Skip to content
SOFTWARE DOWNLOADS & REVIEWS

Valve Details Strategy to Eliminate Stuttering in Windows Games Running on Arm Processors

Even the most powerful, high-end gaming PC can occasionally stutter under heavy loads. But imagine a processor that has to translate an entire game’s instructions on the fly while you are actively playing it, adding an enormous computational burden to every single frame rendered on screen.

That is one of the primary technical hurdles Valve faces in its fascinating, ongoing push to bring Windows games to Arm-based chips. As the company continues its efforts to break Steam free from traditional x86 gaming PCs, it is actively developing a clever plan to tackle translation-related stutter before players even hit the "Play" button.

"FEX code caching is vital for frame time consistency," Valve software engineer Pierre-Loup Griffais told PCWorld in an interview. "In the future, it will also be integrated with Steam’s Pre-Caching feature, so that cached code is available even prior to the first local run of an application."

This approach represents a sophisticated trick when attempting to make thousands of Windows games run smoothly on alternative processors they were never originally designed to support. Furthermore, it dovetails with a broader industry-wide push to stamp out one of PC gaming’s most infuriating and persistent problems: stuttering caused by real-time code translation and compilation.

Translating Games Without the Hiccups

Valve wants Steam to squash Arm gaming stutter before you even hit play

At the heart of Valve’s ambitious effort to liberate Steam from conventional hardware architectures is FEX, a crucial piece of software emulation. Because the vast majority of Windows games are built natively for x86 processors manufactured by companies like AMD and Intel, they speak a fundamentally different instruction language than Arm chips. To bridge this gap, FEX translates x86 CPU instructions into Arm-compatible instructions as the games run.

Fortunately, FEX does not have to shoulder the burden of translating every single component of a running application.

"When provided with a proper ARM execution environment, such as when integrated with Proton ARM, FEX only translates the CPU code of the application," Griffais explained.

Proton, Valve’s established Windows compatibility layer, is capable of switching to native Arm execution at the Win32 API boundary. Meanwhile, the underlying operating system and graphics drivers also run natively on the hardware. This leaves FEX with a more focused task, allowing it to concentrate specifically on translating the game’s own internal CPU instructions rather than wasting cycles on system-level tasks.

Griffais noted that the translation overhead introduced by this process is typically modest, particularly in games that are bottlenecked by the graphics processing unit rather than the CPU. However, modern gaming requires strict timing. A game running at a steady 60 frames per second gives the system roughly 16.7 milliseconds to finish rendering each individual frame. If FEX happens to encounter unfamiliar, untranslated code and needs to process it in the middle of active gameplay, that unexpected computational work can quickly chew through the frame’s strict time budget and manifest as an annoying hitch or micro-stutter on screen.

FEX’s internal caching system helps sidestep this exact problem. By saving translated code so that it can be reused later, the system avoids the need to re-compile the instructions every time a sequence occurs.

Valve wants Steam to squash Arm gaming stutter before you even hit play

In September, developers behind the FEX project introduced a persistent, on-disk JIT cache designed to preserve translated code between gaming sessions. Valve subsequently enabled this feature within Proton Experimental for Arm, with Griffais pointing out that it yields noticeable improvements in 1% low frame rates, especially during repeat runs of the same software.

These low frame rate metrics are vital because they expose the severe performance dips and stuttering moments that average frames-per-second statistics can easily hide. While average frame rates do not tell the whole story about a game’s smoothness, 1% low metrics are significantly more revealing when diagnosing micro-stutters.

Steam’s planned Pre-Caching integration takes this foundational concept even further. Instead of forcing a game to run locally first just to slowly build up its cache of translated instructions, Valve envisions a future where fully translated code is already prepared and waiting before the player launches the application for the very first time.

While Valve has not yet announced a specific timetable or published detailed performance benchmarks for this upcoming integration, the underlying premise is compelling. Eliminating translation-related hitches without requiring users to purchase a faster or more expensive processor is an advancement that the broader PC gaming community can readily get behind.

The PC Industry Has a Stutter Problem

Valve is certainly not the only major technology company attempting to shift expensive code compilation and translation work out of the middle of active gameplay.

Valve wants Steam to squash Arm gaming stutter before you even hit play

Earlier, industry efforts have also targeted similar pain points. Microsoft and Nvidia have been attacking PC gaming’s notorious shader stutter nightmare through Advanced Shader Delivery. Under this system, supported games can receive pre-compiled shaders tailored specifically for their hardware configuration rather than forcing the user’s local machine to compile everything from scratch upon first launch. Similarly, Intel has worked on a Graphics Shader Distribution Service designed to reduce compilation-related delays.

FEX approaches a fundamentally different problem, however. Its cache stores translated CPU code, whereas shader delivery systems deal strictly with GPU shaders. The two technologies are not interchangeable, and none of them serve as a magical silver bullet that will fix every single source of stuttering in modern games. Even with optimized translation, games can still hitch due to asset streaming bottlenecks, poor CPU utilization, or subpar developer optimization.

Nevertheless, the overarching philosophy guiding these initiatives is identical: get the expensive computational work out of the way before it has a chance to interrupt the player’s experience.

Steam is uniquely well-positioned to drive this kind of solution forward. Valve already operates the primary platform for digital PC game distribution and has spent years refining shader pre-caching infrastructure for SteamOS and the Steam Deck. Folding translated CPU code seamlessly into that existing ecosystem could go a long way toward making Arm-based gaming feel less like an experimental compatibility hurdle and more like a viable mainstream alternative.

Promising, But Not Magic

Despite its enormous potential, FEX’s persistent caching technology is still actively evolving. The initial release in September carried a few rough edges, including cache files that could grow unchecked on storage drives and stale entries that were not always properly invalidated when underlying game files changed.

Valve wants Steam to squash Arm gaming stutter before you even hit play

A subsequent FEX update tackled several of those growing pains by introducing a configurable cache limit that defaults to one gigabyte, implementing automatic cleanup routines for outdated caches after FEX updates are applied, and adding an optional memory buffer to speed up access for frequently used code.

There are also broader architectural limitations to consider. Griffais cautioned that Arm-based GPUs can differ substantially from the traditional desktop graphics hardware that PC games are primarily optimized for, which could potentially introduce unforeseen performance costs when handling advanced graphical features. Furthermore, heavy CPU translation workloads can still impact overall battery life and power consumption on portable devices, even in scenarios where the impact on average frame rates remains relatively modest.

Comprehensive independent testing examining how much FEX’s new cache improves gameplay across a wide variety of commercial Arm-based chips is still limited, and Valve’s planned Steam Pre-Caching integration has not yet been demonstrated in public environments.

Even so, Valve has spent nearly a decade helping the FEX project mature to a point where it can successfully run modern PC games with remarkably low translation overhead. The company’s current efforts are now focused on ensuring that whatever overhead remains is rendered virtually unnoticeable while players are actually enjoying their games.