Even without the Eurogamer comparison feature, the internet has been set ablaze by GTA IV PlayStation 3/Xbox 360 face-off discussions. But mere digital incendiaries are not enough for a release of this magnitude. It's time to go nuclear.
The biggest game of the year deserves the most in-depth face-off piece yet produced by anyone, anywhere, so we've massively upgraded the scope of our coverage and the amount of 'assets' you can view. There's still the same 720p and 1080p 24-bit RGB screengrabs taken from both versions of the game of course, this time using the new Digital Foundry TrueHD capture station, so why not take a look with the first of our galleries, featuring general gameplay shots?
Comparison videos appear to be all the rage, even though they're mostly low-res blur-a-thons that demonstrate absolutely nothing, so we decided to take a different approach. Our vids are encoded in supreme quality h264, but more than that, we've zoomed in on the action so you can actually see the difference. Unless otherwise stated, one pixel in the Eurogamer player corresponds to one pixel on your HDTV. Most of the clips are run at 50 percent speed too, again making the job of comparison that much easier, with more bandwidth dedicated to picture quality.
On top of that, there's an extra-special bonus too. Every video is embedded into the article, but there's a full-size, higher resolution link to the same material over at Eurogamer TV, offering a good 30 percent more screen area, with each clip reframed to make the most of the additional digital acreage. Use these if you want to link to our vids 'elsewhere'.
A sampler of what is to come can be found right here:
As there's a colossal amount of stuff to discuss, argue and insult each other about, we've broken down the feature into these areas:
I think the most important thing to say right from the get-go is that there's absolutely no doubt in my mind that Grand Theft Auto IV is a 10/10 game no matter which console you own. Having spent the last six days of my life playing both versions simultaneously, painstakingly matching up the gameplay and recording over 500GB of video captures in the process, everything in Tom's original review stands regardless of the platform you're playing it on. As one leading developer said to me the other day, "I think this game, like Crackdown before it, changes everything." Quite.
Yes there are some significant differences between the two games which we'll be going into in more depth later on in the feature, but it's important not to lose focus of the most important issue - that this release is an astonishing achievement. Content is king here, and until the extent of the Microsoft-exclusive DLC becomes apparent, the two games serve up an equally superb range of entertainment.
Even in terms of internet functionality the games are like for like. Similar to Burnout Paradise, the two games share all the usual online gaming accruements - the same levels, the same options, and very similar levels of performance, although obviously Xbox Live's servers have been coping better than the GameSpy equivalents behind the scenes at the PlayStation Network.
It's also clear to see that Rockstar has tried its level best to get the most out of both systems. Case in point is the Sixaxis support on the PlayStation 3 version. Regular readers of these features will know that I generally tend to find the motion sensor to be too imprecise to be worth bothering with, but at least here there's a sense that the programmers have tried to get the most out of it. There's a tutorial you access via the mobile phone that teleports you to the airport and lets you try out bikes, boats and helicopters - all of which offer Sixaxis support. Also invaluable is the way you can get the game to auto-calibrate Sixaxis each time the motion sensor is engaged, with the current position of the controller judged to be 'neutral'. This shows that some thought has gone into the control scheme, and there's no chance of your chopper suddenly lurching into the sky because your joypad isn't absolutely level.
Whether you actually find the motion sensor control to be anything other than a novelty remains a matter of personal taste. In terms of GTA IV, I find it decent enough with the boats, a trial with the bikes and short-lived and unintentional entertainment with the choppers. I honestly don't think that 360 owners are missing out on anything special by not having the same control system in their game.
Rockstar also makes use of the PS3 hard disk, with a mandatory 3,339MB installation of game data that takes around seven minutes. This doesn't do much for the initial loading of the game, which takes around 100 seconds on both systems, but certainly helps with mission loading, shaving a few seconds off each time a new task is initiated. There are also improvements in texture streaming as you navigate around Liberty City, but the advantages are barely perceptible.
Finally, a nice little touch is that the PS3 version of the game has much higher resolution versions of the various TV shows available to view in Niko's apartment. Additional camera angles are also included to show off the extra resolution - a cool bonus, somewhat reminiscent of a similar feature in Starbreeze Studios' PS3 conversion of The Darkness.
Resolution, Textures and Anti-Aliasing
Here's where things start to get interesting, where the two games really start to diverge. GTA IV has a different look on each machine, dictated by both technical limitations and - apparently - Rockstar's vision for the game.
First things first. Xbox 360 runs at full 720p (1280x720), whereas the PlayStation 3 code takes a 20 per cent hit, being natively rendered at 1152x640 before being software-upscaled. Regular readers of the face-off features know that this approach can work well (check out the comparison gallery for Dark Sector in the last face-off) but in most cases, the PS3 port just tends to look like a blurrier version of the original Xbox code: not impressive considering that typically, PS3 hardware - and sometimes software - is more expensive.
GTA IV is a kind of weird combination of the two. Both versions feature heavily post-processed visuals, in particular when it comes to depth-of-field effects. Objects in the distance on both versions are blurred in an effort to match the natural focus of the human eye. Where post-processed blur meets the upscaling effect of PlayStation 3, the result usually looks very good indeed - a little softer, of course, but rarely distracting. A slight change of hue (particularly noticeable on indoor cut-scenes) also makes the PS3 version look slightly warmer.
Technically speaking, Xbox 360 really should be winning this contest hands-down, but bizarrely, it doesn't. There's support for proper hardware-assisted anti-aliasing, eliminating a great deal of the jagginess of the PlayStation 3 version, plus it's running at full-fat 720p. However, Rockstar has introduced a 360-specific post-processing effect that dithers just about every texture on-screen. It's an effect not present at all on the PS3 version and serves to introduce an oil-painting-like effect to the overall look of the game, particularly on background objects. Unfortunately, it also seems to actively distort the edges of detail in the textures and occasionally looks really ugly.
As it is then, both versions of the game have their strengths and weakness. Ask anyone which look they prefer when presented with comparison shots and you'll find that opinion is divided pretty much straight down the middle. Even Rockstar VP and GTA IV co-writer Sam Houser stepped into the debate, telling 1UP that the 360 version had a "more clinical element" while the PS3 game has "a certain kind of softness without being blurry... [with] some warmth to it". I can see why PS3 would be more appealing to some. The upscale and resultant blur helps to make the game look a touch more movie-like; less rendered and less 'gamey' if you will - a good combination for a mainstream audience.
What is curious to me is that I can see no technical reason why the 360 game shouldn't just be a more detailed, smoother version of the PS3 version. Indeed, if the texture-dither filter could be turned off with a selectable option in a forthcoming patch, I'm almost certain that it would be the superior-looking game simply by virtue of the extra resolution and edge-filtering. But as it is, right now, there's not much in it.
As it is then, it comes down to personal preference, and in my view, Xbox 360 just 'edges' it. It seems that Rockstar didn't appear to issue a single PS3 screenshot pre-launch, which indicates that in-house the 360 game was the lead platform, and obviously the one chosen to best serve the PR effort. Anyone who's ever worked in the games business will know how carefully Rockstar controls its assets, so choosing 360 was a conscious decision. As an aside, it does seem quite ironic that in most cases, the PlayStation press would've been running previews and even reviews with Xbox 360 screenshots.
Regardless, away with such small talk and on with the video.
Alternatively, get the full picture with our 720p gallery of resolutions and anti-aliasing comparison shots.
Lighting, Draw Distance and Effects
This section of the feature mainly came about to prove or disprove a lot of the claims floating about online concerning apparent technical differences between the two games. As the vast majority of them were derived mostly from looking at screenshots other people had posted, or guesswork based on the advantages of the PS3's hard disk installation, I thought they were worth putting to the test.
Part of the reason this face-off has taken so long to put together is because actually matching the two games up is a really tough task. To get truly comparable shots, you really need to be capturing the action at the same time during the game's day-night cycle and more than that, you doubly need to make sure that the weather is identical too - not so easy when it appears to change in an arbitrary manner. Thankfully, the recently-released cheat codes allowed me to match up the weather identically before each capture was initiated, the only variable being the time taken to accomplish the same mission on both versions.
Time for a spot of myth-busting then. Aside from a slight tweak to the colour palette (going back once again to Houser's 'warmer' PS3 comments), the two games feature virtually identical lighting. Day and night cycles are basically the same, weather likewise. A case has also been made online with judiciously chosen screenshots that the PS3 version has better-realised explosion effects, but again, in controlled conditions this is proven not to be the case. In fact, the only lighting issue between the two versions that stands out is a shimmering on shadows on the 360 version. Noticeably improved on PS3, but hardly earth-shaking stuff.
To accompany the video are precision 720p shots of lighting and effects in the game.
Onto the next bone of contention then: draw distance. The theory is that the PS3's mandatory hard disk installation gives the Sony version an advantage here, but once again, the evidence clearly shows that the Xbox 360 game matches it. It's not difficult to see why. If the hard disk is good for anything, it's the swift streaming of texture detail, not the actual geometry.
Despite the hard disk advantage of PS3, I found it very hard indeed to show a tangible advantage in terms of texture streaming, aside from whatever wear and tear on the 360's DVD-ROM unit the game might inflict long-term. Maniacal flying of the helicopter showed obvious pop-in on 360 that the PS3 version coped better with, but in more common game conditions both versions acted in a very similar manner. In fact, both games infrequently exhibited pop-in textures, something you wouldn't expect from the PS3 code.
This video is backed up by 720p precision shots of good draw distance examples taken from the raw, uncompressed captures.
Both versions of the game have support for 1080i and 1080p monitors, even though the actual output is being scaled up from the base resolution (640p for PS3, 720p for Xbox 360). In the case of PlayStation 3, the effect only kicks in at all if 720p is not activated on the XMB, the implication being that Rockstar is happier with your screen handling the resizing work. Certainly, the lower resolution and (dare I say it) 'jagginess' are exaggerated in the scaling process, which isn't particularly good.
On the plus side, while software upscaling more often than not results in a drop in frame-rate, the refresh rate of the PlayStation 3 version of GTA IV is barely affected, regardless of whether 1080i or 1080p is selected.
We're on far more familiar ground with the Xbox 360 version of the game in that it's following the usual form of handing off the scaling duties to the ATI GPU. That being the case, it actually looks pretty good. While the texture dithering problem is still clearly apparent, the scaling does help to smooth off the effect a touch, meaning that the apparently 'clinical' look has had the edge taken off. More than that, the ATI chip works particularly well with an anti-aliased image to process, and as GTA IV supports that, the overall effect is fairly pleasing.
Also worth noting is that both versions have proper, non-letterboxed 4:3 support for standard definition screens, and 360 even has 5:4 compatibility for those running their consoles connected to non-widescreen LCD monitors.
You can see a few examples of how the game looks upscaled on both formats in our upscaled 1080p comparison gallery.
Frame Rate Tests
So far, aside from the different visual approaches, there's not much to tell the two games apart. But the one difference I couldn't help but notice was the sense that as well as possessing a higher resolution, the Xbox 360 version of GTA IV runs noticeably smoother than the PlayStation 3 code.
What is clear is that both versions adhere to no specific frame-rate. They'll pump out as many frames as they can, roughly averaging out to around 30fps (though it can go higher). However, the amount of time any given frame will stay on-screen is variable, leading to a very inconsistent look. Compare and contrast with, say, Project Gotham Racing 4, or Criterion's Black - both examples of games with a rock solid 30fps refresh rate. In terms of GTA IV, its basic inconsistency means it's really difficult to say whether one version drops more frames than the other just by trusting your own eyes. A more scientific approach is needed.
In previous face-off features, where I've felt that it's relevant, I've always mentioned the difference in refresh rate, but now I can show you how I do it. Essentially, the basic method is simple - capture every single frame that is output over the HDMI port in a lossless format, then examine the capture to check for the amount of unique frames. Digital Foundry's coder programmed a very useful little tool that does that count for me; useful considering the first test is based on a video with over 15,000 frames.
For GTA IV, I performed six tests on a varied amount of material. All of the game's cut-scenes are based on the same engine as the gameplay, so a variety of them were chosen, both indoors and outdoors, and of varying lengths. As the cut-scenes are rendering identical scenes on the fly, it's the best test I could come up with. In all cases, the respective 3D engines will be dealing with 100 percent identical source material.
Links to compressed versions of the test material are provided, but please bear in mind that there's some pretty hardcore swearing. Oh, and possibly some spoilers.
Test One: Game Intro
See it on EGTV.
Test Two: Clean Getaway
See it on EGTV.
Test Three: Final Destination
See it on EGTV.
Test Four: Station Face-Off
See it on EGTV.
Test Five: Rigged to Blow
See it on EGTV.
Test Six: Ivan the Not So Terrible
See it on EGTV.
These figures have been updated slightly since they were originally published. Head over to the Editor's blog to find out why.
So the results clearly show that over the course of the entire clip, 360 out-performs PlayStation 3 in every one of the six scenarios presented here. Indeed, on the longer vids, with a larger average sample, we're seeing a good 17 to 18 percent variance. Tests on gameplay (playing through the same mission, but not rendering identical scenes, obviously) saw a similar range of variance too. For example, the 'Ivan the Not So Terrible' stage has a nice range of in-car, on-foot, rooftop and cut-scene action. 30.106fps average on a 360 runthrough, compared with 26.733fps and 26.696fps on two separate PS3 captures of the same mission.
The bottom line is that no matter what material I put through the detector, 360 came ahead in all tests, sometimes dramatically so.
However, it is worth bearing in mind that the frame-rate detector can only produce an average rate. By looking at the other comparison videos in this piece, you'll see that by and large the games generally run at the same speed. However, when the engine struggles to cope, the frame-rate falls and it drops harder on PS3, hence bringing down the overall average.
One curiosity I found was that the PS3 version has v-lock enabled, whereas 360 can produce the odd torn frame. This might be seen to skew results in 360's favour where it not for the fact that typically, a torn frame is only on-screen for 1/60th of a second before v-sync is re-established. More than that, closer examination of the clips reveals that the game actually swaps between two different v-lock modes, only occasionally losing sync. And when it does, in the vast majority of cases, it's a tiny effect evident only on the very top or very bottom of the screen - mostly unseen in the overscan area. More than that, our tests showed that of the entire 60fps output of the 360, the amount of torn frames averaged between 2.35 percent and 6.07 percent. In short then, really not a problem.
Therefore, it's certainly not an issue during gameplay, and regardless, the figures above are compensated anyway to give the true frame-rate. The hosted clips are from the 360 version by the way, in case you want to check.
Truth be told, it's not really the visual differences as such that gravitates me slightly towards the 360 version; it's just that the game runs more solidly on the Microsoft platform, and when the frame-rate does drop, it's not quite so jarring as it is on PS3.
After almost two thousands words of technical discussion, the bottom line is that it's clear that Rockstar had some issues matching the basic performance of the Xbox 360 game on the PS3 hardware. Lower resolution, zero anti-aliasing support and a more variable frame-rate are the bottom line.
On the one hand, it's a touch disappointing that Rockstar's USD 100 million budget couldn't extend to optimising the experience to match Xbox 360, especially when you look at a game like Burnout Paradise that doesn't require a mandatory installation, has a basically rock solid frame-rate, and is technically identical cross-platform.
That said, it's patently clear that Rockstar hasn't handed in a lazy conversion here. Creative decisions have been made to compensate for the technical limitations, and by and large they really work, to the point where you sometimes wonder why they couldn't have been applied to the Xbox 360 version too. PS3 GTA IV looks absolutely fantastic, and even factoring in zero anti-aliasing support and a lower resolution, in many scenarios it looks as good as the 360 version, if not better.
I've not addressed the freezing issue being reported on the game, of course. I'll have to leave that to the people affected as - try as I might to coax the code to fall over and die horribly - GTA IV behaved impeccably for me on both systems. But the chances are that by the time you read, a PS3 patch will be out and about that should sort out those who've been affected. A 360 update after that, please, complete with a tweakable option to use the PS3 post-processing modes. Now that would be interesting.