What was actually tested
The benchmark details come from GameGPU, which relayed the owner's post, and from TweakTown's coverage of it.
- Hardware: An ROG Ally X was paired with an external Radeon RX 9070 XT with 16 GB of video memory. A standard eGPU dock was used, with no proprietary XG Mobile connector, because the handheld has a full USB4 port with Thunderbolt support.
- Game and display: The built-in Cyberpunk 2077 benchmark, version 2.31, ran on an ultrawide 3440x1440 monitor.
- Settings: Ray-traced reflections and shadows were on. Ray-traced lighting was set to Ultra, with no path tracing. FSR scaling was on Auto with sharpness 0.50, plus FSR 3.1 frame generation.
- Software: Radeon Software 26.5.1 on Windows 11 Home.
- Result: GameGPU reports an average of 150.77 FPS, a minimum of 101.67 FPS and a maximum of 192.57 FPS. The run lasted about 64 seconds and processed 9,688 frames.
Microsoft's product page lists the ROG Xbox Ally X with an AMD Ryzen AI Z2 Extreme processor, 24 GB of LPDDR5X-8000 memory and a 7-inch 1080p 120 Hz display. That means 3440x1440 is an external-monitor resolution, not something the handheld's own screen shows.
Why "151 FPS" needs an asterisk
The number is a displayed frame rate with both FSR upscaling and FSR 3.1 frame generation on. Frame generation inserts extra frames between conventionally rendered ones. As TweakTown points out, the figure therefore says little about the game's underlying render rate or input responsiveness.
The run also leaves several things unanswered:
- No result with frame generation disabled was published.
- Path tracing was off, so this says nothing about the game's heaviest mode.
- It was one short run, so there is nothing on sustained performance, thermals, power draw or noise.
- It covers one game, docked, so it says nothing about undocked play or other titles.
- The owner's claim of good SteamVR results has no benchmark, headset or measurement behind it. Treat it as anecdote.
GameGPU says the system stayed stable in demanding Night City scenes despite the limited bandwidth of the external connection. That is the site's characterization. No frame-time data was published to back it up.
The USB4 question
USB4 offers far less bandwidth than a desktop PCIe slot. TweakTown is explicit that this test does not measure how much performance the card loses over USB4. TweakTown's own RX 9070 XT review figures were 54 FPS at native 1440p with the RT Ultra preset, and 92 FPS with FSR 4. These came from a different setup and different settings, so they are not a like-for-like comparison.
Separate testing helps here. Notebookcheck, reporting on a YouTuber's (ETA Prime) test, covered an ROG Ally X with an RX 9070 XT in an Oculink dock. This is a different connection from the USB4 setup above. At 1440p Ultra with FSR 4 on Quality, the eGPU typically delivered 60-70 FPS, but frame rates often dipped below 50 FPS. Enabling frame generation nearly doubled the average to around 120, but with lows of 99 FPS and peaks of 150 FPS, which the YouTuber described as "all over the place."
That is a different dock, different settings and a different test, so it does not validate or contradict the Reddit run. It does show that frame generation can inflate averages while frame pacing stays uneven. It also suggests the 101-to-192 FPS spread in this benchmark is worth reading as more than a single average.
Practical takeaways
- An eGPU can lift a handheld well above its built-in graphics. The tested combination hit triple-digit displayed frame rates with ray tracing on at ultrawide resolution.
- It is not a handheld anymore. You need a dock, a power supply, a desktop GPU and an external display. That adds cost and bulk compared with the handheld alone.
- Judge it on more than averages. Look for 1% lows, frame-time graphs, and results with frame generation off. This run offers none of them.
- Expect variation. Dock, cable, driver and game settings can all change results, as the Notebookcheck comparison suggests.
Bottom line
This is a fun, credible-looking enthusiast demo of what a Windows 11 handheld can do when docked. It is not an independent lab test. It does not show native 3440x1440 rendering at 151 FPS, and it does not quantify the USB4 penalty. Read it as a proof of concept, and wait for tests that include frame generation off, 1% lows and more games.