Valve Steam Frame VR Review vs. Meta Quest 3 | Privacy, Latency, & Benchmarks

The Quest 3 costs $600 currently -- and also your first born, alongside video and voice data from inside of your house, scans of your room, your body movements, facial expressions, reading around 83,000 words of agreements, and depending on how much you spam-agree to, “audio recordings of your child’s voice, “Physical information about or related to your child,” “Headset location,” “your child’s connections to others,” “your child’s activity in Meta Quest devices and apps” (like apps downloaded, gameplay activity and stats, content created as well as the time, frequency, and duration of activities), and, with approval, “the size of walls, surfaces, and objects in your child’s room”

We asked Valve if it needed those things, they didn’t need 83,000 words to answer us. The answer was one word: No.

We have a ton of other tests, too: In latency testing, Valve overall had a favorable position as compared to the Quest 3; however, the Quest 3 had a favorable position as compared to Valve in battery life testing. Thermal testing also came into play.

One of the big tests we did for this was thermals to represent comfort in an objective sense. We used a Fluke RSE600 thermal imaging camera, which is connected to a laptop. The purpose of this test is to look at not just the external temperature of the headset, which isn’t too important, but to examine the temperature of your face when you're wearing it. We then take that data and compare it between the Frame and the Quest 3 to get an objective read on the human factors skin temperature.

And thermal results position the Frame as having some objective advantages in comfort, alongside the subjective opinions from our team.

Mike on the team stated, “The Frame definitely felt a lot more comfortable just as a whole.”

Patrick stated, “This, out of the box, is a lot more comfortable.”

The price is the worst part of the Frame, especially as Valve broke its promise related to price, and it’s not just the Frame alone:

The accessories kit is $60 for straps that we really think are mandatory, seeing as you need them to make full use of some advertised features. The charging cable is $30 and isn’t included, the cushion replacement with gasket and light blocker is $50.

So, the price is the rough part with the Frame, but the price is also the rough part with the far cheaper Meta. It's just that you're paying for it with literally everything about you. If you thought LG's terms were bad? Jesus Fuck! That's all we have to say about this one.

The market started to get bad right when Valve announced the Steam Frame in November, and in the time since, memory prices have gone up on average 350-600%, depending on which spec of memory you look at, and storage has increased by more than double.

We have a lot to test today with the Frame. 

This review is actually the first hardware review where we’ve added privacy & security as a significant section of the content, and that’s all thanks to your backing. 

Steam Frame Overview

Today, we’re focusing on the Meta Quest 3 and the Valve Steam Frame. The reason for this is that, in the 2 weeks we’ve been working on testing, we’ve already fully maxed-out our available hours just for those two headsets. Rather than doing shallow depth across a lot of headsets, we’re focusing our efforts on going in-depth for the main two. PiMax and Bigscreen also exist, but we haven’t tested them yet.

Meta completely dominates the VR market, even on Valve's own PC hardware survey, and the Quest 3 is its flagship. 

What's Included with the Steam Frame

The specs are public, but here's a rundown: the Frame has a Snapdragon 8 Gen 3 8650 SOC, 16GB LPDDR5X (combined system/VRAM), 256GB or 1TB of onboard storage with a microSD card slot for expansion, 2160x2160 LCDs with pancake lenses, refresh rates from 72-144Hz (144Hz is "experimental”), FOV up to 110 degrees, four external IR-sensitive monochrome cameras, two internal eye-tracking cameras, an expansion port with both a single PCIe Gen 4 lane and "8 lanes @ 2.5Gbps MIPI," WiFi 7, Bluetooth 5.4, and a 21.6Wh battery on the headstrap.

The whole thing weighs about 445g, +/- 1g, by our measurements.

The core module containing the computer, optics, screens, and all the other expensive stuff locks into the modular headstrap, which only contains the battery and speakers.

The Frame comes with the headset, a spacer for glasses, two controllers with lanyards, the special wireless adapter, a USB Type-C cable, and a cleaning cloth. 

You don’t get this AC adapter included -- it’s a $30 optional accessory that matches the Deck charger, although any USB power supply will work -- such as the one that comes with a Quest 3. 

A replacement kit for the stock cushions and rubber nosepiece runs $50, in the event they deteriorate or get nasty over time.

The single USB Type-C port is for charging and data (only at USB2 speeds, which is 480 megabits per second). 

Unlike the Quest 3, the Frame lacks a 3.5mm headphone jack. The Type-C port can be used with 3.5mm adapters.

The stock headphones were designed to be lightweight and portable, but an enthusiast headstrap kit will launch in a few months with Index-style headphones and a hot-swappable battery. This will be an add-on kit, not another Frame SKU, so the kit will only include the accessories.

Valve has a number of DLC add-ons, like the "Ergonomic Accessories Kit" for $60 US with controller hand straps, a headset top strap, and an extended light blocker. 

Some of these feel necessary: the hand straps are important for finger tracking since they let you fully open your hands to actually utilize the feature, and the default light blocker may leave a significant open gap depending on your face shape.

Patrick found the light blocker accessory unnecessary, while Steve found it necessary due to the extreme amount of light coming in. The optional top strap just got in the way for some of us, but for Steve, it helped with headset bounce when walking, possibly because it helps keep the strap secure over long hair.

We think some of Valve’s first-party accessories should be included with the Frame, like the hand straps that are necessary to fully utilize the advertised finger tracking feature beyond taking one or two fingers off the controller.

Privacy

Up next, thanks to our LG deep-dive, we have a new section that we’ve never had in reviews before: a review of the privacy settings, the extent of the devices’ spy capabilities, and the privacy policies. After all, both the Frame and Quest are loaded with cameras, microphones, and motion tracking, and Meta’s involved.

Valve Steam Frame Privacy

Starting from the checkout: Valve requires users to accept two agreements when purchasing its hardware – Valve’s privacy policy and Steam Subscriber Agreement – totaling a collective 12,616 words. If you have a Steam account, you’ve likely already accepted these.

Valve’s privacy policy explains how it collects account data, gaming activity and statistics, device information, transaction/payment data, posts/comments/chats on Steam, interactions on the Steam client, and tracking cookies/pixels, which it claims it uses to provide both its services and content recommendations.

Additionally, in a statement to us, Valve clarified:

“Camera and eye-tracking data is processed on-device, and we don't collect it, nor do we collect byproducts of it.

Steam Support has long had the ability for users to choose to submit a system report to help us troubleshoot problems they may be having. For Steam Frame, that report would contain text logs that include some camera and eye-tracking systems data.

As always, Valve never sells user data to third party advertisers.”

Meta Quest 3 Privacy

Meta’s Quest 3 checkout, on the other hand, has seven user agreements totaling over 36,000 words. That includes Shopify’s terms and privacy policy that Meta attempts getting users to unnecessarily agree to.

Then, the Meta Horizon app that Meta claims is required to complete the Quest 3 setup requires users to accept an additional 5 agreements containing 46,579 words, so you’re now at almost 83,000 words of agreements before using the headset.

Meta’s App Store page says: “The following data may be collected and linked to your identity: [...] Health & Fitness, Purchases, Financial Info, Location, Contact Info, Contacts, User Content, Search History, Browsing HIstory, Identifiers, Usage Data, Sensitive Info, Diagnostics, and Other Data.”

In addition to collecting all the same types of information as Valve, Meta also records and processes all voice interactions using its voice services, “including conversations with AI-powered features” and they say, “including if you mistakenly invoke our voice services, as well as any background sound that happens,” so they’re pulling a full LG but they’ve actually written it a little bit better. Meta claims that voice dictation recordings aren’t stored. Meta also collects all video and audio content you create, including “video or audio content and information about this content, such as the date and time you created it,” and also “environmental information and dimensions,” and what it calls ‘abstracted’: hand and body data, gaze data, facial expressions data, and fit adjustment data.

Meta claims that raw image data is processed locally, but says it processes what it calls the "abstracted data" and that it collects and retains certain data collected from those interactions, explaining in its Natural Facial Expressions Privacy Notice:

“When you choose to enable Natural Facial Expressions on the headset or in a specific app, software on the Meta Quest Pro headset analyzes images of your face (“raw image data”) to create an estimate of how your face is moving, producing a set of generic facial expressions, like a broad smile or frown, (“abstracted facial expressions data”) that animate your avatar. The raw image data is processed in real time on your headset and deleted once processing is completed. We do not collect or store raw image data from the Natural Facial Expressions feature on Meta servers. The abstracted facial expressions data is generated and overwritten in real time on your headset, and processed on device or Meta servers to animate your avatar’s facial expressions. For example, if you choose to give an app offered by Meta access to your Natural Facial Expressions data, the abstracted facial expressions data may be processed on Meta servers in order to animate your avatar’s facial movement in a multiplayer scenario.

Meta notes:

“Additionally, when you use Worlds, the last few minutes of your and other users’ most recent audio, video and other interactions in Worlds are recorded in case you want to report anything you’ve encountered. These recordings may be stored on our servers. We do not review these recordings unless you submit a report. If you don't submit a report, the recordings are deleted on a rolling basis.”

Meta also assigns users advertising IDs (enabled by default), has significantly more users agreements, and includes a forced arbitration notice in its terms.

Valve actually removed its mandatory arbitration clause and class action waiver from its Steam Subscriber Agreement in late 2024 after getting “threatened with an avalanche of individual dispute resolution demands,” as reported via Reuters.

In Meta’s “Parent Privacy Disclosure,” the company tells parents what information it collects about their children aged 10-12 years old when using Meta VR products, listing: 

• “audio recordings of your child’s voice” if voice services are enabled
• “video and audio information recorded when cameras and microphones are enabled”
• “Physical information about or related to your child,” if hand and body tracking is enabled
• “Headset location” if location services are enabled
• “your child’s connections to others”
• “your child’s activity in Meta Quest devices and apps” (like apps downloaded, gameplay activity and stats, content created as well as the time, frequency, and duration of activities), 
• And, with approval, “the size of walls, surfaces, and objects in your child’s room”

We wouldn’t want literally anybody having access to that much of a child’s information, and certainly not a company with Meta’s track record.

Only a year ago, two former Meta researchers “studying the welfare of underage users of Meta VR products” told a Senate subcommittee that Meta’s VR products exposed children to adult content and claimed that “Meta disregarded their findings, censored their attempts to do further research and, in at least one instance, deleted evidence of youth being harassed,” as reported via NBC. 

As for why Meta needs so much of your child’s information, one of the researchers explained, “Children drive profits. If Meta invests more in safety to get kids off of there, engagement goes down, monetization goes down, ad revenue goes down. They need them. They need them young. And we know that younger audiences drive later engagement across many types of products, but Meta’s smart, they know that too. They need them – they need children.”

And back to Valve, as we said, it’s the two Steam agreements, they don’t exfiltrate room, object, or body data off-device, and they don’t sell data to third-party advertisers. As far as we can tell, the only time Valve uses data is in optional cases where users submit support tickets, for example, if you complain about eye tracking or head tracking issues or the data that it already uses for your Steam account for things like content recommendations and things that are on the Steam store rather than just the VR headset. 

So there’s a pretty big difference between the two products and maybe a bit of an explanation as to how Meta can price their headset low. 

Steam Frame Fit and Ergonomics

Out of the box, almost our entire team found the Frame is easier to put on and more comfortable on the face than the Quest 3 (and a massive improvement over the original Vive). Multiple members of our team consider the Frame the most comfortable VR headset they've used. 

The openness of the nose area on the Frame makes it easier to, for example, drink from a can while using it (versus a Quest 3 with its nose overhang and enclosure); however, some people may find this positive to be the reason light bleeds into the headset, depending on face shape. The comfort and casual nature of the Frame is its biggest advantage over the Quest, but we also haven't tried any ultra-lightweight headsets like the Bigscreen Beyond 2.

The face gasket and eye spacer use magnets and the rear cushion uses hook-and-loop patches. The magnets take some getting used to as the face gasket pulls off extremely easily. The stock position of the lenses is potentially in eyelash-brushing-range, meaning that the inner glass gets dirty relatively easily even if not touching it just from eyelash contact, although we couldn’t physically feel it. We left the spacer on even without glasses for this reason. Maximum claimed width for glasses is 140mm, but we were able to squeeze in a pair with arms out to 145mm.

We'll cover this in the thermal section but, anecdotally, we didn't feel unpleasantly hot or uncomfortable with the weight and balance of the Frame. There's a very obvious exhaust vent on the top, which we prefer to the hidden cracks on the Quest 3. We managed to overheat a Quest 3 by sticking a label over the wrong joint.

For cleaning, Valve says to use their microfiber cloth and water (no chemicals). We found it difficult to keep the lenses clean, although this didn't noticeably affect vision.

Steam Frame Thermal Testing

We initially attempted thermal testing using thermocouples attached to the face, but abandoned the test due to issues with (1) getting them to stick reliably, and (2) not influencing the results by trapping thermocouples under non-breathable adhesives. We instead switched to a Fluke RSE 600 thermal imaging camera.

The process involved playing Moss locally, at equal settings on both devices, for 30 minutes. Then, upon removing the headset, we instantaneously take the thermal measurement of the wearer’s face. This is obviously imperfect due to human factors; however, we controlled as best we could by taking a control measurement, testing in the same ambient temperature, and testing on the same day (with a brief cooldown).

The results generally favored the Frame for face skin temperature during use. Within 1/3 of a second of taking the headset off, we captured browline numbers -- which land right underneath the gasket for each headset -- at 5 points. Those had the Frame at 33 degrees to the Quest’s 35 at the hottest point. This is our most useful figure, as the larger gap seems more likely a real impact rather than variance. Browline 2-4 also favored the Frame by about half a degree. The Browline 5 measurement had the Frame at 34.2 to 35.0 of the Quest 3.

Eye temperature also ran warmer on the Quest 3, at 34.7-35.1 C versus 33.9-34.2C for the Frame.

The Quest 3 ran noticeably warmer on the cheek, right below the eye and at physical contact with the headset. 33.6 is an advantage against the Quest’s 35.1. The nose hotspot was also noticeably higher with the Quest, at 34.6 to 32.8 on the Frame. Given the headset shape, that makes sense.

We also captured palm temperature hotspots by holding hands up to the camera immediately after setting the controllers down, but the imperfect nature of this measurement makes it unreliable.

The headset lenses ran warmer on the Frame, which indicates to us that its internal ventilation must just be working better than the Quest’s. HMD front external temperatures were also higher on the Frame in areas that don’t contact skin.

Testing Intro

We have a lot to talk about with controls, the UI, and software, but let’s run through some hard testing numbers first and then come back to those.

Latency Methodology

Latency testing is next. For this, we’d ideally want the end-to-end latency of the VR system, so in other words, how long it takes from button press on a controller to an action happening on the screen.

We used the LDAT for this and attempted to solder wires into a button on the controller so that the LDAT would be aware of a press, but the Steam Frame controllers were difficult to disassemble safely and involve glue. We found that soldering the LDAT to the grip button interfered with its function, and we were worried about breaking the controllers to get into the face buttons. Normally, that’s not a concern. But we only have one of these for testing and the controllers use glue that we can’t easily get at.

We also tried a Steam Controller hardwired to the Frame's USB-C port, but we ran into issues there, too. 

Wiring into a button is ideal because we don’t need to count frames and we can run a 100-sample test for accuracy. Since wiring in a lead didn’t work, we realized we could instead use a microphone. The LDAT microphone hears the button click and starts a timer, then when it sees a change on the monitor (or headset, in this case) it stops the timer. It’s not as fast as the velocity of an electron, but 343 meters per second is fast enough for us.

All streaming tests were hosted on our GPU bench with an RTX 5090 and a 9800X3D. The Frame was set to 120Hz since the only higher option (144Hz) is labeled experimental, and 120Hz is also the highest refresh rate supported by the Quest 3 without tweaking. WITH tweaking, the Quest 3 can apparently run at up to 207Hz, or 240Hz with a reduction in resolution.

The easiest test would be to stream the regular LDAT test software in theater mode in VR, which works, but 2D streaming and VR streaming are different pipelines in Steam. For example, multi-link streaming doesn't currently work with non-VR games. We did some initial tests anyway out of curiosity, but these numbers don't apply to VR streaming.

Latency Benchmarks - LDAT Test Software

Here’s the first chart with controlled testing using the LDAT software Vsynced to 120Hz on the PC side, as otherwise it was hit-or-miss whether we’d stream the white frame needed to trigger the LDAT. 

These tests were done at 1920x1080 with video quality set to the “beautiful” setting, unlimited bandwidth, a 240FPS cap set for Remote Play, and a 350Mbps cap in SteamVR. Inputs were issued by a mouse connected to the host PC, so the controllers aren't part of these numbers. This is control testing. Throughout testing, we recorded three baselines with 100 samples each on the 360Hz monitor, which fell between 24.3ms and 25.6ms. At 120Hz, the Frame over the wireless adapter averaged 46.1ms and the Frame over WiFi averaged 47.1ms. That these figures are so close is an achievement, and the adapter is technically better here.

The Quest 3 on WiFi averaged 72.5ms in the same test. The Quest 3 required 57% more time to make the trip as compared to the Frame’s best result. That’s an incredibly strong start for the Steam Frame and a rough one for the Quest, but VR game streaming is the real test.

Latency Benchmarks - Thumper

We’re using Thumper as a test scenario since it can run natively on both the Frame and Quest as well as streaming. We’re also using it because Thumper has clear actions that are photonically observable. 

The disclaimer of test limitations is that we're testing one action in one game, so these are not the set-in-stone latency numbers for each device. 

Here's the chart.

Big picture, the Steam Frame’s latency in this specific test is superior to the Quest 3, with the exception of the Steam Frame on WiFi (without the wireless adapter or multi-link).

We kept the same 120Hz refresh rates and maxed-out stream settings, since the point of the Frame is supposed to be that it can reliably handle high-quality streaming. Our control test with the monitor measured 24.8 ms. Running Thumper local to the Frame at 120 Hz, we measured a 56.5 ms response time to what the player would see. At 72 Hz, predictably, that got worse (with fewer cycles) and climbed to 61.2 ms. With multi-link streaming from a host, non-local, and at 120 Hz, we measured a 65.7 ms latency. Versus local, this streamed test added 9.2 ms, increasing the time measured by 16%. Using only the wireless adapter pushed that to 68.4 ms, another 2.7 ms (or 4%) on top of multi-link. Multi-Link does seem to be better than just the adapter in at least this specific test, and by a measurable amount.

The Quest shows up here: Locally, which uses a different build via the Meta store and may have variance, the response with the default 72 Hz refresh was 75.2 ms. Like-for-like against the 72 Hz Frame locally, that’s about a 23% increase in measured latency versus the 61.2 ms result of the Frame. Running from the host computer rather than locally, we obviously see some slow-down. 

We opted into the beta version of the Steam Link app on the Quest to get plug-and-play streaming over USB using the cable that shipped with the headset. Assuming the beta feature was working correctly, it didn't have much of an effect and landed at 82.9 ms. That was with the cable that ships with the Quest 3, which isn't good for high transfer speeds and Meta wants to sell you an $80 link cable instead. We did not have one of those to test.

With the Quest 3 over WiFi and 120 Hz, we measured 83.5 ms. The Steam Frame on WiFi and 120 Hz ran at 86.2 ms.

Keep in mind that specific results for WiFi will depend on your router solution. Our tests were conducted with like-for-like hardware and in the same spot, so the WiFi solution was fixed for all tests.

The low-persistence displays of the VR headsets make it hard to use the LDAT since it relies on changes in luminance and the screens are rapidly strobing. You can see this effect easily by filming inside the headset. In the original announcement, Valve told us that the duty cycle for its LCD panels was locked at 8% at all framerates, meaning that, at 120Hz, the panels are illuminated for approximately 0.67ms out of every 8.3ms refresh. This reduces blur, and is part of the reason Valve opted for LCD over OLED.

Multi-Link and Foveated Streaming

Up next, we’ll explain how foveated streaming and multi-link work.

The Frame is presently wireless-only. Valve’s included wireless adapter gives the Frame an advantage over the Quest 3 in households with bad wireless signal strength or old routers. 

We found the wireless adapter mostly suited to streaming in the same room. Walking about 8 meters away from the test bench and facing directly away made the signal start dropping out, while facing towards the bench set it back to "excellent." It offers decent range but you’d have trouble piercing through walls. 

The Quest 3's solution to bad WiFi is that it can be used wired. As of August 31st, Valve updated the beta version of the Steam Link app on the Quest to make it plug-and-play.

The Frame can connect to normal WiFi, and it can also connect to the included Wireless Adapter: it has two radios. When you connect the Frame to WiFi, it gives the option to reserve its 6E connection specifically for the Wireless Adapter. 

If you connect the Frame to the Wireless Adapter but not to WiFi, you can still stream games locally through the Wireless Adapter, but your headset won't be connected to the internet. This would isolate your headset away from the internet.

The Frame can do multi-link streaming as well with the simultaneous 5GHz WiFi and 6GHz wireless adapter connections. Valve told us: "All of the links are always active and competing for the lowest latency on a per-packet basis, acting as a single robust connection with built-in redundancy that adapts in real-time. This means your streaming connection is always as powerful as the strongest link, and often better than any single link."

Foveated streaming is a new concept that Valve has been pushing that utilizes the Frame's eye tracking to prioritize high-bitrate streaming directly in front of your pupils, but there's not much we can do to demonstrate it in video recordings. 

There's a debug tool to highlight the foveated area that gets the highest-quality stream, so we know it's active, but it's hard to notice by design. Valve says: "The angular size of the foveated area is a function of game resolution and the encoded video size." We kept the encoded video size slider maxed out at its default value of 1344px for everything in this review.

Steam Frame Battery Life

Battery life and charge testing is next.

Meta doesn't have a battery capacity listed in its technical specs: it just says it should last 2.2 hours overall, or 2.4 when gaming. The most common number is 19.44Wh from an iFixit teardown, but the label on the battery in that video looks like it says 19.12Wh. Either way, it's probably somewhere around 19-20Wh. Valve's battery claim is 21.6Wh, although differences in measurement are always possible. 

Valve told us that "highly optimized games (and streaming) may use <10W, while some un-optimized or higher-spec games may draw up to 25W (peak)." That works out to 2.16 hours of streaming at 10W, so on par with the Quest on paper. Valve also claims that idle (screens off) power draw should be about 3W, and the long-term low power state draws about 250mW.

Battery Life Benchmarks - Moss

For our first battery test, we picked Moss, whose developers sadly shut down 2 days before the Frame launch. 

Starting with the Quest 3, we have control over very few variables on the headset itself: basically brightness and volume. There is a battery-saver mode, but we kept it toggled off for all testing, and we used the LDAT's luminance measurement to roughly equalize the Quest 3 to the Frame's default brightness (~50% on the Quest's slider), so these tests are brightness-normalized. For streaming, we set a 120Hz refresh rate through Steam Link and left the per-eye resolution slider at its default of 150% (2528x2704). 

Then we loaded into the first level and let the headset spin around in circles until it died. Note that this is a relatively abusive test, as it’s constant motion -- the reason we do this is to ensure the test is perfectly repeatable and does not rely on human factors to repeat.

Here’s the chart.

Running the Meta build of Moss locally, we lose direct control over refresh rate, resolution, and in-game graphics settings, but we know it's running 90Hz. The Quest 3 lasted 120.2 minutes locally and 118 minutes when streamed (but with a higher refresh rate of 120 Hz).

We did multiple rounds of tests for the Frame: first, a high-end streaming test with System Profile set to Performance, GPU maxed at 903MHz, all refresh rates (game and display) set to 120Hz, and resolution per eye still at default 150% (2644x2644 on the Frame). With those settings, the Frame died in 84.1 minutes, so the Quest 3 lasted 40% longer in as close to a like-for-like comparison as we can get. Running at 120Hz in Performance mode on battery is harsh and creates a stress test.

We then ran Moss locally on the Frame with the same high settings (other than resolution, which defaulted to 100% for 2160x2160). The headset died in 47 minutes flat. If you're playing games locally, you should switch to the 72Hz "Playtime" mode and preferably stick to lightweight games. We also tried running locally with a 90Hz refresh rate to match the Quest 3, and the Frame lasted 55.8 minutes.

Streaming in Playtime mode with the default resolution (still 2160x2160) and bandwidth cap (230Mbps at that time) resulted in a 111-minute runtime. We saw some tracking glitches towards the end of this run but overall, it was basically 111 minutes. 

With Playtime settings running Moss locally, the Frame lasted a more impressive 124.6 minutes, a lot longer than its best streaming time. Streaming can be fairly intensive on its own, but "Playtime" affects multiple settings, not just refresh rate, so it isn’t purely like-for-like on the settings.

The Quest broadly has better battery life than our Steam Frame. We have limited control over Quest 3 settings for local play, so can’t get a perfect match between them for that, but we can look at one more test.

Battery Life - Real-World Testing

All of that was pretty abusive, torture workload testing with the constant rotation on a lazy Susan. We needed some real-world testing, so we paid Mike to play video games for an entire day and stream Half Life: Alyx to the headsets. We wanted a reasonable middle-ground that would give both headsets a decent chance at beating each other, so we set the Frame to Playtime mode, locked both headsets to 72Hz, capped target bandwidth to 200Mbps, and kept resolution per eye at the default 150%.

We set the Frame on the turntable again as a control, and it lasted 123.6 minutes. With Mike wearing the Frame and playing his way through the story, it lasted 115.8 minutes -- close enough to make us confident in the turntable method, and actually slightly more intensive, although there are now human factors involved. It’s possible that something like the heat from being worn by a real user would reduce battery life. The Frame was in single-link on WiFi for Mike's test, but if that had any effect, it would be to lengthen battery life. Mike's session with the Quest 3 lasted 139.2 minutes, 20% beyond the Frame in a like-for-like scenario. This makes us pretty confident that the Quest 3 has better battery life, at least as we’ve tested it and for the Frame unit we have.

Steam Frame Charging Benchmarks

The charging tests are where we ran into issues with the Frame.

The first test involves waiting until the headsets shut themselves off due to low battery, then plugging them into a meter without turning them on. We also tested the Frame in several other conditions, including:

• Fully powered off
• Idle/standby charge
• Charging while gaming and below the fast charge range
• Charging while gaming and above the fast charge range
• Charging with the Quest charger
• Charging with other Steam Deck chargers

The reason is what you’re about to see.

Recharging the Quest 3 took 146 minutes with the stock 18W charger (9V at 2A), starting with a rate that oscillated between 12W and 17W, increasing to a point where it was charging at 13-19W. At 77 minutes in, the rate smoothed out and decreased in a curve down to under 2W before finishing. It'd make sense to unplug the headset and start using it again after approximately an hour and a half, since charging slows down significantly after that. There was no real difference in charging behavior for the Quest when using the Frame's higher-capacity charger, and 146 minutes is pretty close to the 2.3 hours (138 minutes) that Meta claims.

Valve sent us an official Frame/Deck PSU that maxes out at 45W. The Frame used just under 11W of that capacity to start with, which only increased up to a peak of 12W, then curved down to 3W before finishing. This process took 215 minutes, significantly more than the Quest, and the bulk of the charging took place over roughly the first two and a half hours (versus roughly one and a half on the Quest). We confirmed the charging behavior with a known-good 45W Steam Deck charger and tested again with a Quest charger.

Valve stated that "A dead battery takes around two hours to charge to full capacity. However, the charging is not linear. There is fast charging from 0 to 65%, and regular charging from 65% and up. Charging to 65% is expected to take less than 40 minutes." We didn’t see that.

With that in mind, we drained the battery below 65% and did some spot checks: manually shutting down the Frame (instead of letting it die) and then plugging it in drew 11W, putting it to sleep drew even less at 9W, and simply leaving the headset idle with the displays inactive drew 13-14W. We're confident in our equipment, which has been professionally calibrated, and methods, so that means either Valve is wrong, there's a bug with the charging behavior they don’t know about, or there's a defect with our review unit. We only have one and can only represent what they sent us.

We emailed Valve about all of this to try and compare notes, as we’re confident in our findings and methods and re-ran everything. Valve noted:

“From everything shared, it sounds like your experiment is valid. However, we are seeing different results on our end. Below are test results from last month, where we measured ~100 charge cycles across four Steam Frame headsets (with panels off and SteamVR stopped, simulating standby), and they took between 90 and 120 minutes to fully charge.

Valve continued, “One hypothesis our team has for this discrepancy is that it could be an effect of battery capacity ‘learning’. (In this battery tech, a few full charge/discharge cycles are required to have an accurate estimate of max cell capacity after initial power-on).” We did have a lot of full charge and discharge cycles so we’re not sure if this is the case. 

Valve’s reply continues, “Prior to going through such a cycle, we have seen Steam Frame hit >90% after the usual 90-120 min, but then creep up slowly after that until hitting termination at the new (higher) max capacity.”

The company continued, “It's also possible (of course) that there's been a regression in our charging software, and we'll try to double check that next week. But we are confident the hardware can do a full change (in standby/off) in the originally quoted times.”

Valve also shared this chart. The company noted: “[These] are plots of the raw data from our most recent charging tests. You can observe the inflection point where fast charging ends, after ~35 min (vs the 143 min you observed). And charging asymptotically ‘completes’ in the 90-120 min range. Also note that the battery charge levels in these graphs are not directly visible to the end user in the UX. Our charging protocol uses the industry stand approach to leave a bit of headroom at the top, to extend battery lifetime.”

We think Valve’s information and testing makes sense and Valve thinks our information and testing makes sense, but our numbers disagree. Without a second unit, we can’t verify, but we’re left with either a regression in the software (which Valve did leave as a possibility) or a defective unit, and it’s not the charger, because we tried several.

This isn’t the biggest deal in the world, but something screwy is going on. At this point, we’re very confident it’s not on us as we’ve spent an ungodly amount of time trying to verify this.

As for the controllers: For both the Quest 3 and the Frame, the controller battery life is so far beyond the headsets that we don't have an easy way to test them. We got through the entirety of this review and many, many headset charge cycles with the stock disposable AAs for both brands of controller. These were running for about two work weeks during work hours, so in our opinions it's a non-issue. Valve claims 40 hours, depending on the batteries used.

Steam Frame Software & User Experience

Time to get into the software and user experience.

The Frame uses SteamOS running on ARM, using the FEX emulation layer, with ARM-native builds of DXVK and VKD3D. It's a completely new frontier for Valve, so the most impressive thing we can say about it is that we didn't really notice. The Frame's SteamOS interface is familiar to anyone who's used a Steam Deck or Big Picture mode within the past couple of years. As of this writing, Valve has only confirmed 130 games to be Stand-Alone Verified/Great on Frame, so there are compatibility limitations; however, Valve is marketing this as a streaming headset, and all 2D games and many VR games should stream equally well. Stand Alone Verified is the program and Great on Frame is the library filter that shows those verified games.

Valve previously told us: "APKs can also be side-loadable just like any non-Steam applications on Steam Deck. We expect that VR APKs that don't leverage proprietary APIs to just work." We haven’t tested standalone APKs yet.

For VR games that have separate Windows and "Great On Frame" ARM builds, you can force download the Windows version by going into the per-game settings and overriding the compatibility mode to Proton. This may be worth doing for games like Moss, where the Frame build removes the graphics quality settings we see in Windows. This is already the method needed for downloading Windows builds on Linux, and it's messy. This needs to be integrated into the "Game Versions & Betas" tab, and devs need to actually make use of that tab.

Valve stated that streaming from the Steam Machine wouldn't be enabled until after launch, but we tried it anyway. Streaming 2D games works, streaming VR does not.

Desktop mode is still present. By default, it's the same Linux desktop environment as the Deck and Machine, just in VR. Being able to plug in a keyboard and work on a virtual desktop is neat, but, like the Deck, you'll need a hub if you want to plug in more than one peripheral at a time.

We downloaded a game on our Steam Machine, moved it to a microSD card, and then stuck it in the Frame and launched it, and it worked fine.

Like with inputs, Valve needs to streamline its UI. The Frame is different from past VR headsets in that it can stream from Steam while also RUNNING Steam, so there are two full sets of options to check. Just to run battery tests, we had to check a confusing web of host Steam settings, host SteamVR settings, host PC settings, host game settings, client Steam settings, client SteamVR settings, and headset settings. While streaming, there's a toggle in the SteamVR menu on the headset that switches from showing the local (Frame) SteamVR settings to the remote (PC) SteamVR settings. The audience that's kept up with VR over the past 10 years may be familiar with this; for newcomers, this will be a learning curve and we do thing the UI and software layout here is a mess.

The Frame wireless adapter shows up as a Realtek 8832CU WiFi 6 USB NIC in Windows and it can technically connect to home WiFi like any other wireless NIC, except when Steam is using the adapter.

Steam Frame Bugs

Moving on to the bugs:

On the Frame, the most common issue was that exiting games from the headset while streaming wouldn't close the game on the host-side. When running games locally, sometimes the launch screen would open, then go away after a few seconds without feedback about what went wrong.

Sometimes unplugging and reconnecting the wireless adapter required rebooting, but as long as we left it installed, it worked consistently. We also experimented with using the wireless adapter to connect to WiFi and then opening Steam, which caused a conflict that might be confusing. 

Several times when we tried to stream Open Brush, it worked normally on the host PC but turned into a nauseating, flickering smear on the headset side, but it looks like this was patched with a note from Valve: “Fixed video corruption in some cases while streaming VR.”

Camera Quality & Arcturus Vision Camera

We think Valve may have made a miscalculation with the cameras. In Valve's lineage of VR headsets, the passthrough cameras exist to tell you when you're about to smash your shins and not much else. The original Vive defaulted to showing ghostly outlines of objects, for example. With the Quest 3, the cameras are at the forefront: it acts as an AR headset by default, constantly showing you the real world with menus superimposed.

The Frame creates that same feeling when you first put it on, except the cameras suck. The framerate is low and choppy, which left some of us feeling sick when using it in passthrough mode, not helped by the monochromatic coloring (although that’s useful when lights are out). Likewise, real-world watch, phone screen, and text are illegible from more than a couple inches away, even for those of us with 20/20 vision. There's also distortion as objects come towards the front of the headset between the cameras. Monitors showing the headset viewport represent this well in a YouTube video, where you can see the same flickery, warblyness that the wearer is experiencing. Depending on how used to VR you are, this can range from sickening to simply just a minor annoyance. The Quest 3 has some of this as well, but whatever image-correction math Meta is doing in post-processing works a lot better than Valve's right now. It affects the exact area a couple feet in front of the Frame where you'd naturally hold an object to look at it.

The Quest 3 takes a big step towards making VR feel less isolating and burdensome by just letting you fucking see, while the Frame goes back to locking you into the VR Zone as long as you have it on.

We stuck a camera inside the lens and recorded at 120FPS to get an estimate of the framerate for the passthrough cameras. Moving objects appear at a sickening 30FPS, which is extremely obvious when moving your hands. 

The coolest feature of the stock cameras is that they can see and track in complete darkness using the infrared illumination from the headset, which also means they can pick up infrared light from security cameras, phones, and remote controls. As stupid as this sounds, it’d be a good device for hidden camera detection. 

Both the Frame and the Quest showed flickering in situations with mixed LED lighting.

According to Meta's product page, the Quest 3's dual RGB cameras give 18PPD (pixels per degree), which at 110 degrees horizontal and 96 vertical would calculate to 1920x1728 each. Meta's API page states that passthrough can do 60Hz (which we validated) at 1280x1280 per camera with 20-40ms of capture latency. All of those numbers are painfully low for reproducing reality, but we find it easier walking around in AR using the Quest 3 than in the Frame, and AR gaming is feasible on the Quest if that’s your thing.

Valve made us aware of two third-party accessories before launch: prescription lenses from Zenni Optical, and an RGB camera from Arcturus. The camera (technically two cameras) is $150, and besides being color, it significantly improves image resolution. 

The Arcturus camera made it substantially easier to look at a phone while in VR and solved some of Steve’s issues with Valve’s cameras, although it still has a relatively low framerate. It does feel a lot better, but it also draws on the power a little bit more.

The dual Arcturus cameras take advantage of the high-speed MIPI interface (4 lanes at 2GHz) on the headset and operate up to 72Hz at the maximum resolution of 2464x2464 each; the fine print notes that passthrough is "60 FPS by default; 72–144 FPS with special config." All processing is handled by the Frame: the cameras are just cameras, and the only function they replace is camera passthrough. Hotswapping works immediately with no configuration; the cameras interface with the OpenXR API implemented by Valve, so anything that uses the onboard cameras for passthrough (like AR games) should work exactly the same. The Arcturus cameras can't see in the dark like Valve’s can, although in the future it's possible Valve could update the headset to switch to the onboard cameras in low-light conditions.

We asked Valve about hand or gesture recognition as seen on other headsets like the Quest 3. The company told us that it is not a priority due to compute costs, but said they’re interested in it.

Steam Frame Controls

The external controls for the headset are similar to the Quest 3: volume, power, and IPD adjustment. The Frame's IPD adjustment is superior as the UI shows finer adjustments and the wheel can be locked. The Frame also has a separate "aux button" that brings up a cursor which can be used for basic menu navigation; the Quest 3 combines this with the power button (and also has controller-free hand gestures). 

The addition of a D-pad also gives an alternative to those who don't like the usual point-and-shoot VR menu navigation.

Unfortunately, VR tracking for the Steam Controller wasn't ready in time for this review. Valve has not finished the feature. We don't know how well it'll work in general, since the Steam Controller has a single tracking LED versus dozens on the dedicated VR controllers. The Controller must be connected via Bluetooth (or through the Frame's USB port) currently. If you're streaming, it may make more sense to have the Controller connected directly to the host PC via the Puck.

The finger tracking worked better than we expected (gradients for fingers, on/off for thumb). The biggest limitation here will be dev support: Finger tracking isn’t universal to VR hardware, so it will rely on developers to build games around it. The biggest positive is that the tracking makes it very convenient to flip someone off in VRChat, which is the most we'd ask of it.

The rebinding situation is insane, and maybe it has to be, but we don't recommend engaging with it unless you have to. Consider a situation where you're streaming a VR game from PC to headset, and the game is using an Oculus controller profile because your Steam VR controllers emulate Oculus controllers by default. Your options are to change bindings in the in-game settings, your headset per-game settings, your headset controller settings, your PC Steam client controller settings, PC Steam client game settings, or PC SteamVR controller settings, and it’s not always clear where to do that. You also might not know whether you want system-level bindings or "VRLink" bindings or app-specific bindings, and the buttons don't even match up to the Oculus anyway, and a "click" can mean a literal click or just pressing the button even if it doesn't click, and everything has a capacitive touch sensor, and you have to do half of this with a VR headset on. 

We were frequently greeted by a “Gamepad input is not yet supported while streaming VR” tooltip, but we were able to get a Steam Controller hardwired to the headset and working in VR at least once, and then when we went to reproduce it, we couldn’t easily figure out whatever combination of factors and/or mistakes allowed that.

If you DON'T open the controller menus, things generally just work. The biggest problem we foresee is with VR games that already expect a normal gamepad: for example, Star Wars: Squadrons wants gamepad inputs, but the Frame VR controllers show up as Oculus controllers by default when streaming, so even though they should be able to work by default, they don't and we don’t think it’s intuitive as to how to resolve this issue.

Steam Frame Conclusion

Ultimately, VR headsets are subjective. We had Mike and Patrick from the team join Steve to talk about their subjective opinions. We asked them to just give their thoughts from a pure consumer, non-reviewer standpoint.

Mike thought that the Frame felt a lot more comfortable as a whole. He noted how it didn’t get really hot and said that the weight and its distribution felt more comfortable compared to the Quest 3’s front-heavy design.

Considering Mike wears glasses, he also liked the optional top/middle strap on the Frame as it put less downward pressure on the front of his face. 

Mike also liked the optional hand straps for the Frame as it allowed him to “let go” of the controller while performing grabbing motions. 

In terms of performance tracking, Mike was impressed with the experience overall. He did note that, while gaming, the battery life lasted longer on the Quest 3. 

Mike summarized his experience with the Steam Frame as “unextraordinary in a good way” and didn’t cite any major flaws or issues while he was playing with the headset. 

While the Frame certainly cost a lot more than the Quest 3, he said, if he had to choose between the 2 headsets, the extra privacy would make the Frame more worth it to him. 

Next, we got Patrick’s subjective opinions on the Steam Frame, who also worked on testing the headset and helped write this review.

Patrick has a Quest 3 at home and thought the Steam Frame out of the box felt a lot more comfortable and lightweight. He said he didn’t like the optional top/middle strap, however.  

Looking at games on the Steam Frame vs the Quest 3, Patrick said that he didn’t know if he could tell the difference from a display standpoint, especially if both headsets are running the same game at 120Hz. 

In terms of tracking and controls, Patrick thought the Frame and Quest 3 also felt very similar. He did note, however, that there can be a lot of complications with streaming games to the Frame since the headset is running Steam, the PC that it’s streaming from runs Steam, players have to manage their PC settings, and the game settings on top of that. In addition, the controllers often mimic the Quest’s controllers and users can change their controller bindings, which can make things complicated.

All in all, Patrick said he’s happy that it’s out, but he is pessimistic about how successful it will be given its steep price. 

As for Steve’s subjective opinions, they boil-down pretty simply:

For hardware, Steve was impressed with the weight, comfort, and materials of the headset. He specifically said that the facial material felt like it breathed a bit better. Thermally, he said it felt cool overall and he liked the optional top/middle strap, so much so that he thought it should come included as he felt like it slid around too much otherwise. 

He also thought Valve’s privacy policy and privacy practices are excellent by comparison of what they could collect with the device. 

In terms of things Steve didn’t like, he absolutely hated the built-in cameras for the Frame. He found their low-speed and the weird re-projections in pass-through mode to be bouncy, sickening, and distorted. This is obviously a very opinionated thing to say, and some of users may disagree, but that was Steve’s experience with it. For some reference, Steve has perfect vision, nothing makes him dizzy, doesn't get any kind of vertigo, and is not sensitive to VR for long sessions, and yet, for some physiological reason, he has a very strong reaction to the Valve Frame camera. Just walking for him was bad, where he could notice that bounce. It was improved substantially with the optional head strap, which reduced the bounce. The more expensive color camera also helped as well. While in VR, however, the camera complaint goes away.

Pricing is the big discussion point. The value is bad, period, but it looks worse thanks to Meta subsidizing its pricing with the cost of your privacy. The common defense is to say that it’s nothing more than what a phone can collect -- but one, that might not be true. And two, to quote a viewer of ours when it comes to privacy and tracking: Just because we’ve been fucking it up doesn’t mean we have to keep fucking it up.

We wouldn’t recommend the Quest at all purely on privacy grounds, just like we wouldn’t recommend LG at all and purely on privacy grounds, but that doesn’t mean the Frame is worth it at over $1,000.

One of the biggest problems with VR is that there remains a chicken-and-egg problem where there has to be hardware adoption before developers will commit to making games. As a result, the categories of VR games for the Frame are something like: VRChat, racing game, spaceship game, 2D game that added a token VR mode, adult content, Meta exclusives you can't play, and Half-Life Alyx. There's some gold out there like Superhot VR, but we think this is a big part of the reason Valve has emphasized the ability to play and/or stream 2D games on the Frame.

For people super into VR, you already have your games and your reasons. For people new to it, if spending over $1,000, we’d advise seriously looking over the games first and making sure they’re the kind you like.