Xbox One Resolutiongate: the 720p fallout
How Call of Duty and Battlefield changed the console war.
"The biggest thing in terms of the number of compute units, that's been something that's been very easy to focus on. It's like, hey, let's count up the number of CUs, count up the gigaflops and declare the winner based on that. My take on it is that when you buy a graphics card, do you go by the specs or do you actually run some benchmarks? Firstly though, we don't have any games out. You can't see the games. When you see the games you'll be saying, 'What is the performance difference between them?' The games are the benchmarks." - Microsoft technical fellow, Andrew Goossen.
We couldn't agree more. Despite the myriad gaffes and U-turns, Microsoft always had the right to make the case for its technology and to address its critics in the face of overwhelming criticism of its hardware design decisions. However, in the final analysis, the games are clearly the benchmarks that matter - and in terms of multi-platform offerings, relative system capabilities are a key battleground that Microsoft chose to mire itself in, with damaging results.
As the metrics emerge on key next-gen launch titles, it's clear that Xbox One is under-performing against its rival - not just according to the spec differential, but actually beyond the difference in raw numbers. Our Battlefield 4 Face-Off preview reveals a 50 per cent resolution boost on PlayStation 4 with no appreciable compromise in effects or performance in single-player gameplay, while Infinity Ward's Mark Rubin confirmed rumours that Call of Duty: Ghosts runs at native 720p on Xbox One, with 1080p a lock for PS4. Assuming uniform features and performance, that's a massive blow for Microsoft.
While Digital Foundry has yet to see either next-gen version of Call of Duty, our experience with Battlefield 4 demonstrates that you can easily see the visual difference between them. The Xbox One version holds up well given the gulf in resolution, but it doesn't require a pixel counter to tell that the PS4 game is crisper and cleaner either. At last week's Battlefield 4 review event in Stockholm, we noted that the resolution change from one version to the next was obvious to many of the press in attendance, with some even suggesting on-site that the PS4 version was operating at native 1080p when its actual resolution was 1600x900. Battlefield 4 is a beautiful game generally, but if it has one Achilles' heel common to both next-gen platforms, it is the pixel-crawl and sub-pixel break-up derived from the post-AA technique. Xbox One has bigger pixels and fewer of them, so naturally the most obtrusive element of the presentation is more of an issue when displayed on the same screen.
The reality for Microsoft is that the raw spec differential it has battled against is not only borne out in what is arguably the most technologically advanced multi-platform game of the next-gen launch, but the gulf actually increases on a title that, on the face of it, isn't pushing boundaries to anything like the same degree. Mark Rubin has previously suggested that there is no new Infinity Ward engine for the cross-generational Ghosts - rather that the studio has continued to build upon the existing tech. The situation is interesting in that we have a piece of technology that almost always favoured Microsoft's current-generation hardware now performing in a vastly superior manner on the competing platform in the next-gen era. It's a stunning turnaround.
"As the metrics emerge on key next-gen launch titles, it's clear that Xbox One is under-performing against its rival - not just according to spec, but actually beyond the theoretical teraflop differential."
So what went wrong? The irony is that despite the battering it has received from both the Battlefield 4 and Call of Duty code, Microsoft is right in claiming that - in isolation - more compute power doesn't produce a corresponding level of performance, borne out by virtually any AMD GCN graphics card comparison you care to run in the PC space. In our In Theory piece on flops vs. frame-rate, we replicated XO and PS4 compute performance with PC parts, and saw the 50 per cent compute advantage whittled down to anything from 19 to 33 per cent, depending on the game.
Whether the performance boost is 19 per cent, 33 per cent or any number in between, that is still a tangible PlayStation 4 advantage, but not enough on its own to explain the BF4 and Ghosts differentials - the obvious conclusion being that raw compute power is just one part of the equation. Developers have faced a procession of issues with Xbox One GPU performance in the run-up to launch - some that are being tackled in the short term, others that will prove more difficult to address. Microsoft's "mono driver" for the AMD GPU had been known for months to be delivering sub-par performance prior to Gamescom in August (hence disappointing Ryse and Dead Rising 3 showings at E3) and while improvements have been - and apparently continue to be - delivered, developers have been working around a moving target, unsure exactly what the power of the graphics hardware will be in the final retail box.
Conversely, by that point, Ubisoft Reflections had already confirmed to us during its June presentation on porting The Crew to PS4 that GPU performance on the Sony platform was looking relatively solid, a state of affairs that presumably produced a stronger foundation for optimisation.
"The SDK is changing all the time, [but] it's changing less quickly than it was six months ago," Ubisoft expert programmer Dr. Chris Jenner told us. "We're getting near to the final state, we're not expecting huge performance changes, just finalisation of features. It's a lot more stable than it was early on. We haven't had to do any changes for a while."
"Kinect functionality and other features take up ten per cent of Xbox One GPU resources that developers can't access right now, though plans are afoot to change this."
Driver revisions are clearly one issue, another is the remarkable ten per cent of GPU time reserved by the Xbox One operating system for functions like Kinect skeletal tracking, accounting for precious resources that are inaccessible to game developers. Again, Microsoft is looking to free up that GPU power, but that is clearly no help for developers in the launch period. While OS GPU time for PS4 remains an unknown, it has not been flagged as an issue thus far.
However, the hardware make-up itself could be more troublesome for multi-platform developers in the longer term, despite Microsoft's outline of how the Xbox One tech operates and the theoretical advantages it chose to highlight. In our In Theory piece, we could only address the teraflop difference - we couldn't measure the impact of Xbox One's reduction in memory bandwidth, and we certainly couldn't factor in what was then the big unknown: the controversial 32MB of Embedded Static RAM (ESRAM) built into the Xbox One's central processor.
In the Xbox One Architects interview, Microsoft positioned the Xbox One silicon as a natural successor to Xbox 360, with ESRAM defended as the most power-efficient, cost-effective solution for delivering 8GB of RAM in a console product - an evolution of the eDRAM that had served it so well in the current generation. We hear different stories about ESRAM from virtually every source we speak to, but two gripes are common. Firstly, the notion of operating between two memory pools for render targets is an additional pain that is not an issue on PlayStation 4's unified 8GB of GDDR5. Secondly - and perhaps most importantly - the most common compliant we hear is that developers really want more than 32MB for their high-bandwidth graphics work.
Talking to the Microsoft tech staff, we attempted to tackle the issue of the 32MB ESRAM limit by suggesting a 1080p render target scenario that wasn't that outrageous for a modern game engine, but would bust through the memory ceiling very easily. Microsoft countered by suggesting that those targets could be split between DDR3 and ESRAM, and pushed its own, more memory-efficient, compressed render target formats - similar to the ones utilised on Xbox 360 to great success.
"If studios like DICE and Infinity Ward - whose resources are vast - are having problems with Xbox One, the prognosis can't look good in the short term, though it stands more of a fighting chance with 30fps titles."
In the light of recent events, the question is, will those formats actually be utilised if they can't be easily supported on PC or PlayStation 4? More pertinently, faced with crushing deadlines for next-gen launch titles, isn't lowering the size of Xbox One render targets the much easier option?
So, will things continue to look grim for Xbox One multi-platform games in the launch period? On the one hand, if studios like DICE and Infinity Ward - where resources are vast - are having problems, the prognosis can't look good in the short term. What hasn't helped Microsoft's cause is that both of the key titles where we have known metrics are pushing 60 frames per second - the bleeding edge of performance on console, which rarely works out kindly where one platform has weaker GPU performance, as many of the current-gen Call of Duty titles demonstrate.
After six years of producing Face-Offs, it's safe to say that the 30fps cap we see on many multi-platform titles often acts as a great leveller in the pursuit of platform parity. We say that with some confidence as some 30fps titles can be unlocked, giving an idea of actual top-end performance on each platform. Running BioShock Infinite in unlocked mode, we see a frame-rate advantage for PlayStation 3 (helped in part by its lower resolution) over Xbox 360. Conversely, Batman: Arkham City also runs unlocked by engaging TriOviz 3D mode. There we see frame-rates soar far north of 30fps in many scenarios, but almost always with a clear Xbox 360 lead. All things being equal, a 30fps cap perhaps working in combination with a drop to 900p on Xbox One sounds like a viable approach, providing a level of platform parity closer to the situation we see on current-gen consoles. And indeed, there's much talk now from execs about games on Xbox One "looking great" without referencing image quality or native resolution directly.
Game developers have utilised varying resolutions over the current-gen period and we would venture to suggest that at a commercial level, it has had little impact on game sales. But the next-gen launch is clearly a very different situation - consoles are at their most expensive, gamers want the best deal, and if they have invested in a 1080p display, why wouldn't they want to get the most out of it? Speaking to Guerrilla Games' MD Hermen Hulst in Amsterdam last week, he believed that Killzone fans would not take too kindly to a non-native 1080p presentation. There are areas in Killzone Shadow Fall - particularly in terms of the game's state-of-the-art lighting and material detail, where the case for targeting full HD simply cannot be challenged from an image quality perspective - the results are simply stunning. At the other extreme, 720p is so closely associated with the current-gen standard by core gamers looking for that next-generational leap, that its association with Xbox One on major triple-A games does the console no favours.
"While Microsoft is losing the multi-platform argument, the allure of Forza Motorsport 5 is difficult to ignore."
But pixel counts aside, perhaps the most challenging issue Microsoft faces - and the same one faced by Sony across this generation - is of the more expensive console offering the sub-optimal experience in key titles. In the here and now, few game franchises are more important than Battlefield and Call of Duty. PlayStation 3's Blu-ray drive added significantly to the bill of materials back in 2006/2007, while the next-gen Xbox One comes with Kinect - an expensive addition that failed to gain traction with the core in the current generation. Looking forward to the next, Microsoft and third-party developers appear to be doing little to champion the capabilities of its successor, despite the sheer wealth of system resources dedicated to the technology.
Back in February of this year, in the era before Xbox One and PlayStation 4 were even announced, we knew these consoles only by their codenames - Durango and Orbis - and through their leaked specifications. Our assessment of the machine that was to become PlayStation 4 - even before its 8GB upgrade - was of a "tighter, more powerful, more games-focused design". In parallel, we summed up the precursor to Xbox One as a machine designed to be broader in nature, encompassing media and games - pretty much exactly how the console was presented at its reveal some months later, and a message that failed to resonate with many of the core audience. Yet some might argue that the thinking behind the approach had merit in speaking to a broader audience - Microsoft had seen Xbox 360 receive extensive utilisation as a media playback device, the role of console hardware had adapted across the generation, and Xbox One is, in many ways, a logical response to that.
Since then, Microsoft has reversed many of its setbacks and concentrated on gaming first and foremost - the "TVTVTV" focus in the messaging has all but gone and the platform-exclusive line-up is winning over gamers far more, we suspect, than its attempts to compete with PS4 on platform parity. However, in directly addressing the specs differential, whether in the Xbox One architects interview or in online forum posts, it has seemingly set itself up for another own goal. Microsoft itself has made the story about parity with the competition, when highlighting what makes Xbox One unique in terms of exclusive games, services and functionality - along with more effort in returning some of the magic to Kinect - may have served Xbox One more effectively in the run-up to launch.
Digital Foundry specialises in technical analysis of gaming hardware and software, using state-of-the-art capture systems and bespoke software to show you how well games and hardware run, visualising precisely what they're capable of. In order to show you what 4K gaming actually looks like we needed to build our own platform to supply high quality 4K video for offline viewing. So we did.
Our videos are multi-gigabyte files and we've chosen a high quality provider to ensure fast downloads. However, that bandwidth isn't free and so we charge a small monthly subscription fee of £4.50. We think it's a small price to pay for unlimited access to top-tier quality encodes of our content. Thank you.Support Digital Foundry