During a Steam Deck hands-on event, Valve was asked about the possibility of seeing Steam Deck’s custom AMD processor in a standalone VR headset. While the company didn’t confirm anything outright, they said the Steam Deck hardware could certainly be a good match for a standalone VR headset.

The Verge‘s Sean Hollister published an extensive hands-on with Steam Deck today, saying he’s “nearly a believer” in Valve’s vision for a portable gaming device built with PC hardware. Like many of us, Hollister is curious whether or not Valve’s work in creating a portable could open the door to a standalone VR headset from the company.

Hollister asked if Steam Deck’s custom APU could work in a standalone VR headset, and writes that 24-year Valve veteran Greg Coomer ‘lit up’ at the suggestion, saying he “loved the question.”

“We’re not ready to say anything about [a standalone VR headset], but [Steam Deck’s hardware] would run well in that environment, with the TDP necessary… it’s very relevant to us and our future plans,” Coomer said.

TDP stand for Thermal Design Power which describes how much heat a processing system generates under load and how much power it consumes. Portable devices like phones need a low TDP so they don’t overheat or consume battery too quickly. This is especially important for standalone VR headsets because of the high performance demands balanced against the need to prevent overheating of a worn device… not to mention avoiding too much added weight from a battery.

Coomer’s response indicates that he thinks the AMD APU in Steam Deck—which ranges between 4 and 15 watts—could reasonably scale to the power, performance, and heat needs of a standalone VR headset.

SEE ALSO
Meta Explains Why It Sees Wide Field-of-View Headsets as a 'bad tradeoff'

In total isolation, we might be able to reasonably brush off Coomer’s comment as a fun “what if.” But there’s plenty of evidence that Valve has been investigating standalone VR headsets.

Earlier this year we reported on new patents from Valve that envision wireless and standalone versions of its Index headset, including novel head-mount designs to aid in cooling and weight balance.

It’s important to set expectations however. Steam Deck’s hardware is powerful, but it isn’t magical. While Valve says it can run AAA PC games “really well,” they’re talking about running them at 1,280 × 800 (1MP) and no higher than 60Hz. Contrast that with Quest 2 which runs VR games at 3,664 × 1,920 (7MP), in 3D, and at a minimum of 72Hz, along with additional processing overhead dedicated to running the headset’s tracking and other functions. The only way Quest 2 achieves this is that all the games are built with far less graphical fidelity than what you’d find on a PC.

That is to say: if Valve built the Steam Deck hardware into a standalone VR headset, you won’t be running Half-Life: Alyx on it without completely gutting the graphics to the point that it looks much closer to a Quest 2 game. Granted, developers have shown that games made from the ground up for this kind of low-powered hardware can indeed look great, as long as they aren’t aiming for a highly realistic look.

Even so, it would be great for the entire industry to have Valve enter the standalone VR space which is currently dominated by Facebook. Having a serious alternative on the market would bring much needed choice and competition.

SEE ALSO
'Metro Awakening' to Release on All Major Headsets in November, New Gameplay Revealed

Valve could easily build a standalone VR headset with the same Qualcomm Snapdragon hardware that Facebook uses for Quest 2, but that wouldn’t make much sense for the company because Snapdragon is a different kind of CPU that can’t run PC applications—which is the bread and butter of Valve’s entire business. If Valve went that route it would mean fragmenting Steam into separate PC and mobile versions, with different games on each, which it has little incentive to do (especially now that it’s launching Steam Deck).

Now that Valve has proven it can build a portable gaming device with PC hardware, the door suddenly seems much more open to the possibility of a standalone VR headset from the company. If they took this path, they could continue to sell games right from the PC version of Steam (though they’d probably need to have a special section of the store reserved just for VR games that have been optimized with their standalone VR headset in mind).

Newsletter graphic

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. More information.


Ben is the world's most senior professional analyst solely dedicated to the XR industry, having founded Road to VR in 2011—a year before the Oculus Kickstarter sparked a resurgence that led to the modern XR landscape. He has authored more than 3,000 articles chronicling the evolution of the XR industry over more than a decade. With that unique perspective, Ben has been consistently recognized as one of the most influential voices in XR, giving keynotes and joining panel and podcast discussions at key industry events. He is a self-described "journalist and analyst, not evangelist."
  • Ajedi32

    They’d need something a bit beefier than the Steam Deck to run VR games at acceptable resolutions/framerates on PC hardware. Might be weight issues with trying to pack all that into a headset too.

    If they’re going to enter this space, personally I hope they go with a more modular approach; with the computer/battery mounted on your belt and connected to the headset with a wire. That way they could even sell the headset as a PCVR headset, while still having the headset/comput box bundle available for those who want a more standalone-like experience.

    This also raises questions about how they’d handle tracking. Inside-out is best for standalone. Might Valve be working on a headset with inside-out tracking? Or perhaps they plan to stick with lighthouse and offer some sort of portable battery-powered base stations?

    Lots of interesting possibilities here.

    • kontis

      I think they already said something in the past about adopting self contained computer vision solution if it becomes a better option for them. (Lighthouse is technically also inside out).

      Hopefully, by that time the webcam based full body tracking will be just as good so losing trackers won’t feel that bad for the enthusiasts.

      • wheeler

        Check this, it looks like they’re working on an inside out tracking system (in the average VR consumer’s “no boxes on your walls” sense of “inside out”) that optionally functions with lighthouse. At a base level they have a camera on the headset to track the headset’s position within the play space using CV, and then a new kind of “angle sensitive sensor” for the headset and other tracked devices to mutually track each other via an array of IR LEDs. They describe the system as being interoperable with basestations, which suggests these new sensors could also work with lighthouse.

        Putting it all together, this would let them provide a cheaper and more convenient baseline while continuing to support and remain backwards compatible lighthouse for those that want its benefits. https://www.freepatentsonline.com/y2021/0231962.html

    • MosBen

      I mean, the Steam Deck is going to be rolling out over the next year, and they haven’t even announced anything for VR. So it’s not crazy to think that they might be working on something that’s targeting the next generation of integrated APUs, which will be much more capable of running VR. Keep in mind that the Quest 2 seems to do ok with the Snapdragon XR2, so while we might want something beefier from a Valve standalone that would be expected to run SteamVR games, we’re not super far outside of the world of the possible.

      Agreed on the separate compute unit though. I’ve been hoping for a product like that for a long while.

    • A breakout compute box AND portable battery-powered base stations …??
      Doesn’t sound very standalone to me.
      And who’s gonna wanna deal with lugging all that shit around,
      then having to set it all up, recalibration ….
      Bad idea all over if you ask me.

      • g-man

        Stop using bold

        • Jonathan Winters III

          He’ll ignore your request as he has many of us.

          • Rob Scott

            Let me know if he stops using bold and maybe I’ll unblock him.

        • JakeDunnegan

          May as well tell him to stop upvoting his comments while you’re at it. He’ll likely ignore that, too.

      • Jistuce

        “Bad idea all over if you ask me.”
        No one did.

    • Ad

      If it has any standalone use then it would need inside out. At this point we really have zero safe bets on what Valve is doing besides the fact that it will cost almost as much or more than the index.

    • wheeler

      There are Valve patents that detail an inside out tracking system that optionally interoperates with lighthouse https://www.freepatentsonline.com/y2021/0231962.html

  • kontis

    In 2 years a standalone RDNA3 will run Alyx without any changes.

    As PCVR is slowing down and devs focus more on Quest it may be not just a curse for SteamVR but also a blessing. It will make it much less difficult for them to create a standalone PC Quest competitor “that just works” without too much effort from devs, almost like they did it with SteamDeck.

    The problem is that in a few years Facebook may have hardware tech (optical) that will make alternatives completely undesirable, however the history tells us it rarely works this way. People were saying the same thing in 2014 and then Valve+HTC released superior tech with better tracking using small fraction of the money FB used. Apple after decade of smartphone tech domination uses the same display tech every one else uses (irrelevant differences) and not some 10-years-ahead super displays.

    • Jonathan Winters III

      And also, once they figure out much better thermal dissipation, VR standalones (Quest 3?) will be much much more powerful.

      • Jimmie Hunt

        I am being paid with 85 US bucks per-hr to do easy tasks from a laptop…~nk214~I have certainly not believed like it can possible however my bosom mate produced $25,000 only within four weeks easily completing this task and also she influenced me to try…~nk214~Reach new updates through following page >>> http://employment40.fqrd.com

    • Ad

      I think the bigger issue running alyx is the needed RAM and VRAM and other pipeline concerns. I think facebook will have specific hardware advantages but also a huge price and software advantage. I also don’t think the steam library is as consequential as valve might think it is.

    • Blaexe

      In 2 years, people will want a resolution of 3000×3000 per eye with a wide FoV.

      Requirements for Alyx on “simple” VR headsets are a GTX1070 – roughly double the performance of SteamDeck.

      The only way to do it would be very efficient dynamic Foveated Rendering and extremely good upscaling. Basically some cutting edge innovations. We won’t get it just through better manufacturing processes and architectures.

      Don’t hold your breath for it.

      • Christian Schildwaechter

        People wanted 3000×3000 per eye with a wide FoV in 2016. They had to realize that wasn’t feasible, and obviously they learned to deal with it. And somehow a mobile VR HMD that is much less powerful than even the entry level VR requirements from 2016 has outsold all existing VR HMDs in months and is now the most used headset even for PCVR.

        There is and will be no minimal spec for VR, there is just a minimal level that people are willing to accept or be able to use without nausea. An astonishing number of VR users are fine with not being able to run HL:A at 2.5K per eye, which would be possible today, as long as they can play Beat Saber. So maybe instead of claiming that something cannot possible sell, because it doesn’t run Crysis VR, we should assume that anything that performs good enough for Beat Saber has at least a chance to succeed. Especially if it allows you to play your existing Steam library titles without you having to buy them again or the game developers to invest a lot of time to port to a much less powerful mobile SoC.

        • Blaexe

          You’d have to drop the resolution to sub OG Vive levels with this APU if you want to play the existing SteamVR library. This will not be acceptable for people when they can have a significantly cheaper Quest 3 with a significantly higher render target but simpler graphics (and games will be developed Quest first anyway).

          And let’s be honest. PCVR only has a couple of exclusive apps that are actually really good and high quality anyway.

          • Christian Schildwaechter

            The GPU in the Steam Deck’s Van Gogh APU has 8 compute units and is about half as fast as a GTX 1060 on paper, should be slightly faster in real apps. The CPU part is already fast enough to run e.g. HL:A and estimated to consume 15W at full speed, while the GPU will draw about 10W. In the Steam Deck, both CPU and GPU have to share the 15W TDP Valve targets to get to 2-8h of battery life. Users will be allowed to adjust TDP and how power is shared between CPU/GPU, but this is mainly intended to increase battery life further.

            Nothing is stopping AMD from releasing an APU with the same Zen 2 cores, but 16 (or more) instead of 8 CUs. This system would run HL:A fine and pretty much your whole Steam VR library, at the cost of requiring 35W with everything at max load. Add 5-10W for screens, SSD, controllers etc., and you are at 40-45W, about 1/3rd of what the GTX 1060 consumes alone, and about twice of what the current Steam Deck is allowed to pull.

            And this is for a kind of worst case scenario. I have spend an unhealthy amount of time looking into Steam Deck, APUs, TDP settings, FSR in the recent days, and I think that people seriously underestimate how powerful these things are, and how well they could work for VR, esp. with advanced upscaling. The Steam Deck will not be able to play your full VR Steam Library, but a mobile Valve VR HMD is probably at least a year off and could use a newer AMD APU with more CUs esp. for VR. If it ends up as an AIO, they would have to add a much beefier battery, which might be a good idea anyway, as it makes a great counterweight already in the Vive Focus 3.

          • Blaexe

            …do you know how much 40W are? It’s not possible to cool this in a standalone headset form factor. The XR2 has a TDP of only 5W, and yet HTC using its full power with the Focus 3 has a very hard time cooling it already.

            Not to mention it would need a huge battery. Do you want a 2kg headset or what?

          • Christian Schildwaechter

            I am aware. 40W is the worst case scenario with no further improvements, just scaling up what the current APU uses. It also includes a hardware wouldn’t be present on a HMD. There are already other mobile PC consoles like the GPD Win 3 that is about the size of a Nintendo Switch and allows a TDP up to 30W with matching cooling, and they don’t burn your hands.

            I’m not saying that Valve will take the APU from the Steam Deck, double the CU and stuff all that in a Quest 2 case without any changes (this would be that worst case scenario). I’m actually pretty sure that they will make it work with way less than 16 CUs and have already found smarter ways to improve performance.

            But Valve just showed that they can actually build a very tiny machine that can run very demanding PC titles at low temperature, and they said that the whole idea is that you can use your existing Steam library without everybody having to performance tune their games. So it is a very safe bet that they will follow the exact same strategy with a mobile HMD, they will want users to be able to play their existing games without big compromises. And when they get exited about the question and say that

            … [a standalone VR headset], but [Steam Deck’s hardware] would run well in that environment, with the TDP necessary

            , they pretty much already hint that they will increase the TDP to make it fast enough for VR.

          • Blaexe

            This could also mean lowering the TDP, which is imo more likely if you want to put it on your head.

            These PC titles run at 720p and at least 30fps. That’s a very, very different matter.

            Your argument is basically “they’ll find a way” leaving all the technical limitations out.

          • Christian Schildwaechter

            Take a look at the hand-on videos released yesterday, most titles are running at way faster than 30fps at medium settings. There are a number of articles already out, according to which Valve will release a lot more information in the near future, including teardowns etc.

            I’d suggest we wait until we get benchmarks, because this now turns into pure speculation about the limits of a machine that neither you or I have seen. I will assume that it works, partly based on what others that have actually used the Steam Deck or configured similar machines for tests have said, but mostly because Valve said an AMD APU will work well for VR. They didn’t say it will work with all your existing games, but IMHO that is the only goal that would make sense for them.

          • Nancy Branch

            I’m bringing in 88 US bucks per hour to work part time online…~nk202~I not at all believed like it’s possible but one of my bosom mate was obtaining $27k only within four weeks by completing this best offer and she had satisfied me to try…~nk202~Find updated info visiting following web-page >>> https://abre.ai/jobs50

          • Sharon Vega

            I’m able to get paid 85 US bucks per-hr for completing easy work from the comfort of home…~j303~I have certainly not realized like it’d possible but one of my bosom friend got $25,000 just within four weeks by completing this easy job and she influenced me to join…~j303~Have updated details reaching following web-page >>> https://cut.gy/Cash2021

    • JB1968

      Who cares about Alyx? 95% of Quest users are satisfied with Beat Saber and bunch of 2017 “VR hits”. That’s the current state of VR entertainment.

      • Christian Schildwaechter

        And an x86 based Valve HMD would be the only mobile VR device other than Quest 2 where Beat Saber is actually available.

      • Jeremy Kins

        That’s just such a reductionist and not at all true take on VR right now. And if you think it is, well, I have a bridge to sell you.

      • Cragheart

        Who would buy $2500 hardware to play one game Half-Life Alyx?

  • MosBen

    I still maintain that having a separate compute unit is the way to go. That allows the headset to be much lighter and not impacted by the heat generated by computing parts. The compute unit, could be clipped to a belt or pants pocket, which would allow for better thermals. And best of all, the two parts, the HMD and the compute unit, could be upgraded separately as the user saw fit. Want to really push your games to the maximum refresh rate? Upgrade that compute unit. Ready to increase your FOV or resolution? Upgrade the HMD. It would also mean that you could bring your own compute unit, with all of your profiles, saved games, etc., over to a friend’s house and just use their HMD and have all of your stuff immediately available.

    • Wiley t.h.e Coyote

      I’ve heard that valve is aiming to a hybrid hmd just like oq2 is with airlink.

      They want to put their product directly against facebook and oculus.

    • jeff

      you are not a developer, are you ;)

      • MosBen

        I’m definitely not! But what about the idea of a separate compute unit is particularly bad from a developer perspective?

        • Christian Schildwaechter

          The problem is less the separation of the two and more that you can update them independently. One of the reasons why developers like consoles is that they have a fixed set of components. You know exactly what a PS4/PS4 pro can achieve, all PSVR run at the same resolution, so you can optimize your app to really get the max performance possible. Which is one of the reasons why consoles usually have a seven year cycle, now with one mid generation pro version, which is much slower than the development of the PC hardware they are based on.

          All this extreme optimization all falls apart on PC, where people have different CPUs, GPUs, amounts of RAM, SSDs or HDDs and HMDs from CV1 to Vive Pro 2. You now have to choose a target, which mostly means you make sure that the low end works and only add higher resolution textures and maybe a few models for the high end. You wouldn’t e.g. write a game that absolutely needs an NVMe SSD, even if this allowed you to massively increase the graphical experience. You can’t even use some features that e.g. work on NVidia GPUs, but not on AMD, and vice versa, while PS4 and Xbox One developers can. This is partly compensated by the PC being the most powerful platform, but in effect most games don’t fully utilize a high end PC beyond running at higher resolutions or FPS, and making sure that a game runs everywhere takes way more effort than on consoles.

          So for a developer fixed specs are very desirable, esp. on an underpowered platform like mobile VR. A Quest developer can target Quest and Quest 2, adding improved frame rates and textures for the newer version, and it will run as expected on all machines. The option to select or update to a different compute unit is great from a user perspective, but creates a lot of headaches for developers.

          • MosBen

            I’m not suggesting that the compute unit would be completely open. The hardware manufacturer would simply release new compute units on some regular cycle, and developers could target whatever level of compute unit they wanted to support. It’s certainly more complex than an entirely closed system like a console, but now the Wild West of the PC world either.

          • Christian Schildwaechter

            I personally absolutely love the idea of having a separate compute unit, because it solves so many heat/power/performance problems. And after Valve announced that they are working with AMD on FSR for the Steam Deck, I’m almost willing to bet that something like this will happen rather soon. We will possibly see FSR upscaling implemented natively into HL:A before the Steam Deck is released, and this way it could work as a compute unit for the Quest 2 not only for Beat Saber, but also (low settings) HL:A, basically as a proof of concept that compute units for VR will work in principle, without ever calling it that.

            This would allow Valve to later release a very slim and light HMD without integrated compute, ideally along side a more powerful compute unit. This HMD you could a) still connect to a high performance PC, b) use with the matching compute unit to play most of the Steam library at low or mid settings or c) possibly use with an existing Steam Deck, of which by that time a few millions may have been sold, to run a subset of the library that runs on lower end PCs and has been optimized by e.g. implementing FSR.

            You’d end up with three performance targets: Steam Deck, which developers had quite some time to optimize for, the new compute unit, that might use a newer, faster version of the same APU, and the still very varied VR PCs, that now also include lower end PCs that can run the same light, optimized titles as a Steam Deck as the well defined entry level.

            This way Valve can start in December to “unofficially” compete with Quest 2, has a large potential buyer base once their mobile HMD is ready for release, and lots VR developers that have already optimized their apps for AMD APUs and FSR. They would continue to make the money from selling the PCVR titles on Steam, while Facebook would involuntarily provide the cheap HMDs used until the Valve mobile HMD is ready.

            And by then they might even be able to squeeze an alternate tiny compute unit with the power of the Steam Deck into a small package that can be placed at the back of the HMD as a counter weight, turning it into AIO without any wires. People could freely choose how much power they need and how many wires they are willing to accept.

            A lot of speculation based on very limited facts, but as a Valve strategy against Facebook this would be very beneficial, and by now they have dropped so many hints that, as mentioned, I’m almost willing to bet on it.

          • Arno van Wingerde

            Your suggestion would be somewhere between a Quest2 and a quest2 with wireless connection to a PC. Yes there are disadvantages to the last option, you cannot pick up your VR set and play somewhere else, have to switch on/update/upgrade things separately and not always have a standard matched set and a slightly slower response time, but the CPU/GPU power outguns any compute unit you have. A separate compute unit would rive up the price and possibly not offer performance advantages.

          • Christian Schildwaechter

            A compute unit is basically a portable PC, so yes, this is like a Quest 2 connected to a VRPC. The main difference is that these compute units are portable and battery powered. If you want to use them with a tether, you can wear the unit at the waist and still have the same movement freedom as with an AIO. If you want to use them wireless, you can carry both Quest and compute unit in a small backpack for your camping trip, place the compute unit on the tent floor and play HL:A in the middle of the forest. Of course you could use a similar powerful gaming laptop for this, but the Steam Deck is available for USD 400, which includes screen, controllers and speakers, none of which would be needed for the compute unit. You will not find a gaming laptop with a GTX 1050Ti for this price.

            Compute units aren’t the ultimate solution for everything. They simply add an option to have a more powerful, but still fully mobile VR solution than current AIOs, and they cost less than buying a gaming laptop or PC. This isn’t relevant if you are fine with the performance of the Quest 2 or if you already have a VR PC and only use VR at home, but there are numerous situations where they would work better than what we have today.

    • You’re talking about an AI2 [“All-In-Two”] which
      HTC’s big mouth CEO said was gonna make a big splash this year.
      I’m still waiting …. lol

    • Ad

      I still don’t think this thing will have standalone use in the traditional sense, but we’ll see.

      • Yvonne Guajardo

        I’m making 85 US dollars every hour to complete easy tasks from the comfort of home…~gk253~I have not ever thought like it could even achievable but one of my close buddy collected $27k just in six weeks just working this super career as well as she influenced me to try…~gk253~Reach new instructions going this article >>> https://link.gy/Jobs100

    • RFC

      Was fortunate to have several sessions with Magic Leap, the separate compute (Lightpack) and tether worked better than I anticipated despite the simple strap.

      Offloading compute from the head made them feel very comfortable with less craniofacial pressure and lower inertia compared to Hololens.

      Building a proper hip harness for a separate compute unit could be very effective; perhaps a temporary form factor until wireless glasses are reality?

    • Ming

      The market for a separate VR compute unit might not be high enough to make financial sense, seeing as the market for VR headsets already barely makes financial sense. But hopefully it will be possible to just strap a Steam Deck onto a belt and plug it into an inside-out VR HMD. Who knows, maybe Sony’s PSVR2 will be compatible with the Steam Deck too.

    • care package

      Everything has its ups and downs. A belt strapped CU involves more wires for one, and more expensive.

      • MosBen

        Why would it be more expensive? Sure, there’s some extra plastic involved, but that’s fractions of a penny per unit. And the fact that it’s not built into the HMD seems like it would save time and money developing and implementing cooling solutions.

  • g-man

    I think they’ll release a headset with enough compute to do inside-out tracking and local reprojection, which is necessary for lower bandwidth wireless like WiFi, rather than a full standalone headset that runs PC games. The latter would have too many compromises.

    • Ad

      They could ship more features on a device like that too, like a quest that didn’t use any of its on board compute for the user’s software.

  • Ad

    Why… would they do this though? Rather than ARM? Compatibility would be very mixed and while valve definitely believes in the supremacy of the steam library, that is far less important for this when all the games that would run well on it are already ported to arm. Like the only advantage of this is that facebook can’t stop you from running beat saber on it.

    • Christian Schildwaechter

      The Steam library is the key here, because it allows Valve to compete with Facebook’s nearly unlimited resources. If Valve build an AIO VR HMD, they are pretty much in the same position as the HTC Focus 3 or the Pico Neo 3 Pro: slightly better hardware than the Quest 2, but no software compatibility to either Quest nor SteamVR. They would have to convince developers to port, which they will only do if they know that the HMD attracts a sufficiently large number of potential buyers for their games.

      Porting from Quest would be even harder if Valve decided to use their SteamOS experience and build an ARM based HMD running SteamOS, while all other AIO run some form of Android. Running existing x86 on ARM via emulation takes a huge performance hit on any non-Apple ARM SoC, and Apple bought several ARM design companies and an unlimited ARM license for several hundred million USD to be able to add special instructions to their M1 silicon to be able to run existing x86 at near native speed.

      There are fundamental limits to the x86 architecture that cannot be easily circumvented, making them less efficient at low TDP, but this is only the CPU part, while VR is usually GPU limited. The Quest 2 even underclocks the CPU to run the GPU faster, and AMD have been the performance leader in x86 iGPUs for years, so using an AMD APU for a VR headsets is much less critical than in an Ultrabook running CPU heavy productive apps, while still allowing compatibility to the existing VR library without having to convince any developer to port their games.

      From a technical point, an AIO HMD based on ARM looks more attractive, esp. if they used something like the upcoming Samsung SoCs that actually integrate versions the RDNA2 GPUs featured in Steam Deck or PS5. But from an “how to compete with Facebook without having their deep pockets” point, using an x86 APU and running the existing library is absolutely the correct decisions. People will buy it simply because the will be able to use all the software they already payed for, many of which will never be ported to Quest 2.

      • Blaexe

        You presume that we’ll see significant improved efficiency in the near term – because even the 15W SteamDeck APU won’t cut it for existing PCVR games. And it’s not even close (if you want to use resolution and refresh rate people expect in 2022+), we’ll see when people have SteamDeck in their hands.

        • Christian Schildwaechter

          No, I assume that they will use a variant of the current APU with more GPU CU units and use something like FSR upscaling, which was added to SteamVR on SteamOS by users within days in a very patchy way. Doubling the number of compute units would add about 10W and could put the APU at current VR entry level. They would probably have to lift the TDP limit, which was set to 15W to conserve battery, not because the APU couldn’t run faster.

          This is based on what the APU does today, without any further improved efficiency, without FSR being implemented by the game developers. AMD already published timelines for future APUs, including models with 16CU, so this isn’t exactly science fiction. I’d suggest waiting for actual performance reviews of the Steam Deck, including the inevitable attempt to run HL:A on it before assuming that this isn’t a feasible VR platform. I’m pretty sure you will be surprised.

          • Blaexe

            As I said in the other comment, you won’t be able to cool an APU that powerful in a standalone headset. That’s a pipedream.

            I’m sure people will get HL:A to run as a proof of concept – but not in a way people will actually play it

          • Ad

            FSR is very blurry and doesn’t work in all games. I highly doubt they’ll push it as a general use thing in VR. Also alyx needs more RAM and VRRAM than the Deck has and the bottlenecks from the LPDDR5 will probably mean it doesn’t run well even at very low and blurry loads.

          • Christian Schildwaechter

            FSR is only blurry if you add it at the end of the rendering process, like in the patched SteamVR version. The proper way to use FSR is adding it into the render pipeline before applying any post processing or rendering the UI, so for it to work well, it has to be implemented by the game developers. Luckily it is open source and adding it to the pipeline is very simple. And even just patched in FSR can be useful for games with a lot of sharp geometry, while it will pretty much fail with anything that has blur added on or a lot of fuzzy structures like the vegetation in Skyrim VR.

            FSR also requires a rather high resolution to begin with. In tests upscaling by about 33% to 2160p or at least 1440p works great, upscaling by 50% is usable, but once your native resolution drops below 720p, the image gets very mushy. Which is why FSR doesn’t really help for playing on the Steam Deck at 720p, but would be great for upscaling an image rendered on an AMD APU from at least 720p to something bearable in VR. It is not a solution for everything, e.g. the TAA in Unreal can often deliver better upscaling than FSR.

            HL:A needs 12GB RAM, the Steam Deck has 16GB, shared with the GPU. HL:A actually runs on machines with 8GB RAM, but very badly. The LPDDR5 in the Steam Deck has a bandwidth of 88GB/sec, which is low compared to GPUs used for VR, but very high compared to typical iGPUs. A lack of bandwidth would most certainly become a problem with high settings involving large textures, but the GPU isn’t fast enough for these, you will be pretty much stuck to mid or low settings, where the speed should be sufficient.

          • Ad

            FSR has been implemented natively in a few games on flatscreen and it’s far from perfect, not something you can lean that much on.

            HLA needs 12GB of PC RAM and 6GB of VRAM. You seem to understand that the LPDDR RAM isn’t fast enough, but then hand wave it away for no reason. I can see you want this all to make sense but you’re having trouble with the fact that it doesn’t.

            There’s also this assumption that this would be a noticeably better experience than a Q2, even just with these legacy games whose developers have moved on to quest, and I’m not seeing it.

      • Ad

        The Steam library matters outside of VR, but in VR there are real issues with valuing it that much. Think about this, if they use ARM they can just sell openXR builds which will work out of the box, but if they use x86 they have just the steam VR library. Try something, make a list of the top ten games you like in VR besides Boneworks and Alyx, and then see if they’re on Quest or coming (and the boneworks sequel will be on quest). Developers are just going to facebook. Also a steamOS mobile would just be android since android comes from linux and they wouldn’t need proton.

        • Christian Schildwaechter

          OpenXR is great for developers, because it reduces the amount of time it takes to port an app. It isn’t all that helpful directly for users or publishers. OpenXR builds will only work out of the box if they were compiled for the same architecture (like ARM) and the same operating system (like Android), if they use no special APIs (like SteamWorks) and if a VR HMD allows installing them. That is a lot of ifs.

          So if Valve uses ARM, they could start to also sell ARM binaries, which then might run not only on the Valve HMD, but also on Focus 3, Pico 3 or Quest 2, but at least on Quest they wouldn’t be allowed to install the Steam store. At best they could sell them through App Lab, with 30% of every sale going to Facebook, or user would have to use something like a Steam version of Sidequest. Not a particularly attractive option. Using x86 they can simply use their existing store and sell the existing apps, and any new app that will target a new mobile Valve HMD can also be sold to any existing PCVR user. Without any ifs.

          A quick check of my Steam library showed that I have 434 VR supported titles (out of 2086 titles total). That is probably more than all Quest store and App lab titles combined. A quick check on the Steam store revealed there are currently 5439 VR supported titles available. I’m not even able to try all the titles I already own. The huge library is Valves real treasure, and nobody will be able to compete on that level for a very long time.

          I don’t know what Valve will do. I just extrapolate from their current hardware release and the strategy they have with it. I wouldn’t expect anybody else to release an x86 HMD, because ARM has so many benefits, but then nobody else could benefit from an x86 HMD in the way Valve can. I know that they will not get access to Apple Silicon and that Qualcomm SoC is very inferior to both x86 and Apple ARM regarding the GPU. The XR2 is based on the SD865, which was released in 2019, while the Steam Deck is based on AMD APUs that will not be available to other manufacturers until 2022.

          For me all that indicates that a Valve HMD based on an AMD GPU is pretty smart idea, and Valve seems to think so too.

          • Ad

            Your logic is just bullshit. I’m sorry, you seem to have a very weak grasp on all of this. Valve’s existing library is a ticking time bomb to irrelevance. Obviously ARM is the better option, it could run on other hardware (sideloaded on quest as well), it would get ports from ARM android (the set up every other developer is building for) vs an x86 platform where they have their existing library (half of which is on quest, half is abandoned or will run terribly on this). You’re obsessed with this mythical steam library that will kill the quest but it’s nonsense, all of the notable games are either impossible on either mobile hardware solution or already ported or being ported. Also you don’t seem to get that qualcomm’s chips are getting core VR functions built into the board the way we have mpeg codecs on x86 chips now.

    • kontis

      A few months ago I would agree with you but after listening to Jim Keller interview and then checking it myself I realized there is no magical advantage to using ARM.

      Turns out all the magic of super long battery life of ARM is only relevant in low loads and especially when idle. This is huge for phones, but NOT for VR which has to constantly render graphics.

      1. iPad M1 lasts only 1.5 hour under heavy gaming/rendering load
      2. under moderate to heavy load Intel Macbook lasts only 15% less than M1 Macbook. I thought, due to all the myths, the difference would be easily 100% (2x), but nope. It’s very close.

      • Ad

        Do you really think that the SteamVR experience, which is already kind of strained and buggy on my 1080ti, will work right at this performance level? If Valve needs devs to optimize their games for this new device then the library advantage evaporates.

  • A PC-based standalone system would be interesting, but also more complicated for developers that want to port Quest games to it (unless OpenXR saves us, of course)

    • Ad

      The weird thing is whether Valve believes in its own mythology enough to make such a weird choice. And while we could get OpenXR ports, wouldn’t be really pointless to go to all this effort to make PCVR content usable and then users just install ports from mobile that just run less efficiently?

  • Joe Austin

    I am producing 88 dollars hourly to do job off a home computer…~ak281~I not ever realized that it’s even achievable but my bosom mate was receiving $27,000 just in five weeks just completing this super project also she had influenced me to avail…~ak281~Get further instructions by reaching this web >>> https://lst.to/ov6ck

  • wheeler

    I would still caution people depending on what their expectations are for “standalone”. Valve’s patents explicitly point out using in-headset compute for a variety of edge processing functions (which makes sense) but there is a big difference between “used primarily for edge processing functions to complement the PCVR experience–and oh you can run some applications on it too” and “Quest-like standalone functionality”.

    It’s going to take some unheard of combination of foveated rendering, motion smoothing, and upscaling to make PCVR games work “as is” on an in-headset processing unit.

    • And fairy dust.
      Don’t forget fairy dust.

    • Ad

      there is a big difference between “used primarily for edge processingfunctions to complement the PCVR experience–and oh you can run some applications on it too” and “Quest-like standalone functionality

      Exactly. I still don’t think this thing will actually be fully standalone unless it’s like streaming from anywhere in your house.

  • ZarathustraDK

    According to their patents the thing they are working on in regards to VR is a Wireless HMD that outsources some of the things that make sense (like tracking) to the HMD while the hefty computations remain done on the desktop pc.

    When they say “it’s very relevant to us and our future plans” it probably means they’re planning to shove an AMD APU in the HMD that takes care of tracking and crosses the t’s and dots the i’s in the wireless transmission. Perhaps some kind of dedicated upscaler like FSR to keep the bandwidth lean between the desktop and the headset, then let the headset finish it off?

    • Ad

      I agree but it’s a weird choice to use x86 for that purpose when Virtual Desktop is already doing some of this on ARM.

  • JB1968

    The actual version fo Steam deck is a VR joke from the performance point of view. Looks like Valve is trying to pitch the product at any cost event to the hopeless PCVR crowd.

  • Ad

    Disqus is the absolute most terribly coded piece of software I’ve ever used. I didn’t even realize some of these bugs were possible.

    • JustNiz

      Apparently you have never used anything written by Microsoft.