync, but simply drops it when the game goes below the capped 30fps update. In fact, the game could well be soft v-synced (whereby tearing constantly occurs in the overscan area at the top of the screen), whilst doing the same thing.
Looking at the video it's pretty obvious that both versions constantly fail to meet the targeted 30fps without throwing out out scores of incomplete (torn) frames. Although the game does stay mostly around the intended level - on PS3 at least. The 360, with its uncapped frame-rate is the smoother of the two. But this comes at the expense of visibly more screen tearing throughout the entire experience. Whereas, on the other hand the PS3 game features more frame drops (particularly in heavy scenes), but with less intrusive tearing.
All in all, performance on both formats comes as a disappointment, especially given the trade-offs made in order to ensure as smooth an experience as possible. Unsurprisingly, the PS3 benefits from having slightly smoother controls given its more stable frame-rate. However, when things get busy the opposite is true, making it six of one, and half a dozen of the other in this case.
Far more impressive, is the game use of lighting and the amount of atmosphere it provides. When combined with the sense of scale, and detail of the surrounding environment, it goes along way to sucking you into a world swamped in turmoil. Lens flair is just the one of the initial focal point of the game's lighting implementation, whilst the range provided by the use of HDR delivers ample contrast to the scene in light or dark areas.
The 360 also benefits further in getting extra depth provided by what looks like additional light occlusion, which in turn seems to accentuate the game's use of SSAO. By contrast, while the PS3 game still features plenty of depth also, SSAO has less of an impact on the scene. This can be seen above - not only does the PS3 version appear brighter, but in corners where all three walls meet, there is significantly less in the way of ambient shadowing in some scenes.
In addition the game's use of SSAO on both platforms is distinctly buggy. A noticeable halo effect can be seen around characters as they stand near walls and other objects, along with various environmental objects too as you approach them from differing angles. Additionally, this haloing can also cause shadows on characters to flicker and glitch in and out of view on occasion.
Homefront is a pretty interesting idea, with a potentially engrossing premise and certainly the right artistic design, but one might indeed wonder whether or not Kaos have actually chosen the right tech to do the job properly.
Certainly, a few things stick out from the muddy-looking mess the game sometimes seems to resemble. The lighting for example, in combination with the detailed environmental construction and the game's musical score certainly creates a sense of immersion. And the motion blur/screen distortion effects add impact to the combat when pinned down and being fired upon. But this is continuously counter-balanced with a fuzzy, upscaled framebuffer and blurry looking low resolution effects.
The very same thing applies to both the Xbox 360 and PlayStation 3 games. While 360 owners benefit from having higher levels of AF and additional lighting, plus anti-aliasing, on the PS3 we see better texture steaming and what amounts to a more consistent frame-rate - despite larger drops when the engine is put under load. In which case it's pretty hard to recommend one over the other, or either version in general.
In conclusion, I'd perhaps say that the PS3 delivers the most consistent experience overall. The lack of AA doesn't harm the game as much as it would in other titles - due to the heavy upscale making things look rough on both - and the missing lighting doesn't always dramatically change every scene. But in terms of performance, the steadier frame-rate and lesser amount of screen tearing is a bonus, even if performance under load suffers.
Perhaps, rather than fighting over which version is 'the' one to go for (when neither are great), you should instead think about which version your friends will be playing online. The campaign in Homefront is definitely a second-tier experience - multiplayer online is exactly what this game was made for. And in that respect therein is where your final decision should rest.
Once again, thanks go out to Richard Leadbetter for use of Digital Foundry analysis tools.
Sadly Kaos Studios take on things leaves a lot to be desired. While the multiplayer modes are filled with plentiful options and neat touches which balance out the game for both the hardcore elitist and the casual player, the campaign mode of Homefront feels rough around the edges and tightly pre-scripted, let down by poor design choices and the unwillingness to let the player really take control in driving forward the action. It is rather atmospheric however.
On the other hand, Kaos' use of the Unreal Engine 3 commands at least some respect. The developer has taken a engine which is most suited for smaller, more enclosed spaces and customised it to handle much larger draw distances, and wider expanses of terrain which deliver an added sense of scale to the proceedings. It's not the first time that this has been done: Frontlines: Fuel of War used the engine as a building block for something bigger and better, whilst Epic's own Gears Of War 2 showed how the tech was capable of breaking free from its original constraints.
A fair amount of work has also been put into the multiplatform aspect of the game's development, tayloring certain graphical elements for each console. As a result we see some noticeable plus points for the PS3 over the 360, and vice-versa. However, both games are visually underwhelming, with a range of issues from sub-HD framebuffers to a few poorly implemented effects.
It's fair to say that neither version holds a candle to the very best shooters which utilise the very same engine, though not without some merit.
The game world itself provides much in the way of atmosphere, with the developer's impressive use of lighting, and the building up of a detailed, middle American town scape ravaged by war both helping to connect you with what's going on inside your TV screen. However, that sense of immersion is disrupted by the game's inconsistent visual nature.
The framebuffer for example, comes in at a lowly 1024x576 (basically PAL DVD resolution) on both formats. There's a distinct blur over the entire image which helps hide intricate texture details, along with plenty of noticeable edge shimmering. Pretty much everything looks hideously fuzzy, marred by the hefty upscale that's going on here.
Xbox 360 owners get the benefit of 2xMSAA helping in smoothing over some of the jaggies and reducing a few upscaling artefacts. Whilst on PS3, the game runs without any kind of anti-aliasing whatsoever, thus making it look even rougher as a result. Both are far from delivering anything close to native HD goodness, but its clear that the 360 version features slightly better IQ.
The drop in resolution may come as a shock to some, but in the context of producing a highly detailed and reasonably open landscape, there are significant costs involved - especially when you take into account the game's stellar HDR lighting implementation and real-time shadowing. The additional memory bandwidth and GPU cycles spared by rendering in a sub-HD resolution allow for longer draw distances and considerable more intricate environments. And therein lies the compromise.
Of course, in upping the amount of rendering you have to do for any given scene, this often means carefully implementing LOD set-up and object/texture steaming in order to maintain a high level of detail without impacting on the core graphical make-up of the game. Older titles that used the UE3 featured prominent texture streaming issues, whereby higher resolution textures would pop-in extremely close to the player. Although since then, this has become far less of an issue, with changes in the way transitions between textures are displayed.
In Homefront we see that these refinements aren't present at all, and the end result is that both versions feature some noticeable texture pop whist the engine loads in the higher quality assets. You can see this in the screenshots above. But in this case it is the PS3 game which suffers far less from these issues - sometimes higher quality assets never load in on the 360 either.
We can almost certainly say that this is a direct result from the game imposing a mandatory install onto the system's hard drive, thus allowing content to be streamed in much faster than directly of off the DVD on 360.
On the flip side we find that most of the game's art assets are basically identical, with only a few instances of less detailed textures gracing the 360 version (sometimes it appears that higher resolution assets never load in). Use of texture filtering on 360 also appears to be better too, with higher levels of AF (anisotropic filtering) being present. Although, this does very little to improve the look of the game.
Moving on, there are a few other differences between the two games.
Starting with the foliage, and we can see above that certain plants and bushes are being rendered in a lower resolution in the PS3 game, whilst others are completely missing from the scene altogether. As only a few small objects have been cut back on, it's often hard to see any difference, which thankfully means that it has little to no impact on the look of the PS3 game in any meaningful way.
Secondly, there is the use of lower resolution alpha buffers on both formats. Effects have been filtered accordingly to look smooth, but some - like fire in particular - look very flat and unconvincing. Smoke other the other hand, has a reasonable amount of depth to it.
The main reason looks to be centered around the heavy bandwidth requirements of the customised UE3, and Kaos's use of a modified lighting system, both of which impacts on both system's available resources.
A look at Homefront's performance, and it's fair to say that most of these compromises are based around getting the game up and running at a relatively smooth frame-rate - at the usual target of 30fps.
As is now standard practice here at IQGamer, we've put together an analysis video showcasing performance across a range of scenes from both versions of the game.
The results are hardly complementary, revealing a slew of frame-rate and screen tearing issues pertaining to both builds. What's clear however, is that across the run of play the 360 game manages to maintain a smoother frame-rate, but tears far more often as a direct consequence. Whilst on PS3, the game drops under the 30fps mark slightly more frequently but features a still noticeable, although reduced amount of tearing.
Performance profiles differ across both platforms. The 360 game runs without v-sync and an uncapped frame-rate - thus we see near constant tearing as the engine delivers loads of unfinished frames to be displayed. But this results in the smoothest controller response possible outside of the game slowing down. By contrast, on the PS3 Homefront employs v-sync, but simply drops it when the game goes below the capped 30fps update. In fact, the game could well be soft v-synced (whereby tearing constantly occurs in the overscan area at the top of the screen), whilst doing the same thing.
Looking at the video it's pretty obvious that both versions constantly fail to meet the targeted 30fps without throwing out out scores of incomplete (torn) frames. Although the game does stay mostly around the intended level - on PS3 at least. The 360, with its uncapped frame-rate is the smoother of the two. But this comes at the expense of visibly more screen tearing throughout the entire experience. Whereas, on the other hand the PS3 game features more frame drops (particularly in heavy scenes), but with less intrusive tearing.
All in all, performance on both formats comes as a disappointment, especially given the trade-offs made in order to ensure as smooth an experience as possible. Unsurprisingly, the PS3 benefits from having slightly smoother controls given its more stable frame-rate. However, when things get busy the opposite is true, making it six of one, and half a dozen of the other in this case.
Far more impressive, is the game use of lighting and the amount of atmosphere it provides. When combined with the sense of scale, and detail of the surrounding environment, it goes along way to sucking you into a world swamped in turmoil. Lens flair is just the one of the initial focal point of the game's lighting implementation, whilst the range provided by the use of HDR delivers ample contrast to the scene in light or dark areas.
The 360 also benefits further in getting extra depth provided by what looks like additional light occlusion, which in turn seems to accentuate the game's use of SSAO. By contrast, while the PS3 game still features plenty of depth also, SSAO has less of an impact on the scene. This can be seen above - not only does the PS3 version appear brighter, but in corners where all three walls meet, there is significantly less in the way of ambient shadowing in some scenes.
In addition the game's use of SSAO on both platforms is distinctly buggy. A noticeable halo effect can be seen around characters as they stand near walls and other objects, along with various environmental objects too as you approach them from differing angles. Additionally, this haloing can also cause shadows on characters to flicker and glitch in and out of view on occasion.
Homefront is a pretty interesting idea, with a potentially engrossing premise and certainly the right artistic design, but one might indeed wonder whether or not Kaos have actually chosen the right tech to do the job properly.
Certainly, a few things stick out from the muddy-looking mess the game sometimes seems to resemble. The lighting for example, in combination with the detailed environmental construction and the game's musical score certainly creates a sense of immersion. And the motion blur/screen distortion effects add impact to the combat when pinned down and being fired upon. But this is continuously counter-balanced with a fuzzy, upscaled framebuffer and blurry looking low resolution effects.
The very same thing applies to both the Xbox 360 and PlayStation 3 games. While 360 owners benefit from having higher levels of AF and additional lighting, plus anti-aliasing, on the PS3 we see better texture steaming and what amounts to a more consistent frame-rate - despite larger drops when the engine is put under load. In which case it's pretty hard to recommend one over the other, or either version in general.
In conclusion, I'd perhaps say that the PS3 delivers the most consistent experience overall. The lack of AA doesn't harm the game as much as it would in other titles - due to the heavy upscale making things look rough on both - and the missing lighting doesn't always dramatically change every scene. But in terms of performance, the steadier frame-rate and lesser amount of screen tearing is a bonus, even if performance under load suffers.
Perhaps, rather than fighting over which version is 'the' one to go for (when neither are great), you should instead think about which version your friends will be playing online. The campaign in Homefront is definitely a second-tier experience - multiplayer online is exactly what this game was made for. And in that respect therein is where your final decision should rest.
Once again, thanks go out to Richard Leadbetter for use of Digital Foundry analysis tools.
Subscribe to:
Posts (Atom)