Valve's Rube Goldberg Machine

Obviously I have to write about the deckard right now, because while I have many ideas for blog posts in the future, the deckard has become an increasingly time sensitive subject.

We are currently mere weeks away from a deckard announcement and as such I have decided that right now would be the best time to encapsulate the state of affairs.

The Hardware

Let's establish common understanding of what "the deckard" is before proceeding: The deckard is the codename for Valve's highly anticipated stand-alone VR headset, the final product name which is believed to be "The Steam Frame" based on a recently published US trademark. As a stand-alone headset it, as expected, is powered by an Arm-based SoC (System on a Chip), specifically it's believed to be the SM8650, more widely known as the Snapdragon 8 Gen 3 chipset from Qualcomm. This makes for an odd choice, as this SoC is both not a dedicated XR chipset (akin to qualcomm's XR Gen X series) and also not a terribly recent piece of silicon either. The headset's display tech has not yet been confirmed, but if it truly will be going for ~1200 dollars then, barring any subsidizing from Valve's end, it seems likely that we're not looking at some terribly new or high-end display technology - as a prediction based on pure hearsay to base the rest of this blog post off of, I'm going to predict it's housing a 120hz LCD at a squared 2160 pixels Deckard also comes with a set of controllers codenamed roy, these we have all the details on you could possibly want as their 3d models have existed within public steamVR builds for a while and seen hcontinous updates to reflect their latest revisions as time has gone on.

The double-A powered roy controllers' feature a very traditional set of inputs that mirror the legacy console ABXY-controller layout and adds on top the inputs expected of a vr controller like grips and capacitive touch. This set of inputs support the rumors of deckard's intended use cases; as both an XR and 2d flatscreen game consumption device, with a focus on allowing the user to create virtual environments to enjoy their entire steam library within - potentially supporting integration of social features as well hooking into steam's already present and matured game streaming feature-set.

Lastly there's some contention sorrounding the tracking methods of the headset, while there's mutual agreement that it will support some form of inside-out tracking, there's also been strings found in steamVR builds that suggest the headset itself will be capable of functioning as some sort of lighthouse to enable tracking of older lighthouse-tracking-based accessories made for users who have been hanging onto their HTC vive or index devices.

With a few exceptions this doesn't seem like all that exceptional of a device, which begs the question: What makes this thing worth a blog post anymore than an HTC focus vision or a Meta Quest 3? It's just another snapdragon powered VR headset with a set of battery powered wand-shaped controllers that feature a bog-standard LCD running Android.

Wait- that's not right, hold on... -I'm sorry, its running Linux!?

The Software

This is where a large part of the deckard's intrigue stems from. Running on Linux, it's software stack is composed of several well known foss solutions (as well as a few proprietary garnishes from Valve) that together must create a palletable user experience. Anyone familiar with Linux, XR and Arm-powered devices will tell you, that is a very difficult goal to achieve. The right software in the wrong place can make all the difference in the world and so Valve has to move some small mountains to develop the right kind of compatibility layers for an environment as wrong as linux that will make it a true "console-like experience".

The Steam Deck was the culmination of years upon years of work between internal Valve devs and many, many contractors to develop a foundation upon which Windows games could be somewhat reliably played on Linux, this effort needs continous financial support in order to keep up with current day releases and as such Valve has been required to keep up the pace to maintain the value of their handheld devices and by extension steamOS.

The challenge faced by Valve for the deckard however is two-fold: they need to make sure that Arm-based linux XR works at all which is a goal that cannot just be taken for granted - simultaneously - They need to make sure that Linux on Arm can run regular x86 based Windows games. The solution to the latter is the combination of two Valve-backed FOSS projects; Wine/Proton and fex-emu, while Proton has had the conundrum of playing Windows games on Linux solved for years now, fex-emu, the emulation layer that allows software compiled to a generic x86 target to run on Arm-based processors, is still rather young and doesn't inspire confidence that Valve will be able to pull off it's flat screen ambitions without hiccups. Yours truly has collected a few Arm-based devices capable of running linux over the years and can only claim that right now, while the experience has improved at an impressive rate, it absolutely cannot be considered consumer ready in any regard.

Crashes are all too common, there's an inconsistency in frametimes that can be blamed on the sheer number of slowdown causes: whether its the game being particularly CPU intensive in a certain spot, something unoptimized about the emulator or a certain loop of instructions that are inefficient to translate over to Arm - the experience will absolutely take you out of any game (hope those steamVR environments keep you immersed!). Certain games will barely or just not work at all because they utilize a problematic library and/or framework, an example being S&box which utilises dotnet curtesy of it's dependency on C#.

Waydroid is another FOSS project Valve plans to utilize for the deckard, it is an Android Application translation layer to Linux (Wayland) similar to Wine for Windows Applications, in fact it seems to be the analouge of Proton's purpose in the Steam Deck. The Waydroid project currently, does not support any XR components, but low and behold there have been publicly inaccessible Waydroid Steam packages spotted curtesy of steamDB, they even feature custom branding. A big part of Valve's strategy to solve the XR software problem for the deckard seems to be relying on Android's vast XR library developed by, what ostensibly serves to be their competition, Meta and HTC, who have been developing standalone Android devices for years and convincing developers to make the jump. Obviously Valve doesn't intend on you running those apps intended for other headsets in their exact form, but by asking devs to make vendor agnostic versions of their already existing Android XR apps, they're dramatically lowering the effort required from them to invest into their platform.

Once again we do not have anything on this supposed "Steam Edition" of Waydroid, all we have are references to the packages through steamDB which is not a lot; some updates and lists of related packages that give us slight insight into pace of development and types of tests being ran - which isn't much of anything at this point in time.

But perhaps one of the things with the widest impact to come out of the deckard software-wise upon announcement/release is the Linux Arm64 build of Steam and its associated software packages. There's recently been a growing number of projects that have sprouted in the Android world trying to run bog-standard windows games through a combination of termux, wine and fex-emu or box64. They've become quite popular as the standard performance expected out of your typical mobile chipset has increased as well as the rise in popularity of Android handhelds with dedicated cooling. As such the benefits to the android world of "PC gaming" that an Arm native steam client with all the bells and whistles could bring speak for itself.

Which bring us to the now - currently the deckard has not yet been announced by Valve, it has not explicitly been acknowledged nor have we seen the HMD in any form apart from patent drawings of questionable validity - this also plays a large part in the intrigue of the deckard. Because with most devices made by valuable companies known by hundreds of millions around the world, you get leaks. You get real leaks by humans in the loop. You get insiders swapping spit, giving info and sharing pictures. You get people with devices in hands early, employees with preview units leaving models laying around that get accidentally recorded.

Yet this time around we have not seen this thing in it's physical form, we have no actual proof that it has ever been manufactured before, only word of mouth - there's an alternate universe in which this thing happens to be an elaborate hoax or long cancelled or misunderstood to be a completely different type of product. Of course - we have key pieces of information that disprove it's inexistance, but that lingering feeling that it still somehow may not be real does not stray far from the current reality - that is what has made the deckard achieve "faith level status" as one supply chain insider excellently put it.

The deckard codename has persisted for years and nothing has been shown, it's made by Valve who are the end-all-and-be-all of cancelling long-wanted and -running projects so there's a level of doubt that has slowly persisted and been gradually replaced with increasing levels coping with the fact that it has not yet materialised.

All this to say that a lot is riding on the shoulders of this product, the chain of components that go into this stack, orchestrated by a team of merely ~350 people, has to serve an estimated initial costumer base of 500,000 - it's a little terrifying to think mainstream reviewers have to figure out how to talk about the deckard and give something, which fundamentally differs in a myriad of ways from it's perceived contemporaries, a rating which the general audience will inevitably compare and contrast with other headsets on the same axis as if they are interchangeable in functionality.