Tuesday, 8 March 2011

Tech Analysis: Dragon Age 2 Demo (360 vs PS3)

It's very late and has been sitting on the table for a few days now while I finish up something else I've been working on. Not really had the time to be fair. But here's our demo analysis of Dragon Age 2.

The PS3 usually gets the short end of the stick when it comes to multiplatform conversions. Just look at the likes of Red Dead Redemption or Bioshock, with their sub-HD rendering resolutions and use of screen blurring anti-alising techniques. Or, in the case of the latter; no AA with a simple blurring of the image instead. Of course, on the other side of the coin you have titles such as Vanquish, which commanded a tiny lead of its 360 counterpasrt with slightly better overall performance. And Mass Effect 2, with improved shadowing and athestically more pleasing lighting.

The original Dragon Age Origins almost found itself in the latter catagory; a game in which PS3 owners gained additional graphical upgrades in the form of increased texture filtering and better depth of field. While neither version looked particularly great - a fault of the game rather than the consoles inhability to run it - it was Sony's black box of tricks that, at first glance, came out on top. Except for in terms of performance: it was pretty dire in this regard.

Given the mediocre graphical look of the last game, the team at Bioware have stated on various occasions that they planned to improve on the visual make-up and engine powering the series for Dragon Age 2. And in the recent XBL and PSN demo, we can clearly see a few key changes. But most of all a tangible improvement all round, with the 360 build more than matching up to - and exceeding in certain areas - the PS3 code.


360


PS3

The most obvious of these upgrades on both formats, is that the characters and environments in the game feature better use of normal mapping. There are a range of surfaces that appear curved and have a noticeable depth to them. By comparison, in the original Dragon Age many parts of the game featured fairly low resolution, flat textured areas of the environment, with characters that lacked some of the additional effects present in this sequel.

Of course, some of the low resolution elements remain. But here we also see an abundance of mixed hi and low resolution textures, and some key, higher quality elements interwined.

The armour on some of the characters for example, features low resolution textures, whilst some of the surrounding clothing uses higher quality ones. It's quite a visual discrepancy to say the least, with some very obviously poor texturework shoehorned into more detailed areas. On PC this varience is further enlarged. But, thankfully, on consoles appears much less of an issue.


360


PS3

In terms of the framebuffer, we see both versions rendering in 720p. The main difference here, comes in the form of anti-aliasing used in each version's graphical make-up. On 360, we get the standard issue 2xMSAA solution, while PS3 owners are granted a custom take on another MLAA implementation.

While we all know how MSAA works, MLAA is a somewhat lesser known beast. Despite being given a specific title, MLAA can feature a wide variety of implementations, each dealing with the scene a little differently. Universally though, they all have one thing in common: they only work on the finished rendered scene - basically as a post process on the framebuffer before it gets displayed on screen.

The choice of opting for MLAA makes sense when dealing with both consoles impacting memory contrasints compared to the PC, and especially on the PS3, whereby the additional processing power of the CELL allows you to free up more GPU cyclers for other tasks. And in Dragon Age 2, this seems to be the key reason for its uptake on Sony's platform: Better edge smoothing with less impact on memory and processsing requirements compared to MSAA.


360


PS3

As you can see above, this is definitely true. Pixel edges are smoothed over incredibly well, matching MSAA sampling well over the 4x limit a select few 360 titles employ. Naturally, you don't get any sub-pixel coverage or accuracy when dealing with MLAA. Although this seems to be far less of an issue in Dragon Age 2. On the PS3, the game looks even cleaner compared to its 360 counterpart. The AA is superb, and not just for a console title.

That said, there is another small niggle that comes with Bioware's custom form of MLAA. And that is the unwanted side-effect of blurring texture details. Essentially, all MLAA-based solution work around a purely edge-detect method of sampling: you find the pixels which look like an edge, and then apply the smoothing algorithm. However, with no geometry information to extract additional data from, you are left at the mercy of how effective your algorithm is at deciphering a legitimate edge, and faux-edges of sorts found in contrasting parts of a texrture map.

It's evident from the screens that in Bioware's case, the end result has very little impact on the image overall. Yes, there is some blur, but it is nothing overtly noticeably, nor does it really impact in any meaningful way. It is largely incosequencial.

So, in that respect, the PS3 versions gains the upper hand with regards to overall IQ, with better edge-sampling coverage, and with little in the way of annoying sub-pixel issues. Then again, DA2 looks clean and sharp on both platfroms. But it is definitely cleaner on the PS3.


360


PS3


360


PS3

Outside of the framebuffer make-up there is very little to seperate the two versions at all with regards to their graphical look. Lighting, like with Mass Effect 2 appears to be handled a little differently across both versions, although this has little imapct with no one version appearing better than the other.

You can see the key indicator in the screenshots above. Notice how the specular reflections are in different places in both versions in the second shot, whilst the actual frame captures are the same. Colour tone also appears to be different too, seen clearly in the first batch of screens. Compared to the huge variance in Mass Effect 2, the lighting between versions is significantly closer.


360


PS3

Also, the use of alpha-based effects, such as the flame effects in the cut-scenes see some small changes when comparing both games. They seem to be represented a tad differently on both platforms.


360


PS3

Dragon Age 2 also features the use of self-shadowing across all platforms (PC, PS3 and 360). However, the effect looks to have been given a weaker implementation on the PS3. You can see in the shots above. The PS3 game appears to feature much lighter self-shadows, but also missing them on certain parts of the characters. The the impact from the beads around the character's neck, for example, barely cast a complete shadow in the above screens by comparison.

Outside of this, bar the way certain effects are handled, both versions are actually visually very close. There are a few passing oddities, but nothing remotely enough for either version to command a solid lead in this respect. The use of MLAA is by far the most impacting factor in the PS3 code.


360


PS3

One area of the original Dragon Age that left a lot to be desired on the PS3, was its performance. It's safe to say that BioWare's first outing suffered at the hand of numerous framerate drops causing heavy slowdown. By comparason, the 360 game, although slowing down, maintained its baseline 30fps for longer stetches than on the PS3.

Here in the demo code, we find similar traits manifesting them selves during gameplay. Although it is also apparent that progress has definitely been made in this regard. Seeing as the demo is rather short, and that I've only been able to access the first part of it, its far to early to draw any solid conclusions. However, the fact remains that performance still the biggest factor seperating the two versions.

Dragon Age 2 opts to target a 30fps update, and runs with v-sync solidly enabled. As a result, we find that the game never tears but instead drops frames as it struggles to keep up with the renderer. Rather than the 360 leading on all fronts, here we see a difference in performance in both gameplay and in cut-scenes.

Gameplay-wise, both builds start out at 30fps. And both maintain this very well in the absence of any load. Simply running into a group of enemies, hacking and slashing away rarely leads to much, if any slowdown. Throw in a large foe into the works surrounded with five or six smaller adversaries, whilst strafing around and spining the camera, and we find that both versions will slow down to around the 20fps mark or so. But the defining element appears to be alpha-based. Effects like smoke and fire cause the PS3 game to drop frames more often, and lower down than its 360 counterpart.

Cut-scenes operate a little differently. Here we see heavy drops below both the 30fps and 20fps mark. However, it also appears that these drops are slightly less severe on the PS3. Even though at times, both slow down in equal measure. Just like during gameplay, large creatures and specific effects are the root cause - when there's little going on to stress the engine, both run smoothly at 30fps.


360


PS3

From what we've played so far Dragon Age 2 is hardly representative of technical excellence. Low res textures, unimpressive effects, and a unstable framerate - mainly in the cut-scenes - can make for a unsightly experience at times. However, there's no doubt that the game looks much, much better than the first. Characters contain more detail, and the 360 build features an all round improvement in IQ. While PS3 users get the benefit of a vastly superior AA solution.

However, the demo is but a mere taster of what the game will have to offer. At least both versions seem to be almost on a par now. Performance aside - which still suffers more in gameplay on the PS3 - things are looking a lot better for Bioware's latest. But for an engine that is boardering on being five years old, upgraded for this sequel, one can only wonder just how far it will take them.

With the final game out in only a couple of days, I guess that we won't have long until we find out. If we get time, then expect an update. Hopefully we can also bring along a few comparison vids to the table too.

1 comment:

  1. Thanks for another great analysis. I need to learn to wait for these before purchasing games!

    ReplyDelete