The leak of unredacted Microsoft documents lays bare the firm’s plans for console hardware for this year and beyond. We’re getting a new, impressive-looking controller, mid-generation refreshes of the existing Xbox Series consoles, and we have a strong idea of Microsoft’s thinking for the next generation. There are some controversial ideas presented in terms of its revised Series machines but I’m more concerned about the plans for the Gen 10 consoles. Xbox needs to be radical, it needs to align more closely with the best in PC innovation, and the vision presented is either problematic and unambitious.
Before going any further though, Phil Spencer has already said that these documents are out of date and that we’ve not yet seen the ‘real plan’. On the one hand, it makes you wonder why Microsoft would present irrelevant documents to the FTC if its strategy is actually quite different. On the other hand, there is precedent for Microsoft leaks turning out to be very much out of date, bearing little resemblance to the final product – read up on the old Fortaleza leak if you need your memory jogged. This was allegedly a vision for Xbox One, except it bore little resemblance to the final product. Kinect glasses, anyone?
The revised Series consoles in the leaked PDFs are another matter entirely. The chances are that this is exactly what we’re going to get since console design and establishing supply lines can’t be rushed or massively tweaked this far along in the production process, even a year out from launch. Xbox Series S effectively remains unchanged on the outside, bar improvements in sustainable materials in its manufacture. Inside, storage doubles up to 1TB, while the main processor is shrunk to TSMC’s 6nm node from the current 7nm process. The Xbox One era Southbridge retained for Series hardware is replaced while wireless I/O gets a nice upgrade.
- 0:00:00 Introduction
- 0:02:48 The end of physical media?
- 0:14:55 Revised Series X, Series S consoles incoming
- 0:36:02 Updated Xbox controller in the works
- 0:41:51 Microsoft’s next-gen console proposal
- 1:04:53 Supporter Q1: Can Sony really pull off a cost-effective, powerful PS5 Pro console to put Microsoft under pressure?
- 1:11:02 Supporter Q2: Could Microsoft scrap some of their console plans because of this leak?
- 1:14:50 Supporter Q3: Which leak is bigger, the Xbox FTC leak or ‘Jensen’s Prophecy’?
Xbox Series X gets a more thorough revamp. It receives the spec enhancements of the Series S refresh, it gets the 6nm silicon and a doubling of storage, but it’s also losing the optical drive, bringing the end of physical media one step closer. This is another topic altogether, but if this is the route Microsoft wants to take, it needs to come up with a way to provide a competitive marketplace for buying console games from different vendors, if only to head off the antitrust/regulatory nightmares that will inevitably follow. Still, the lack of an optical drive prompts Microsoft to re-style the unit with a Mac Pro-like ‘trash can’ design, making this the first Xbox that isn’t actually a box.
The design validates comments made by ex-Xbox chief architect Andrew Goossen back in 2020 when we met him at the Microsoft campus
That’s if it can, of course. The fact that Microsoft is using 6nm technology and not the 5nm or 4nm processes used by the latest PC components is telling. It suggests that the cost and yields of this more cutting-edge silicon are uneconomical, while the ‘same great price’ messaging suggests that three years on from launch, Xbox consoles are still very expensive to make. I’m going to be interested to see how Microsoft markets these boxes: a console refresh is typically a great impetus to renew interest in the hardware and to cut prices. However, based on what’s shown here, it’s difficult to find much excitement. The only tangible gain to the consumer is more storage, while the removal of features from a console – the optical drive on Series X – is not a great look.
- 00:00:00 Introduction
- 00:01:10 When did the DLSS 3.5 Ray Reconstruction project start and why?
- 00:04:16 How did you get DLSS 3.5 Ray Reconstruction up and running?
- 00:06:17 What was it like to integrate DLSS 3.5 for Cyberpunk 2077?
- 00:10:21 What are the new game inputs for DLSS 3.5?
- 00:11:25 Can DLSS 3.5 be used for hybrid ray tracing titles and not just path traced ones?
- 00:12:41 What is the target performance budget for DLSS 3.5?
- 00:14:10 Is DLSS a crutch for bad performance optimisation in PC games?
- 00:20:19 What makes machine learning specifically useful for denoising?
- 00:24:00 Why is DLSS naming kind of confusing?
- 00:27:03 What did the new denoising enable for Cyberpunk 2077’s graphical vision?
- 00:32:10 Will Nvidia still focus on performance without DLSS at native resolutions?
- 00:38:26 What prompted the change internally at Nvidia to move away from DLSS 1.0 and pursue DLSS 2.0?
- 00:43:43 What do you think about DLSS mods for games that lack DLSS?
- 00:49:52 Where can machine learning go in the future for games beyond DLSS 3.5?
Also included in the PDFs are Microsoft’s tentative plans to attain market leadership in gaming by 2030, along with initial thoughts on the design of its Gen 10 console. What’s interesting here are the timelines, which again emphasise how long it takes to create a console. Based on the docs, Microsoft should have already committed to what kind of silicon it’s going to make, from a choice of ARM or x64 CPU cores to the decision to co-design a GPU with AMD, or else dip into the AMD parts bin and use their upcoming graphics hardware – Navi 5 is mooted (we’re currently on Navi 3 in the PC space).
There are two key elements to Microsoft’s next-gen plans and neither are particularly inspiring. The first is the concept of ‘cohesive hybrid compute’, the idea being to interface cloud hardware with the client much more closely to create experiences we’ve never seen before. Remember ‘the power of the cloud’? It’s back. Microsoft never made it happen back in the day (the Crackdown cloud concept never came to pass) but I see more practical limitations. Starfield launched with hundreds of thousands of concurrent users on Steam alone. Can Azure really spin up that many cloud instances to ensure that games actually work? Secondly, the level of bandwidth required to fully integrate cloud and client hardware remains onerous to say the least, while latency will always be an issue. Internet connections have improved but not to the extent that everyone everywhere (or even a majority?) could get a good experience.
In short, almost everything we said about this concept back in 2013 remains the same today. Practical applications for combining the cloud and client hardware are limited and if the concept of integration cannot guarantee you access to your games whenever you want to play them, it’s a non-starter as the basis of a vision for an entire console generation. Maybe there’s some kind of nuance or context I’m not aware of, and there’s every chance I’ve got this one wrong, of course. Perhaps Xbox’s collaboration with Hideo Kojima will be the proof of concept we really need.
Microsoft’s second objective is to fully leverage machine learning with the addition of an NPU, a neural processor. It’s essential to the make-up of a new console precisely because we are already seeing many of Microsoft’s ‘next-gen’ ideas play out in the PC space. Microsoft talks about super resolution, frame generation and xCloud latency compensation, while mooting the possibility of AI-powered NPCs. For me, the concern here is the lack of vision. This is Microsoft’s roadmap for the next generation that’ll carry through into the 2030s, when all of those features would have been on Nvidia’s roadmap in 2018 or even earlier and are with PC users right now. Releasing a console in 2028 with today’s PC tech signals that console hardware will always be a step behind and I’d argue that Microsoft needs to be at the forefront on both PC and Xbox. For me, it’s the obvious nature of this vision that is concerning. Of course we want all the DLSS stuff – but what’s next?
The design choices presented in the document for the new consoles are arguably too limited. First of all, in creating an AMD-based console, Microsoft once again puts itself in the very real, very dangerous position of creating hardware that has much – perhaps too much – in common with whatever Sony is going to come up with. The last generation spectacularly demonstrated that if you have a set budget for a console and you go to the same manufacturer for its core components, you’re going to end up with much the same hardware. Microsoft needs to be more radical and more innovative – and there’s no shortage of that in the PC space. The thing is, it’s not AMD delivering that innovation. As the release of Cyberpunk 2077 2.0 demonstrates, it’s literally years behind the state of the art.
So, how can Microsoft differentiate itself from PlayStation, while at the same time bringing Xbox consoles closer into line with the PC space? A collaboration with Nvidia is the obvious choice. Everything Microsoft wants to achieve with ML in its 2028 console is already happening on Nvidia hardware right now. Aligning with Nvidia would also open the door to the firm’s continuing innovations in the PC space to migrate more quickly and more gracefully to the console space. If innovation is being driven by Nvidia and around 80 percent or higher of gaming PCs already run on Nvidia hardware, a partnership there seems like a better bet.
Maybe there are political problems that prevent this from happening – issues between Microsoft and Nvidia dating back to the OG Xbox days have not been forgotten. Perhaps there are more practical issues, specifically that an Nvidia console processor could not use x86/x64 CPU cores, creating obvious back compat problems – though it’s interesting to note that only forward compatibility is mentioned in the PDFs and Microsoft is considering ARM already. Maybe there are silicon issues: Nvidia’s integration of RT and ML cores directly into the GPU is more expensive in terms of die area than AMD.
On that latter point, this hasn’t dissuaded Nintendo from embracing Nvidia architecture for its Switch successor, even though AMD is now putting out some extremely impressive mobile silicon. Of course, maybe Microsoft knows something we don’t and is banking on Navi 5 technology to catch up with Nvidia, but even then, I’d argue that innovation is driven not just by hardware but software too. As our recent roundtable with Nvidia and CD Projekt RED (embedded above) demonstrates, DLSS 3.5 is as much about software engineering as it is about hardware design.
Back in the day, it was the consoles that offered game-changing, exotic custom silicon, delivering experiences PC couldn’t match. Today, the situation has reversed and presenting what PC is doing today along with the ‘power of the cloud’ doesn’t excite me as a compelling vision for an Xbox console due in 2028. Maybe we’re missing nuance and context, maybe this was just one potential roadmap from a range of options. Hopefully, as Phil Spencer says, these are not the real plans.