Wednesday, 24 November 2010

Tech Analysis: Need For Speed: Hot Pursuit (PS3 vs 360)

After taking a fairly in-depth technical look at the demo for Need For Speed: Hot Pursuit, we came away decidedly impressed with the superb quality of the work on offer. The use of a highly optimised multi-platform engine, whereby through parralelisation of code - in the form of scalable modules which can run on multiple cores/SPU’s – not only yielded solid results across both platforms. But also confirmed that Criterion’s approach to development was indeed the right one.

In terms of platform parity, the demo was for all accounts, identical on both platforms, with next to nothing in separating them. Texture detail, filtering, lighting, and almost every graphical effect had been careful replicated on both the PS3 and the 360, whilst performance was surprisingly rock solid - 30fps being upheld near constantly, with no screen tearing taking place.

The most interesting parts of our analysis focused us on the game’s use of anti-aliasing - what looked like an additional technique had been included over and above the standard 2xMSAA solution, and the use of an incredibly impressive dynamic lighting system. Both of which were points we wanted to investigate further outside of the night time track we had only access to in the demo. And these are exactly the things we’ll be taking another look at here today, along with another look at performance in the company of the final game.

First of all, just a quick recap. NFS:HP renders in 720p on both platforms with edge smoothing being provided by use of 2xMSAA. Extra smoothing is also present from a currently unknown, custom form of AA, which is used to help reduce sub-pixel aliasing issues on thin strips of geometry and objects that appear far off into the distance.

In the demo we reported on the noticeable improvement in overall image quality present from the additional implementation of custom AA, although at the same time where left with only half the picture. For our demo analysis, we only had access to the night time portion of the game, which as regular readers of IQGamer should know, low contrast areas often provide a best case scenario for most anti-aliasing techniques (supersampling and MLAA aside). The real test of how well Criterion’s custom solution actually works, is in the higher contrast daytime sections of the final code.

As you can see in the screenshot above, like in the demo, power lines and small objects far off into the distance get a huge amount of edge smoothing not possible just by using 2xMSAA on its own. Edge shimmering is noticeably reduced as a result, and the scene has a more solid look to it.

Not all elements of the scene are covered however. Some objects, like the telegrapth poles and small fences at the side of the road, suffer from both high contrast aliasing and subpixel issues, where by the samples created by the MSAA are insufficient to deal with such things. These do result in some jagged and shimmering edges being present, sometimes unavoidably so. Although, overall sampling coverage, and indeed jaggies reduction is very good considering the look of the game.

Despite the high contrast nature of some of the daytime scenes, aliasing is indeed kept in check, with Criterion’s technique successfully aiding the 2xMSAA solution also apparent. As expected, jaggies aren’t completely eliminated - they can crop up frequently throughout the trackside scenery. But the overall result is more than satisfactory given the make up of the game engine – the lighting, huge draw distances etc.

Another thing that we were thoroughly impressed with in the demo, and that we wanted to check out in the finished game, in full daylight environments no less, was the title’s use of dynamic lighting. In NFS:HP the cars are lit and shaded in real time by the surrounding environment, with elements such as cloud coverage dramatically changing the lighting applied to the scene at any given time.

Image based lighting is used to do this, where the actual environment and lighting scheme are rendered first, before the cars are rendered afterwards in a separate pass. This allows the cars to be accurately lit and shaded at all times, changing constantly with regards to their position in the game world itself.

At night the range of different light sources in combination with cloud coverage gave way to an incredibly realistic look, with lighting that brought about a certain amount of naturalness to the overall look of the game. In the daytime we can see this effect being heightened even more. Various elements: such as the sun rising above hilltops as you come speeding around corners, and the constantly moving cloud dramatically impacts on lighting present in the scene, reacting instantly with the cars as they are being driven around the track.

With shadows constantly shifting, lighting is never the same across both versions at any given time. There is more range, and indeed scope on offer here too, making a noticeable difference. Although, the actual quality and implementation of the effect is the same on both PS3 and 360. This accounts for any lighting differences apparent in our comparison screens.

Interestingly, not every aspect of the game appears exactly like for like. In the demo we noticed that the specular maps on both the road and some environment surfaces seemed to be rendered in a slightly lower resolution on the PS3, and as we can see in the screenshot above, the same thing can indeed be found in the final game.

However, this oddity is only present under certain circumstances – the difference simply doesn’t exist when racing in full daylight conditions, nor does it appear above on the wall of the tunnel either (look to the left, it's the same on both). So maybe something else is interfering with it in some way. Either way, signs point to the reduced quality effect only being present when the car's headlights are directly shining on the road, in either dark parts of the track, or at night time.

In terms of performance, the final game, is as expected exactly like in the demo. NFS:HP runs at a rock solid 30fps (the game is framerate locked) at all times during gameplay on both formats, with the only slowdown occurring in takedown or car crash scenes, along with on some cinematics before and after the race. As these segments aren’t controllable by the player, the slowdown makes no real impact on the proceedings, other than visually, so there is no loss of controller responsiveness to be found during gameplay.

The use of v-sync is also fully apparent, with neither version exhibiting any screen tearing whatsoever. The high contrast nature of the daylight courses make tearing easier to see without having any equipment to measure it – that is to say that I saw none to be present at any point when playing the game.

So, performance is remarkably solid – a point we mentioned back when taking a look at the demo. But how does the game’s handling fair? Usually the lower the framerate, the greater the amount of latency has an affect on controller responsiveness, with any increase in lag being noticeable compared to games that run at 60fps. In the demo we sighted handling which felt slightly unresponsive, although actual controller feedback felt responsive.

As we first surmised, and experienced first hand in other games, the use of lower spec cars meant that fast turning or quick Burnout style drifting wasn’t really as easy as it should be. In fact, the handling model felt a little bit like Split Second – that is to say, that it felt a little unrefined. However, in the final game - with cars not a mile better than the ones given to us for use in the demo - we can see a marked improvement. The handling on the whole is far, far better, whilst maintaining that Burnout meets Split Second feel, without the compromise of feeling slightly laggy due to using those underpowered starting vehicles.

Of course, there’s no question that 60fps games, like Burnout Paradise, provide an ample improvement in controller responsiveness – lower latency means more instant feedback. Although, in this regard NFS:HP still feels incredibly responsive. More so it seems than many other 30fps racers, with the initial handling mechanic accounting for the difference.

Interestingly, it has been said that the PC version of the game can be made to run at 60fps with what looks like very little in the way of high-end hardware. Apparently, it is possible to acheive what I consider to be the benchmark framerate in which developers should strive for, with a simple mid-spec gaming rig. Unfortunately, my new PC isn’t ready yet, so I wasn’t able to test this out directly. But perhaps I might look into doing an update of sorts over the Christmas period if I have the time.

Either way, while it is obvious that NFS:HP would indeed benefit greatly from running at a higher framerate – sometimes the game is so fast that the 30fps update seems to lose its fluidity, even if it hasn’t actually changed – the experience is still a highly enjoyable one at that. Maybe the game isn’t quite as silky smooth as I would have liked – not at the super high speeds present. But of all the choices made, the compromise of having a rendering engine which draws massive vistas way into the distance, along impressive use of advanced dynamic lighting to boot, is a worthy one at that.

If its any consolation, the game’s use of motion blur often helps in making the game feel a tad smoother than it is. This is a common side effect of motion blur in general, in which the distorted nature of images in the scene can help to blend the separately rendered frames together, much like the way shutter speed affects the viewing of individual frames in film projection.

Given the insanely high speeds you are often driving at, the effect is a subtle one at best, felt more in the daytime races than at night. And sometimes, even it cannot make the game feel any smoother than it actually is. Thankfully, the constant 30fps update with no screen tearing keeps things nice and fluid. Although, that’s not to say a 60fps experience, with certain cut-backs made to the game’s advanced lighting system, wouldn’t be preferable.

In conclusion, the finished version of Need For Speed: Hot Pursuit is indeed as solid and technically accomplished as the demo - on both platforms no less, with only one slight difference doing absolutely nothing to tip the scales of balance in either way. Both versions come highly recommended, and although the lack of 60fps may come as a disappointment to ardent fans of Criterion’s past racers, that shouldn’t be enough to prevent you from screeching off that starting line in their company once again. Well… not if you want the best Need For Speed game in years that is.

Once again thanks go out to AlStrong for pixel counting and for their screens. The full gallery of higher quality shots can be found here.