Linux Gaming with ProtonDB: What the Numbers Mean in 2026

The old joke about Linux gaming was that it had always been "almost there." A platform permanently one year away from maturity. A hobby operating system for people who preferred troubleshooting Wine prefixes over actually finishing their game library. That characterization is now outdated by a significant margin, and the data to prove it has been quietly accumulating on a community website called ProtonDB.

Boiling Steam's analysis of ProtonDB's crowd-sourced reports, published in October 2025, found that approximately 89.7% of Windows titles now at least launch on Linux systems. That number represents a legitimate inflection point, not a rounding error. But here is the more important finding underneath it: the remaining 10.3% is not primarily a technical problem. It is a policy problem. The infrastructure to run those games largely exists. The decisions not to enable it belong to publishers, anti-cheat vendors, and studio legal departments — not to the engineering teams who built Proton. Understanding that distinction changes how you think about the entire landscape of Linux gaming in 2026. This article builds toward that argument from the ground up.

What ProtonDB Actually Is

ProtonDB is not a Valve product. That distinction matters. The site was built and is maintained by a single developer, Buck DeFore, originally as a spare-time project he funded himself. On the project's Patreon page, DeFore has described the effort as something that at times required stepping away from paid employment to sustain it. The site started life as Steam Play Community Reports (SPCR) before rebranding, and it now holds over 150,000 reports covering more than 20,000 games. All of that data is publicly exported to a GitHub repository and will never be paywalled.

The mechanism is straightforward: players run Windows games on Linux using Proton — Valve's compatibility layer built into Steam — and then report back to ProtonDB. They document which version of Proton they used, their hardware configuration, what operating system distribution they're on, whether the game needed any tweaks, and what those tweaks were. The site aggregates these reports and assigns a single summary rating from a tiered scale.

The Rating Tiers

The ProtonDB rating system has six categories, and understanding them is essential to reading the data correctly:

• Native — The game has an official Linux build. No Proton involved.
• Platinum — Runs flawlessly out of the box. No tweaks needed, no user intervention required.
• Gold — Runs well, but requires minor adjustments to get there. A launch option, a specific Proton version, a config file tweak.
• Silver — Playable, but noticeably imperfect. May have performance issues, crashes in certain areas, or missing features.
• Bronze — Runs, but with significant problems that will affect the experience.
• Borked — Does not launch, or is completely unplayable even after troubleshooting.

One critical nuance: ratings can become stale. A game rated Borked from two years ago may run perfectly today after a Proton update. A Platinum title can become Borked after a publisher deploys a new anti-cheat update. ProtonDB shows the date of the most recent reports alongside each rating, which is the first thing an experienced Linux gamer checks before trusting any summary score.

How Proton Works Under the Hood

Proton is not magic. It is an engineering project with a specific, describable architecture. According to Wikipedia's documentation on the software, Proton is a compatibility layer developed by Valve in cooperation with developers from CodeWeavers. It is, at its foundation, a patched and extended version of Wine

Proton was first released on August 21, 2018. Its initial announcement included a list of 27 certified games, among them Doom (2016), Quake, and Final Fantasy VI. The goal from the start was not just compatibility in principle but parity in practice: these titles were expected to perform like their native Windows counterparts without requiring any user intervention.

The Graphics Translation Stack

The critical performance challenge for any Windows-to-Linux compatibility layer is graphics. Windows games are written against DirectX — Microsoft's proprietary graphics API. Linux doesn't have DirectX. It has Vulkan

DXVK handles DirectX 8, 9, 10, and 11. It translates Direct3D calls into Vulkan API calls at runtime. This is significant because Vulkan provides lower-level access to GPU hardware, which means DXVK doesn't just make older games compatible — in many cases it makes them run faster on Linux than they do through the Windows graphics stack. Proton's Wikipedia entry notes that D9VK, which handled DirectX 9 separately, was merged into DXVK in December 2019.

VKD3D-Proton handles DirectX 12. It is a fork of the original VKD3D project, developed separately to prioritize gaming performance over broad compatibility. The project's GitHub documentation is explicit about this tradeoff: performance and game compatibility take deliberate priority over support for older drivers, with the project aggressively exploiting modern Vulkan extensions to achieve both. In November 2025, VKD3D-Proton 3.0 shipped with a completely rewritten DXBC shader backend and added support for AMD FSR4 upscaling — with Red Dead Redemption 2 specifically called out as now running correctly in DirectX 12 mode.

Beyond Graphics: The Other Components

DXVK and VKD3D-Proton address graphics, but a Windows game is more than a render loop. Proton also includes:

• Steam Linux Runtime — A container environment that provides a consistent set of libraries regardless of which Linux distribution the user is running. This is a major reason why Proton "just works" across Arch, Fedora, Ubuntu, and others without per-distro maintenance.
• esync / fsync — Synchronization primitives that reduce CPU overhead from multithreaded games. Upstream Wine has not accepted these patches; they remain Proton-specific. This is one reason Proton generally outperforms vanilla Wine.
• Steam Input — Controller remapping handled at the Steam layer, meaning most gamepads work transparently.
• Media Foundation patches — Handles video cutscene playback, which was a common failure point for otherwise-compatible games.

GamingOnLinux editor Liam Dawe, reflecting on seven years of Proton in August 2025, captured the shift concisely: Linux gaming had gone from occasional indie titles and ported scraps to tens of thousands of games available with a click — a range that, in his words, left players "spoilt for choice."

— Liam Dawe, GamingOnLinux, August 2025

GE-Proton: The Community Fork

Proton ships in two official variants: stable release builds (currently the Proton 10.x series, with Proton 10.0-4 released in early 2026) and Proton Experimental, a rolling build that often contains fixes for newly released games before they land in a stable version. But there is a third option that the Linux gaming community treats as essential: GE-Proton, maintained by a developer known as GloriousEggroll.

GE-Proton is a custom build that tracks bleeding-edge versions of every Proton component simultaneously — Wine, DXVK, DXVK-NVAPI, and VKD3D-Proton — while also incorporating patches and game-specific fixes that haven't yet been accepted upstream. As of March 2026, GE-Proton 10-33 is current, adding WiVRn-based VR support for non-Steam titles, improved umu.exe behavior to help third-party launchers, fixes for Star Citizen launch errors, and FSR 4.0.3 and 4.1.0 DLL updates for games using AMD upscaling. GE-Proton builds also added aarch64/ARM compilation support in early 2026 — experimental, but a meaningful signal that the compatibility layer is already looking past x86.

The tradeoff is stability. Because GE-Proton pulls from multiple bleeding-edge branches simultaneously, a new build can occasionally introduce regressions for specific titles. The community recommendation is to use official Proton for games in a stable, working state and reach for GE-Proton when a game is broken on official builds or when you need fixes that haven't landed in an official release yet.

Worth acknowledging honestly: that recommendation and actual community practice frequently diverge. Many Linux gamers run GE-Proton as their default for everything, reasoning that the broader fix coverage outweighs the occasional regression risk. The right answer probably depends on your tolerance for debugging. If a surprise regression at the wrong moment would be genuinely disruptive — mid-playthrough of a long game, for example — the conservative approach holds. If you'd rather chase the latest fixes and occasionally need to roll back, the community default of "always GE" is defensible.

The Numbers: What 89.7% Actually Covers

The headline figure of 89.7% compatibility, reported by Boiling Steam from ProtonDB data in October 2025, requires context to be useful. Several layers of nuance sit underneath it.

The Platinum Surge

The most telling data point is not the overall compatibility rate but the Platinum rate trajectory. According to the Boiling Steam analysis, Platinum-rated games — titles that install, run, and save without any user intervention whatsoever — made up 42% of new releases tracked in October 2025. One year prior, that figure was 29%. That is not a marginal improvement. It means the burden of proof has effectively inverted. A few years ago, Linux support for a new Windows release was an afterthought you'd patch in with tweaks and configuration work. Now, for a well-structured game without aggressive DRM or anti-cheat, Linux compatibility at launch is increasingly the default outcome rather than a deliberate achievement — a consequence of Proton's maturity rather than any developer action.

The Borked Floor

At the other end of the scale, the share of games that refuse to launch at all — the Borked category — has fallen to approximately 3.8%. A small subset of that figure consists of games that have deliberately coded Wine and Proton detection into their executables and exit immediately when they detect the compatibility layer. The Boiling Steam analysis cited March of Giants as an explicit example of a title that detects Wine or Proton and deliberately terminates.

The persistence of this 3.8% floor is worth sitting with. It is largely a policy problem, not a technical one — the same framing that applies to the broader 10.3% ceiling. The technical infrastructure to run these games exists. The decision not to enable it belongs to publishers. This is not a gap Valve can patch, and it is not a gap the Linux engineering community can close on their own. It is a commercial and contractual wall, dressed up as a compatibility problem.

The Steam Deck Effect

The Steam Deck's commercial trajectory is inseparable from these numbers. Market research firm IDC estimated Valve had shipped approximately 3.7 to 4 million Steam Decks by early 2025 — the lion's share of a combined ~6 million total handheld gaming PCs sold across all vendors since 2022. Valve's Deck Verified program — which tests and certifies games for SteamOS specifically — shows 21,694 games rated at least "Playable" as of November 2025. The Steam Deck did not create Proton, but it created something Proton had never had before: hardware-level accountability. Every incompatible game was a customer complaint on a device Valve manufactured, sold, and shipped. That accountability loop is the specific reason Proton development velocity changed, not just the reason adoption grew. Desktop Linux gaming had existed for years without generating this kind of pressure. Sold hardware, with receipts, changed the calculus entirely.

Steam's hardware surveys recorded Linux at 2.89% of active users by July 2025, compared to 1.44% in late 2022 — a near-doubling in under three years. That figure climbed further through the rest of the year: Valve's revised December 2025 survey put Linux at 3.58%, an all-time high for the platform, as reported by Phoronix. January 2026 settled at 3.38%. February 2026 then showed an apparent drop to 2.23% — but Phoronix and GamingOnLinux both flagged that this coincided with a simultaneous 30% surge in Simplified Chinese language users in the same dataset. The mechanism is straightforward: that segment is overwhelmingly Windows, so a sudden large influx of Chinese users mathematically compresses the Linux percentage without any actual change in the absolute number of Linux users. It is a denominator problem, not a numerator one. This pattern has historically produced temporary downward distortions in Linux percentages, and the English-only Steam survey view still showed Linux at approximately 8.3% for the same month. Valve has previously revised survey figures when these regional anomalies have been identified.

The Anti-Cheat Problem: A Policy Wall, Not a Technical One

The remaining hard obstacles to Linux gaming are not primarily technical. They are political and contractual. Anti-cheat middleware sits at the center of this.

Easy Anti-Cheat and BattlEye

Both Easy Anti-Cheat (EAC), developed by Epic Games, and BattlEye have formally announced support for Proton on Linux. Both middleware solutions are technically capable of running in the Proton environment. The catch — and it is a significant one — is that enabling this support is not automatic. Each individual game developer must opt in by enabling the Linux support flag in their EAC or BattlEye configuration. Valve cannot force this. The middleware vendors cannot force this. It is a decision that belongs to each studio, and many studios have chosen not to make it.

According to Engadget's November 2025 reporting on Are We Anti-Cheat Yet data, of the 1,136 games tracked that require anti-cheat software, 682 do not function under Linux or Proton — meaning more than half of all anti-cheat-protected titles remain inaccessible on SteamOS. (Engadget, November 2025)

What Would It Actually Take to Fix This

The anti-cheat problem gets discussed endlessly in the Linux gaming community, but the proposed solutions rarely go deeper than "Valve should pressure developers more" or "if Linux market share grows, developers will follow." Those observations are true but incomplete. Here are the actual levers — most of which are currently in motion, to varying degrees:

TPM-based hardware attestation. Valve has already enabled Trusted Platform Module (TPM) APIs on the Steam Machine specifically to give anti-cheat developers a path forward. TPM-based attestation allows a device to cryptographically prove to a game server that it is running unmodified, verified software — a different security model than the Windows kernel-module approach, but one with real precedent. Google's Play Integrity API and Apple's App Attest use similar mechanisms on mobile. The key insight is that attestation doesn't require a ring-0 driver. It requires a hardware root of trust, which SteamOS on Steam Machine hardware provides. Valve developer Pierre-Loup Griffais confirmed on the Friends Per Second podcast in late 2025 that Valve has been working with anti-cheat vendors on exactly this approach, enabling the infrastructure that would allow developers to build solutions that don't depend on a Linux kernel module. The limitation is that this approach is almost certainly Steam Machine-specific, not applicable to general Linux desktops, because it depends on controlled hardware.

Server-side behavioral analysis. The industry's overreliance on client-side kernel inspection as the primary anti-cheat vector is itself a design problem. Server-side behavioral detection — statistical modeling of player performance, movement trajectory anomalies, aim-assist detection, and session-level pattern analysis — does not require any special client-side access. It works identically on Linux and Windows because it doesn't touch the client at all. Games that have invested heavily in server-side analysis, like Valve's own Counter-Strike 2, have demonstrated that this model can be effective without locking out non-Windows players. The objection from publishers is usually that server-side detection is slower and catches fewer cheaters per unit time than a kernel driver doing live memory inspection. That is probably true. The question is whether the marginal detection improvement from a kernel driver is worth permanently blocking an entire class of users — and whether that calculation changes as SteamOS market share grows.

EA Javelin's Linux development signal. In early March 2026, a job listing at EA explicitly called out Linux and Proton support as a target for the Javelin anti-cheat system — the in-house solution EA is developing to eventually replace Easy Anti-Cheat across its portfolio. GamingOnLinux reported on the listing in March 2026. This is the first time a major publisher has publicly signaled active engineering work on Linux-compatible anti-cheat as a staffed priority rather than a stated intention. EA's portfolio — Battlefield, Apex Legends, EA Sports — includes some of the highest-profile Linux holdouts. If Javelin ships with Proton support, the knock-on effect for the competitive landscape could be significant. The caveat is that the listing may refer to server-side infrastructure for ARM-based hardware rather than client-side Proton support — community analysis of the listing's wording is genuinely ambiguous. But the signal itself represents a qualitative shift in how one major publisher is thinking about the problem.

The dual-boot and GPU passthrough path. For users whose libraries are split between anti-cheat-blocked titles and everything else, the practical solution today is not a software fix but a hardware one. Dual-booting Linux and Windows on separate drives remains the standard recommendation for competitive multiplayer players who also want a Linux primary environment. A more sophisticated approach — GPU passthrough via VFIO, running Windows in a virtual machine with dedicated GPU access — allows Windows games to run with near-native performance while the host system remains Linux. This approach is technically demanding to configure and not something most users will attempt, but it eliminates the incompatibility entirely at the cost of hardware complexity. For users with two discrete GPUs, or a discrete GPU plus integrated graphics, the setup is more practical. It is not a solution the broader ecosystem can rely on, but it correctly frames the problem as solvable in hardware terms even when the software wall holds.

The Proton Wine-Detection bypass problem. Some games — the Boiling Steam analysis cited March of Giants as an example — actively detect Wine or Proton and terminate on detection. This is distinct from the anti-cheat problem: it is a developer choice to refuse the compatibility layer, not a technical limitation of the layer itself. The community response has been tools like proton-launch-options that mask Proton's presence from detection routines, though this approach occupies ethically and legally contested ground. Valve's own position is that Proton should be transparent about its nature and that games should not be blocked on the basis of the underlying compatibility layer alone. The resolution here is ultimately a developer policy decision, but the commercial pressure of SteamOS device sales is the most likely mechanism for changing those decisions over time.

What won't work. It is worth being precise about the approaches that sound viable but aren't. A generic Linux kernel-mode driver for anti-cheat — a single driver multiple vendors could use — faces maintenance problems that are qualitatively different from Windows: the Linux kernel API is not stable across versions, unsigned third-party modules face Secure Boot barriers, and the open-source nature of the kernel makes the driver's internals visible to the cheaters it is trying to stop. Security researcher Samuel Tulach's analysis of the problem, widely cited in the Linux gaming community, concludes that even a dedicated Linux kernel driver for anti-cheat would face fundamental circumvention problems that the closed Windows kernel makes harder to exploit. The approaches that can work — attestation, behavioral analysis, developer opt-in for EAC/BattlEye — are already available. The block is not that the solutions don't exist. It is that publishers have not yet judged the user base large enough to justify building them.

Are We Anti-Cheat Yet is an essential companion resource to ProtonDB for anyone whose game library skews toward competitive multiplayer. It tracks the current compatibility status of every anti-cheat-protected game against Linux and Proton.

High-Profile Refusals

The most prominent cases are instructive. Fortnite, powered by Epic's own EAC, has never been playable on Steam Deck or Linux. Valorant, which uses Riot's Vanguard anti-cheat operating at kernel level in Windows, remains blocked. Rockstar disabled Grand Theft Auto V on Linux despite BattlEye offering support that the developer chose not to enable. Destiny 2 actively bans Proton users.

The Apex Legends situation illustrates how quickly the picture can change in the wrong direction. EA's battle royale ran on Linux through Proton for years. On October 31, 2024, an update silently removed the Linux Easy Anti-Cheat binary — effectively deactivating support with no advance notice to players. EA's developer account on X (formerly Twitter) stated that the decision would "meaningfully reduce instances of cheating," even though the anti-cheat infrastructure technically supports Linux when the appropriate binary is included. Players who had been using the game lost access overnight. The game's Linux compatibility status went from functional to Borked in a single patch cycle.

The Windows 10 end-of-life milestone, reached October 14, 2025, introduced a new dimension to this conversation. With approximately 240 million devices facing the choice between upgrading hardware to meet Windows 11's TPM 2.0 requirements or finding an alternative, Linux gaming adoption gained a new pool of prospective users. For single-player gamers, the Proton library makes the migration highly practical. For competitive multiplayer players whose preferred games remain blocked, the situation is considerably more complicated — and there is an irony worth naming. The users most technically capable of making this migration, and most motivated to care about platform principles, are often exactly the community-oriented online gamers whose libraries sit behind the kernel-level anti-cheat walls. The people best positioned to benefit from switching are partially blocked from the benefit. For now.

Using ProtonDB Effectively

ProtonDB is most useful when you treat it as a field guide rather than an answer machine. A single rating number tells you the community consensus at a moment in time. The individual reports tell you the actual story.

Before You Buy

Checking ProtonDB before purchasing a game you intend to play on Linux is good practice. Navigate to protondb.com, search for the title, and look at three things: the overall rating, the date of the most recent reports, and whether reports from users with hardware similar to yours exist. A Platinum rating from reports all dated 2022 is less reliable than a Gold rating from someone who played last month.

Reading Reports for Fixes

When a game is rated Gold or Silver, the individual user reports are often the most valuable resource available anywhere. Linux gamers have a strong culture of documenting exactly what they did to get a game working. A report might specify that adding PROTON_USE_WINED3D=1 %command% to the launch options resolves a specific crash, or that Proton Experimental works where the stable version doesn't, or that a particular DXVK version introduced a regression that the latest build fixed.

# Use OpenGL renderer instead of Vulkan (resolves some D3D11 issues)
PROTON_USE_WINED3D=1 %command%

# Enable Gamemode for CPU/GPU performance optimization
gamemoderun %command%

# Force large address aware flag (helps some 32-bit games)
WINE_LARGE_ADDRESS_AWARE=1 %command%

# Disable NVAPI (fixes some NVIDIA-specific crashes)
PROTON_HIDE_NVIDIA_GPU=1 %command%

# Combine multiple options
gamemoderun DXVK_ASYNC=1 %command%

Launch options in Steam are set per-game: right-click the title in your library, select Properties, and enter the options in the Launch Options field. The %command% token represents the game's actual executable and must be present for arguments to work correctly.

Checking Hardware Match

ProtonDB reports include the reporter's GPU, CPU, and distribution. AMD and NVIDIA users can have meaningfully different experiences with the same game. AMD's open-source Mesa driver stack integrates natively with Vulkan, and AMD GPUs on Linux sometimes outperform their Windows equivalents — Counter-Strike 2 has been documented showing approximately a 9.6% performance gain on AMD hardware under Linux compared to Windows. NVIDIA's proprietary driver has improved significantly since 2020, including better Wayland integration, but some overhead remains for specific workloads. If a game runs well for AMD reporters and poorly for NVIDIA reporters, that is data worth having.

Reading Reports for Signal, Not Just Rating

The summary rating is the headline. The individual reports are the story. Here is how experienced Linux gamers actually process ProtonDB data — a reading method that matters more than knowing which rating tiers exist:

Sort by recent, not by rating. The default view surfaces highly-voted reports, which can be years old. Filter to the last three to six months. A game with a Platinum average from 2022 reports that has three recent Borked entries has almost certainly regressed — a publisher update, a new DRM layer, or an anti-cheat change happened between then and now.

Look for the word "update" in negative reports. When a game breaks on Linux, the reports that document it almost always say something like "broke after the [Season X] update" or "stopped working after the October 24th patch." This language pattern is the single most reliable signal that a previously working game has regressed and the cause is publisher-side, not configuration-side. There is usually nothing you can do about it until Proton catches up.

Read the fix reports, not just the ratings. A Silver-rated game with twenty reports that all include the same three-word fix — say, PROTON_USE_WINED3D=1 — is functionally a Gold game for anyone willing to spend fifteen seconds applying that option. The rating aggregation system doesn't capture this. The reports do. A Bronze rating with one fix that resolves 90% of the issues is a much better gaming experience than a Silver rating where the problems are genuinely random.

Cross-reference with Proton version. Reports from eighteen months ago that reference Proton 7.x or 8.x are not necessarily predictive of behavior on Proton 10.x. The compatibility layer changes substantially across major versions. Old Borked reports for a game often become irrelevant once the underlying regression in Wine or DXVK that caused them gets patched. Old Platinum reports can become less reliable if a new DRM system has since been added to the game.

Check the "Native" reports if they exist. ProtonDB tracks native Linux builds alongside Proton reports. A game with a native build that has worse ratings than its Proton reports is a real phenomenon — port quality varies, and some studio Linux ports have shipped with older rendering paths or broken audio that the Proton-translated Windows version handles correctly. The "native is always better" assumption does not always hold. There is an uncomfortable implication here worth naming: for some titles, the best thing the developer could have done for their Linux players would have been to not port the game at all and let Proton handle it. A well-maintained Windows build running through Proton 10.x can outperform a neglected Linux port that hasn't been touched since 2019.

Contributing Back

The site's value comes entirely from users submitting reports. Creating a free account takes two minutes, and submitting a report after playing takes less than five. Reporting a working game as Platinum helps other users skip unnecessary troubleshooting. Reporting a specific fix that resolved a Silver-rated game's worst issue elevates the collective knowledge available to everyone who plays that title on Linux after you.

Performance: Closing the Gap

The persistent myth that Linux gaming is inherently slower than Windows has been repeatedly undermined by benchmarking data. On well-optimized titles, the performance delta between Linux and Windows on identical hardware is typically within 5-10%. For some games, particularly older DirectX 9, 10, and 11 titles where DXVK's Vulkan translation outperforms the original Windows rendering path, Linux can be the faster platform.

The distribution of performance outcomes roughly follows the anti-cheat distribution: single-player and story-driven games, which represent the bulk of titles that aren't blocked by kernel-level protection, tend to run well. Online competitive titles — particularly those from large publishers who have not opted into EAC or BattlEye Linux support — are the primary source of the remaining performance and compatibility gaps.

GPU Driver Considerations

AMD users on Linux benefit from a fully open-source driver stack (Mesa + RADV for Vulkan) that is actively co-developed by AMD engineers and the open-source community. Driver updates ship through standard distribution package managers. Intel's open-source ANV Vulkan driver follows a similar model.

NVIDIA users should install the proprietary NVIDIA driver rather than the open-source Nouveau driver for gaming workloads. The proprietary driver provides full Vulkan support, better performance, and is required for DLSS. NVIDIA has improved Linux support considerably, including kernel module signing for Secure Boot and Wayland compositing improvements, but the driver must be managed separately from the distribution's main package tree on many distributions.

Beyond Steam: Non-Steam Games and Heroic Launcher

ProtonDB's data is weighted toward Steam titles because the Steam API makes it straightforward to associate reports with specific game IDs. But Proton and its underlying components are not Steam-exclusive.

Lutris is a game manager that supports GOG, Epic Games, Battle.net, Ubisoft Connect, Origin, and Amazon Prime Gaming alongside Steam. It provides community-maintained install scripts for titles not available on Steam, using Wine and Proton under the hood with a GUI that abstracts the configuration work.

Heroic Games Launcher provides a similar interface specifically for Epic Games and GOG titles, using either Wine-GE or Proton as its compatibility backend. For users who left Windows partly because of Epic's anti-consumer practices but still want to play titles in their Epic library, Heroic provides a practical path.

One important filesystem note: games installed on NTFS-formatted partitions — the filesystem Windows uses — frequently cause issues under Linux. The combination of NTFS case-sensitivity handling and permission semantics does not map cleanly to how Wine and Proton expect to interact with game files. Moving games to an ext4, Btrfs, or other native Linux filesystem partition resolves a large class of otherwise-puzzling compatibility problems.

A longer-term development worth tracking: the FEX emulator project enables x86 and x86-64 translation on ARM hardware. Valve has been investing in FEX as part of preparing Proton for ARM-based devices. GE-Proton 10-30 included early experimental aarch64 compilation support — still explicitly work-in-progress, with 32-bit game compatibility described as "hit or miss." But the direction is clear: the Proton compatibility layer is being extended toward ARM, which has implications both for future Valve hardware and for the growing category of ARM-based handheld PCs.

What This Actually Means for You

Linux gaming in 2026 is not a universal solution. It is a highly capable platform with a well-defined ceiling — and the ceiling is not where you might expect it to be.

For single-player and co-op gaming, the practical compatibility rate across a typical Steam library is high enough that many users report not needing Windows at all. Major titles — Elden Ring, Cyberpunk 2077, Baldur's Gate 3, Red Dead Redemption 2, The Witcher 3 — run through Proton with performance comparable to Windows. ProtonDB's Platinum and Gold tier coverage of the top 100 Steam games by player count exceeds 90%.

For competitive online gaming, the picture depends almost entirely on which games matter to you. If your primary titles use kernel-level anti-cheat (Valorant, Fortnite) or if publishers have specifically removed Linux support (Destiny 2, Apex Legends as of October 31, 2024), Linux is not a viable primary gaming platform without dual-booting. This is a policy limitation, not a technical one. And it is the same policy limitation that has been at the center of this article's argument from the first paragraph.

One emerging factor worth watching: Valve has been actively pushing SteamOS beyond the Steam Deck itself. The Lenovo Legion Go S launched with a SteamOS edition in early 2025, the first time Valve-powered hardware appeared in retail stores like Best Buy. A new Steam Machine and Steam Frame were announced in November 2025. In February 2026, Valve disclosed that memory and storage component shortages had forced a delay of their target release window from Q1 to the first half of 2026. Valve has also been explicit about what it expects the Steam Machine to accomplish on anti-cheat: speaking to Eurogamer, the company acknowledged that developer opt-in is still required, but argued that the incentive to enable anti-cheat would be higher for the Steam Machine than for the Steam Deck — precisely because living-room gaming skews toward multiplayer. Valve stated directly that it hopes the Machine's launch will shift the broader equation around anti-cheat support. (Eurogamer / Engadget, November 2025)

ProtonDB is the tool that makes the difference navigable. It converts the theoretical question of whether Linux gaming works into a specific, verifiable answer for specific titles on specific hardware. Before buying a game, before reporting a bug, before spending two hours troubleshooting a crash that someone else already solved and documented — the first stop should be protondb.com.

That sentiment was echoed by members of the GamingOnLinux community in August 2025, with a number of long-time Linux users noting they had stopped booting into Windows entirely — not as a statement of principle, but simply because they no longer had a practical reason to.

The infrastructure built between 2018 and today — Proton, DXVK, VKD3D-Proton, GE-Proton, the Steam Linux Runtime, and the ProtonDB knowledge base — represents one of the more remarkable open-source engineering achievements of the last decade. But the meaningful question for 2026 is not whether Proton works. It works. The meaningful question is whether the remaining holdouts face any structural pressure to change. Right now, the honest answer is: not enough. Valve can incentivize but cannot compel. Linux market share is growing but hasn't reached the threshold where blocking Linux support represents a meaningful lost sale for a live-service shooter. The 89.7% ceiling is not an engineering limit. It is a negotiation that hasn't happened yet — and the outcome of that negotiation will be determined by how many SteamOS devices Valve ships, not by how good Proton gets.

Sources

• [1] Boiling Steam / Tux Machines — "Nearly 90% of Windows Games Now Run on GNU/Linux," October 2025. tuxmachines.org
• [2] Wikipedia — "Proton (software)." en.wikipedia.org
• [3] GamingOnLinux — "7 Years Later, Valve's Proton Has Been an Incredible Game-Changer for Linux," August 2025. gamingonlinux.com
• [4] Valve Software — Proton GitHub Repository (VKD3D-Proton 3.0 release notes). github.com/ValveSoftware/Proton
• [5] HansKristian-Work — VKD3D-Proton GitHub Repository. github.com/HansKristian-Work/vkd3d-proton
• [6] TechReviewer — "Linux Hits 90% Windows Game Compatibility Milestone," October 2025. techreviewer.com
• [7] Engadget — "For the Steam Machine to Change PC Gaming, Valve Must Solve Linux's Anti-Cheat Problem," November 2025. engadget.com
• [8] Slashdot — "Nearly 90% of Windows Games Now Run on Linux, Latest Data Shows," October 2025. linux.slashdot.org
• [9] Are We Anti-Cheat Yet — Community database of anti-cheat compatibility with Linux/Proton. areweanticheatyet.com
• [10] GloriousEggroll — GE-Proton Release History. GE-Proton 10-33 released March 18, 2026; GE-Proton 10-30 released February 10, 2026. github.com/GloriousEggroll/proton-ge-custom
• [11] Buck DeFore (ProtonDB creator) — Patreon page describing the site's history. patreon.com/protondb
• [12] Yahoo / ATT (sourced from VKD3D-Proton 3.0 coverage) — "Valve's Proton Brings Better DirectX 12 Support and FSR4 to Linux," November 2025.
• [13] Linux All Day — "Linux Gaming: The Essential Guide to Maximizing Performance with Steam & Proton," October 2025. linuxallday.com
• [14] Phoronix — "Revised Steam Survey For December 2025 Puts Linux Gaming Marketshare At 3.58%," January 6, 2026. phoronix.com
• [15] The Verge / IDC — "Three Years On, the Steam Deck Has 'Quite Possibly' Sold 4 Million Units," February 2025. Steam Deck estimated at 3.7–4 million units per IDC market research commissioned by The Verge. theverge.com
• [16] PlayApex on X (formerly Twitter) — Developer statement on Apex Legends Linux EAC removal, October 31, 2024. x.com/PlayApex
• [17] GamingOnLinux — "Steam Survey for January 2026 Shows a Small Drop for Linux and macOS," February 2, 2026. gamingonlinux.com
• [18] Phoronix — "Steam Survey Results Published For February 2026," March 2, 2026. Linux share at 2.23% attributed to Simplified Chinese user surge anomaly. phoronix.com
• [19] GamingOnLinux — "Steam Survey for February 2026 Shows a Big Swing to Simplified Chinese," March 2026. English-only split shows Linux at ~8.29%. gamingonlinux.com
• [20] Engadget — "Valve's Steam Machine Launches in 2026: Everything We Know So Far," March 2026. Includes Valve's direct statement on anti-cheat incentives for the Steam Machine. engadget.com
• [21] GamingOnLinux — "Valve Released Proton 10.0-4 Bringing More Game Compatibility to SteamOS / Linux," January 27, 2026. gamingonlinux.com
• [22] Linux Compatible — "GE-Proton10-33 Released," March 18, 2026. Documents WiVRn VR support, Star Citizen fixes, FSR 4.1.0 upscaler, and umu.exe improvements. linuxcompatible.org
• [23] GamingOnLinux — "EA Javelin Anticheat Job Listing Mentions Future Support for Linux and Proton," March 2026. gamingonlinux.com
• [24] Tweaktown / Friends Per Second podcast — "Valve Working with Devs to Get Anti-Cheat on Steam Machine," November 2025. Pierre-Loup Griffais on TPM APIs and developer collaboration. tweaktown.com
• [25] Samuel Tulach — "The Issue of Anti-Cheat on Linux." Security researcher analysis of kernel-level anti-cheat circumvention on Linux. tulach.cc