Results

My pick is run 04: 40W, Low, frame generation off, for 65 fps average with a 48 fps 1% low and no reported stutter. At 30W, Low with VSync on still averaged 62 fps, but its lows were less consistent. Frame generation 2x lifts Medium to 73 fps average, but the logged 1% and 0.2% lows fall. Rows use different presets, VSync and targets, so not every pair is directly comparable, and these are single runs with recording running on the same handheld.

Source: E-Day Benchmarks workbook · Watch the video ↗

Weight and size

Neither handheld weighs what its press figures say. The Mini Pro weighed 1,288 g ready to play, against a 1,090 g hands-on figure. The Apex body weighed 814.6 g against 699 g, but that figure predates the liquid-cooled edition.

Power on the road

Cold boot means starting a fully powered-off handheld from that charger alone. Hot swap means pulling the stock adapter while the handheld is running and plugging in that charger, with no shutdown or brown-out.

Temperatures and noise at 55W on air

At the same 55W, the Mini Pro keeps its CPU cooler but runs the same GPU hotter and louder in the Wukong loop. I don't know why yet, and I'm not claiming a cause before Part 2. Noise was measured at 7 cm with a room floor of 45.8 to 45.9 dBA. Idle readings were too close to that floor to count, so they're left out. Neither handheld reports a fan RPM, only a fan percentage in OneXConsole.

Why is it called the X2 Mini Pro?

X is the large-screen branch of OneXPlayer's 3-in-1 idea, which started with the OneXPlayer 2 in 2023: 8.4 inches, detachable controllers and a keyboard. The X1 took it to 10.95 inches in 2024. 2 is the line's second generation, in 2026. Mini is the 8.8-inch size class, first used by the X1 Mini in June 2024. Pro has twice meant a better chip in the same body: OneXPlayer 2 to 2 Pro (6800U to 7840U, 2023) and X1 to X1 Pro (8840U to HX 370, 2024). This time it marks the platform: the X2 Mini is Intel, and the X2 Mini Pro is AMD.

Three wrinkles. The Pro went on sale on Indiegogo on 15 June, before the X2 and X2 Mini were unveiled on 23 June, and shipped in September. It was teased as the "X2 Mini" with the Ryzen AI Max+ 395 in April and gained "Pro" for the global launch by 28 May, after which "X2 Mini" went to the Intel Arc G3 Extreme model. And it was announced with the 395 (16 cores) but shipped with the 388 (8 cores), with the same 40-CU GPU and the same TDP. The OneXPlayer 3 isn't an X3: it succeeds the OneXPlayer 2 and launched beside the X2.

V: confirmed against OneXPlayer, Notebookcheck, Liliputing, VideoCardz, Tom's Hardware, Retro Handhelds and TweakTown. Press: launch month as reported in the press. U: not independently verified.

Sources: X2 Mini Pro vs Apex physical workbook, v0.4, and the Part 1 opening deck, v0.7 · Watch the video ↗

On my unit the stock cable was the limit. These were short runs, and multi-hour runs on each pairing come next. The review's four did-not-finish results still stand, because that's how the unit behaved out of the box.

Watch the follow-up: AYANEO Next 2 85W Drops Solved ↗ · Video page →

Source: G3E Family Feud workbook, v1.24 · Watch the video ↗ · Video page →

A pass means the game reached real gameplay, controls, audio and graphics were usable through an opening section, and a save loaded correctly after a restart. A menu, an intro or an FPS counter alone didn't count. Frame rates are readings I saw or reported during play, not CapFrameX averages.

The most useful discovery: check CPU Boost. OneXConsole showed 40W while Turbo, which OneXPlayer says turns on CPU Boost, was off. After turning Turbo on, God of War III held a steady 30 fps in combat and passed save/load, and Gran Turismo 5 finished an Arcade race, bought a car and still had it after a reload. Emulator builds and game settings also changed along the way, so this isn't a controlled measurement of Turbo on its own. It does mean the Turbo state belongs in the test record next to the power limit.

God of War III settings for the passing run: Turbo on at 40W; PPU and SPU decoders LLVM; SPU threads Auto; SPU block size Mega; XFloat Approximate; Vulkan at 720p, 100%; shader precision Ultra; Write Color Buffers on; patches Disable MLAA v3.0 and Disable SPU MLAA v1.5. Dropping to 50% resolution made it look worse without helping.

Uncharted 3 and The Last of Us both look far better with the in-game MLAA disabled and GPU Lighting enabled, and both reach gameplay, but they kept freezing. Uncharted 3's 29.04 fps screenshot was captured at the moment it froze, which is why an FPS counter on its own proves nothing.

These are first results from the sections I tested, not full playthroughs, official RPCS3 compatibility ratings or a guide to Android performance. The test procedure carries over to the AYN Odin 3; the Windows settings don't.

Sources: my RPCS3 test report, logs and screenshots · OneXConsole guide (Turbo) ↗ · RPCS3 wiki: God of War III ↗, The Last of Us ↗, Uncharted 3 ↗

Native 1440p: no upscaling, no frame generation

Mortal Kombat 1 shows why averages aren't enough. Both machines average about 60 fps against the game's cap, but the RTX 4060 holds a 46.6 fps 1% low while the Apex drops to 20.8, with a 14.1 fps 0.2% low. The RTX setup holds the cap, and the Apex stutters.

Ray tracing and upscaling at 1440p: DLSS on the RTX, FSR on the Apex, frame generation off

The pattern: with light or engine-level ray tracing the Apex keeps up or leads. With path-tracing-class RT, the RTX 4060 pulls away. In Wukong at matched settings it's 70% faster. The upscalers differ by platform, so these rows compare each machine at its best, not like for like.

Synthetic, creator and AI tests

Cinebench GPU: Cinebench 2024 won't run on the Radeon 8060S, and Cinebench 2026 crashed in AMD's HIP runtime three times out of three, so the Apex has no score. DaVinci Resolve: Studio 21 on both. The Apex crashed the whole system twice before three clean runs, then won the overall score. Llama 2 13B: the model doesn't fit in the RTX 4060's 8 GB of VRAM, and the Apex's 96 GB of unified memory has no such wall.

Noise at 50 cm: the Apex on the Frost Bay dock measured 61.0 dBA idle and 69.2 dBA under sustained load. The RTX setup measured 64.0 dBA under sustained load, and at idle it was at or below 48.7 dBA, too close to the room's 46.7 dBA floor to measure. The handheld and dock are the louder system at every state.

Source: Apex vs RTX 4060 benchmark workbook, v12.38 · Watch the video ↗